Doc. no. N1853=05-0113
Date: 2005-08-26
Project: Programming Language C++
Reply to: Howard Hinnant <howard.hinnant@gmail.com>

C++ Standard Library Defect Report List (Revision R38)

Reference ISO/IEC IS 14882:1998(E)

Also see:

This document contains only library issues which have been closed by the Library Working Group (LWG) after being found to be defects in the standard. That is, issues which have a status of DR, TC, or RR. See the Library Closed Issues List for issues closed as non-defects. See the Library Active Issues List for active issues and more information. The introductory material in that document also applies to this document.

Revision History

Defect Reports


1. C library linkage editing oversight

Section: 17.4.2.2 [lib.using.linkage]  Status: TC  Submitter: Beman Dawes  Date: 16 Nov 1997

The change specified in the proposed resolution below did not make it into the Standard. This change was accepted in principle at the London meeting, and the exact wording below was accepted at the Morristown meeting.

Proposed resolution:

Change 17.4.2.2 [lib.using.linkage] paragraph 2 from:

It is unspecified whether a name from the Standard C library declared with external linkage has either extern "C" or extern "C++" linkage.

to:

Whether a name from the Standard C library declared with external linkage has extern "C" or extern "C++" linkage is implementation defined. It is recommended that an implementation use extern "C++" linkage for this purpose.


3. Atexit registration during atexit() call is not described

Section: 18.3 [lib.support.start.term]  Status: TC  Submitter: Steve Clamage  Date: 12 Dec 1997

We appear not to have covered all the possibilities of exit processing with respect to atexit registration.

Example 1: (C and C++)

    #include <stdlib.h>
    void f1() { }
    void f2() { atexit(f1); }
    
    int main()
    {
        atexit(f2); // the only use of f2
        return 0; // for C compatibility
    }

At program exit, f2 gets called due to its registration in main. Running f2 causes f1 to be newly registered during the exit processing. Is this a valid program? If so, what are its semantics?

Interestingly, neither the C standard, nor the C++ draft standard nor the forthcoming C9X Committee Draft says directly whether you can register a function with atexit during exit processing.

All 3 standards say that functions are run in reverse order of their registration. Since f1 is registered last, it ought to be run first, but by the time it is registered, it is too late to be first.

If the program is valid, the standards are self-contradictory about its semantics.

Example 2: (C++ only)

    
    void F() { static T t; } // type T has a destructor

    int main()
    {
        atexit(F); // the only use of F
    }

Function F registered with atexit has a local static variable t, and F is called for the first time during exit processing. A local static object is initialized the first time control flow passes through its definition, and all static objects are destroyed during exit processing. Is the code valid? If so, what are its semantics?

Section 18.3 "Start and termination" says that if a function F is registered with atexit before a static object t is initialized, F will not be called until after t's destructor completes.

In example 2, function F is registered with atexit before its local static object O could possibly be initialized. On that basis, it must not be called by exit processing until after O's destructor completes. But the destructor cannot be run until after F is called, since otherwise the object could not be constructed in the first place.

If the program is valid, the standard is self-contradictory about its semantics.

I plan to submit Example 1 as a public comment on the C9X CD, with a recommendation that the results be undefined. (Alternative: make it unspecified. I don't think it is worthwhile to specify the case where f1 itself registers additional functions, each of which registers still more functions.)

I think we should resolve the situation in the whatever way the C committee decides.

For Example 2, I recommend we declare the results undefined.

[See reflector message lib-6500 for further discussion.]

Proposed resolution:

Change section 18.3/8 from:

First, objects with static storage duration are destroyed and functions registered by calling atexit are called. Objects with static storage duration are destroyed in the reverse order of the completion of their constructor. (Automatic objects are not destroyed as a result of calling exit().) Functions registered with atexit are called in the reverse order of their registration. A function registered with atexit before an object obj1 of static storage duration is initialized will not be called until obj1's destruction has completed. A function registered with atexit after an object obj2 of static storage duration is initialized will be called before obj2's destruction starts.

to:

First, objects with static storage duration are destroyed and functions registered by calling atexit are called. Non-local objects with static storage duration are destroyed in the reverse order of the completion of their constructor. (Automatic objects are not destroyed as a result of calling exit().) Functions registered with atexit are called in the reverse order of their registration, except that a function is called after any previously registered functions that had already been called at the time it was registered. A function registered with atexit before a non-local object obj1 of static storage duration is initialized will not be called until obj1's destruction has completed. A function registered with atexit after a non-local object obj2 of static storage duration is initialized will be called before obj2's destruction starts. A local static object obj3 is destroyed at the same time it would be if a function calling the obj3 destructor were registered with atexit at the completion of the obj3 constructor.

Rationale:

See 99-0039/N1215, October 22, 1999, by Stephen D. Clamage for the analysis supporting to the proposed resolution.


5. String::compare specification questionable

Section: 21.3.6.8 [lib.string::compare]  Status: TC  Submitter: Jack Reeves  Date: 11 Dec 1997

At the very end of the basic_string class definition is the signature: int compare(size_type pos1, size_type n1, const charT* s, size_type n2 = npos) const; In the following text this is defined as: returns basic_string<charT,traits,Allocator>(*this,pos1,n1).compare( basic_string<charT,traits,Allocator>(s,n2);

Since the constructor basic_string(const charT* s, size_type n, const Allocator& a = Allocator()) clearly requires that s != NULL and n < npos and further states that it throws length_error if n == npos, it appears the compare() signature above should always throw length error if invoked like so: str.compare(1, str.size()-1, s); where 's' is some null terminated character array.

This appears to be a typo since the obvious intent is to allow either the call above or something like: str.compare(1, str.size()-1, s, strlen(s)-1);

This would imply that what was really intended was two signatures int compare(size_type pos1, size_type n1, const charT* s) const int compare(size_type pos1, size_type n1, const charT* s, size_type n2) const; each defined in terms of the corresponding constructor.

Proposed resolution:

Replace the compare signature in 21.3 [lib.basic.string] (at the very end of the basic_string synopsis) which reads:

int compare(size_type pos1, size_type n1,
            const charT* s, size_type n2 = npos) const;

with:

int compare(size_type pos1, size_type n1,
            const charT* s) const;
int compare(size_type pos1, size_type n1,
            const charT* s, size_type n2) const;

Replace the portion of 21.3.6.8 [lib.string::compare] paragraphs 5 and 6 which read:

int compare(size_type pos, size_type n1,
            charT * s, size_type n2 = npos) const;
Returns:
basic_string<charT,traits,Allocator>(*this, pos, n1).compare(
             basic_string<charT,traits,Allocator>( s, n2))

with:

int compare(size_type pos, size_type n1,
            const charT * s) const;
Returns:
basic_string<charT,traits,Allocator>(*this, pos, n1).compare(
             basic_string<charT,traits,Allocator>( s ))

int compare(size_type pos, size_type n1,
            const charT * s, size_type n2) const;
Returns:
basic_string<charT,traits,Allocator>(*this, pos, n1).compare(
             basic_string<charT,traits,Allocator>( s, n2))

Editors please note that in addition to splitting the signature, the third argument becomes const, matching the existing synopsis.

Rationale:

While the LWG dislikes adding signatures, this is a clear defect in the Standard which must be fixed.  The same problem was also identified in issues 7 (item 5) and 87.


7. String clause minor problems

Section: 21 [lib.strings]  Status: TC  Submitter: Matt Austern  Date: 15 Dec 1997

(1) In 21.3.5.4 [lib.string::insert], the description of template <class InputIterator> insert(iterator, InputIterator, InputIterator) makes no sense. It refers to a member function that doesn't exist. It also talks about the return value of a void function.

(2) Several versions of basic_string::replace don't appear in the class synopsis.

(3) basic_string::push_back appears in the synopsis, but is never described elsewhere. In the synopsis its argument is const charT, which doesn't makes much sense; it should probably be charT, or possible const charT&.

(4) basic_string::pop_back is missing.

(5) int compare(size_type pos, size_type n1, charT* s, size_type n2 = npos) make no sense. First, it's const charT* in the synopsis and charT* in the description. Second, given what it says in RETURNS, leaving out the final argument will always result in an exception getting thrown. This is paragraphs 5 and 6 of 21.3.6.8 [lib.string::compare]

(6) In table 37, in section 21.1.1 [lib.char.traits.require], there's a note for X::move(s, p, n). It says "Copies correctly even where p is in [s, s+n)". This is correct as far as it goes, but it doesn't go far enough; it should also guarantee that the copy is correct even where s in in [p, p+n). These are two orthogonal guarantees, and neither one follows from the other. Both guarantees are necessary if X::move is supposed to have the same sort of semantics as memmove (which was clearly the intent), and both guarantees are necessary if X::move is actually supposed to be useful.

Proposed resolution:

ITEM 1: In 21.3.5.4 [lib.string::insert], change paragraph 16 to

    EFFECTS: Equivalent to insert(p - begin(), basic_string(first, last)).

ITEM 2:  Not a defect; the Standard is clear.. There are ten versions of replace() in the synopsis, and ten versions in 21.3.5.6 [lib.string::replace].

ITEM 3: Change the declaration of push_back in the string synopsis (21.3, [lib.basic.string]) from:

     void push_back(const charT)

to

     void push_back(charT)

Add the following text immediately after 21.3.5.2 [lib.string::append], paragraph 10.

    void basic_string::push_back(charT c);
    EFFECTS: Equivalent to append(static_cast<size_type>(1), c);

ITEM 4: Not a defect. The omission appears to have been deliberate.

ITEM 5: Duplicate; see issue 5 (and 87).

ITEM 6: In table 37, Replace:

    "Copies correctly even where p is in [s, s+n)."

with:

     "Copies correctly even where the ranges [p, p+n) and [s, s+n) overlap."


8. Locale::global lacks guarantee

Section: 22.1.1.5 [lib.locale.statics]  Status: TC  Submitter: Matt Austern  Date: 24 Dec 1997

It appears there's an important guarantee missing from clause 22. We're told that invoking locale::global(L) sets the C locale if L has a name. However, we're not told whether or not invoking setlocale(s) sets the global C++ locale.

The intent, I think, is that it should not, but I can't find any such words anywhere.

Proposed resolution:

Add a sentence at the end of 22.1.1.5 [lib.locale.statics], paragraph 2: 

No library function other than locale::global() shall affect the value returned by locale().


9. Operator new(0) calls should not yield the same pointer

Section: 18.4.1 [lib.new.delete]  Status: TC  Submitter: Steve Clamage  Date: 4 Jan 1998

Scott Meyers, in a comp.std.c++ posting: I just noticed that section 3.7.3.1 of CD2 seems to allow for the possibility that all calls to operator new(0) yield the same pointer, an implementation technique specifically prohibited by ARM 5.3.3.Was this prohibition really lifted? Does the FDIS agree with CD2 in the regard? [Issues list maintainer's note: the IS is the same.]

Proposed resolution:

Change the last paragraph of 3.7.3 from:

Any allocation and/or deallocation functions defined in a C++ program shall conform to the semantics specified in 3.7.3.1 and 3.7.3.2.

to:

Any allocation and/or deallocation functions defined in a C++ program, including the default versions in the library, shall conform to the semantics specified in 3.7.3.1 and 3.7.3.2.

Change 3.7.3.1/2, next-to-last sentence, from :

If the size of the space requested is zero, the value returned shall not be a null pointer value (4.10).

to:

Even if the size of the space requested is zero, the request can fail. If the request succeeds, the value returned shall be a non-null pointer value (4.10) p0 different from any previously returned value p1, unless that value p1 was since passed to an operator delete.

5.3.4/7 currently reads:

When the value of the expression in a direct-new-declarator is zero, the allocation function is called to allocate an array with no elements. The pointer returned by the new-expression is non-null. [Note: If the library allocation function is called, the pointer returned is distinct from the pointer to any other object.]

Retain the first sentence, and delete the remainder.

18.4.1 currently has no text. Add the following:

Except where otherwise specified, the provisions of 3.7.3 apply to the library versions of operator new and operator delete.

To 18.4.1.3, add the following text:

The provisions of 3.7.3 do not apply to these reserved placement forms of operator new and operator delete.

Rationale:

See 99-0040/N1216, October 22, 1999, by Stephen D. Clamage for the analysis supporting to the proposed resolution.


11. Bitset minor problems

Section: 23.3.5 [lib.template.bitset]  Status: TC  Submitter: Matt Austern  Date: 22 Jan 1998

(1) bitset<>::operator[] is mentioned in the class synopsis (23.3.5), but it is not documented in 23.3.5.2.

(2) The class synopsis only gives a single signature for bitset<>::operator[], reference operator[](size_t pos). This doesn't make much sense. It ought to be overloaded on const. reference operator[](size_t pos); bool operator[](size_t pos) const.

(3) Bitset's stream input function (23.3.5.3) ought to skip all whitespace before trying to extract 0s and 1s. The standard doesn't explicitly say that, though. This should go in the Effects clause.

Proposed resolution:

ITEMS 1 AND 2:

In the bitset synopsis (23.3.5 [lib.template.bitset]), replace the member function

    reference operator[](size_t pos);

with the two member functions

    bool operator[](size_t pos) const;
    reference operator[](size_t pos);

Add the following text at the end of 23.3.5.2 [lib.bitset.members], immediately after paragraph 45:

bool operator[](size_t pos) const;
Requires: pos is valid
Throws: nothing
Returns: test(pos)

bitset<N>::reference operator[](size_t pos);
Requires: pos is valid
Throws: nothing
Returns: An object of type bitset<N>::reference such that (*this)[pos] == this->test(pos), and such that (*this)[pos] = val is equivalent to this->set(pos, val);

Rationale:

The LWG believes Item 3 is not a defect. "Formatted input" implies the desired semantics. See 27.6.1.2 [lib.istream.formatted].


13. Eos refuses to die

Section: 27.6.1.2.3 [lib.istream::extractors]  Status: TC  Submitter: William M. Miller  Date: 3 Mar 1998

In 27.6.1.2.3, there is a reference to "eos", which is the only one in the whole draft (at least using Acrobat search), so it's undefined.

Proposed resolution:

In 27.6.1.2.3 [lib.istream::extractors], replace "eos" with "charT()"


14. Locale::combine should be const

Section: 22.1.1.3 [lib.locale.members]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

locale::combine is the only member function of locale (other than constructors and destructor) that is not const. There is no reason for it not to be const, and good reasons why it should have been const. Furthermore, leaving it non-const conflicts with 22.1.1 paragraph 6: "An instance of a locale is immutable."

History: this member function originally was a constructor. it happened that the interface it specified had no corresponding language syntax, so it was changed to a member function. As constructors are never const, there was no "const" in the interface which was transformed into member "combine". It should have been added at that time, but the omission was not noticed.

Proposed resolution:

In 22.1.1 [lib.locale] and also in 22.1.1.3 [lib.locale.members], add "const" to the declaration of member combine:

template <class Facet> locale combine(const locale& other) const; 

15. Locale::name requirement inconsistent

Section: 22.1.1.3 [lib.locale.members]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

locale::name() is described as returning a string that can be passed to a locale constructor, but there is no matching constructor.

Proposed resolution:

In 22.1.1.3 [lib.locale.members], paragraph 5, replace "locale(name())" with "locale(name().c_str())".


16. Bad ctype_byname<char> decl

Section: 22.2.1.4 [lib.locale.ctype.byname.special]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

The new virtual members ctype_byname<char>::do_widen and do_narrow did not get edited in properly. Instead, the member do_widen appears four times, with wrong argument lists.

Proposed resolution:

The correct declarations for the overloaded members do_narrow and do_widen should be copied from 22.2.1.3 [lib.facet.ctype.special].


17. Bad bool parsing

Section: 22.2.2.1.2 [lib.facet.num.get.virtuals]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

This section describes the process of parsing a text boolean value from the input stream. It does not say it recognizes either of the sequences "true" or "false" and returns the corresponding bool value; instead, it says it recognizes only one of those sequences, and chooses which according to the received value of a reference argument intended for returning the result, and reports an error if the other sequence is found. (!) Furthermore, it claims to get the names from the ctype<> facet rather than the numpunct<> facet, and it examines the "boolalpha" flag wrongly; it doesn't define the value "loc"; and finally, it computes wrongly whether to use numeric or "alpha" parsing.

I believe the correct algorithm is "as if":

  // in, err, val, and str are arguments.
  err = 0;
  const numpunct<charT>& np = use_facet<numpunct<charT> >(str.getloc());
  const string_type t = np.truename(), f = np.falsename();
  bool tm = true, fm = true;
  size_t pos = 0;
  while (tm && pos < t.size() || fm && pos < f.size()) {
    if (in == end) { err = str.eofbit; }
    bool matched = false;
    if (tm && pos < t.size()) {
      if (!err && t[pos] == *in) matched = true;
      else tm = false;
    }
    if (fm && pos < f.size()) {
      if (!err && f[pos] == *in) matched = true;
      else fm = false;
    }
    if (matched) { ++in; ++pos; }
    if (pos > t.size()) tm = false;
    if (pos > f.size()) fm = false;
  }
  if (tm == fm || pos == 0) { err |= str.failbit; }
  else                      { val = tm; }
  return in;

Notice this works reasonably when the candidate strings are both empty, or equal, or when one is a substring of the other. The proposed text below captures the logic of the code above.

Proposed resolution:

In 22.2.2.1.2 [lib.facet.num.get.virtuals], in the first line of paragraph 14, change "&&" to "&".

Then, replace paragraphs 15 and 16 as follows:

Otherwise target sequences are determined "as if" by calling the members falsename() and truename() of the facet obtained by use_facet<numpunct<charT> >(str.getloc()). Successive characters in the range [in,end) (see [lib.sequence.reqmts]) are obtained and matched against corresponding positions in the target sequences only as necessary to identify a unique match. The input iterator in is compared to end only when necessary to obtain a character. If and only if a target sequence is uniquely matched, val is set to the corresponding value.

The in iterator is always left pointing one position beyond the last character successfully matched. If val is set, then err is set to str.goodbit; or to str.eofbit if, when seeking another character to match, it is found that (in==end). If val is not set, then err is set to str.failbit; or to (str.failbit|str.eofbit)if the reason for the failure was that (in==end). [Example: for targets true:"a" and false:"abb", the input sequence "a" yields val==true and err==str.eofbit; the input sequence "abc" yields err=str.failbit, with in ending at the 'c' element. For targets true:"1" and false:"0", the input sequence "1" yields val==true and err=str.goodbit. For empty targets (""), any input sequence yields err==str.failbit. --end example]


18. Get(...bool&) omitted

Section: 22.2.2.1.1 [lib.facet.num.get.members]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

In the list of num_get<> non-virtual members on page 22-23, the member that parses bool values was omitted from the list of definitions of non-virtual members, though it is listed in the class definition and the corresponding virtual is listed everywhere appropriate.

Proposed resolution:

Add at the beginning of 22.2.2.1.1 [lib.facet.num.get.members] another get member for bool&, copied from the entry in 22.2.2.1 [lib.locale.num.get].


19. "Noconv" definition too vague

Section: 22.2.1.5.2 [lib.locale.codecvt.virtuals]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

In the definitions of codecvt<>::do_out and do_in, they are specified to return noconv if "no conversion is needed". This definition is too vague, and does not say normatively what is done with the buffers.

Proposed resolution:

Change the entry for noconv in the table under paragraph 4 in section 22.2.1.5.2 [lib.locale.codecvt.virtuals] to read:

noconv: internT and externT are the same type, and input sequence is identical to converted sequence.

Change the Note in paragraph 2 to normative text as follows:

If returns noconv, internT and externT are the same type and the converted sequence is identical to the input sequence [from,from_next). to_next is set equal to to, the value of state is unchanged, and there are no changes to the values in [to, to_limit).


20. Thousands_sep returns wrong type

Section: 22.2.3.1.2 [lib.facet.numpunct.virtuals]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

The synopsis for numpunct<>::do_thousands_sep, and the definition of numpunct<>::thousands_sep which calls it, specify that it returns a value of type char_type. Here it is erroneously described as returning a "string_type".

Proposed resolution:

In 22.2.3.1.2 [lib.facet.numpunct.virtuals], above paragraph 2, change "string_type" to "char_type".


21. Codecvt_byname<> instantiations

Section: 22.1.1.1.1 [lib.locale.category]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

In the second table in the section, captioned "Required instantiations", the instantiations for codecvt_byname<> have been omitted. These are necessary to allow users to construct a locale by name from facets.

Proposed resolution:

Add in 22.1.1.1.1 [lib.locale.category] to the table captioned "Required instantiations", in the category "ctype" the lines

codecvt_byname<char,char,mbstate_t>,
codecvt_byname<wchar_t,char,mbstate_t> 

22. Member open vs. flags

Section: 27.8.1.7 [lib.ifstream.members]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

The description of basic_istream<>::open leaves unanswered questions about how it responds to or changes flags in the error status for the stream. A strict reading indicates that it ignores the bits and does not change them, which confuses users who do not expect eofbit and failbit to remain set after a successful open. There are three reasonable resolutions: 1) status quo 2) fail if fail(), ignore eofbit 3) clear failbit and eofbit on call to open().

Proposed resolution:

In 27.8.1.7 [lib.ifstream.members] paragraph 3, and in 27.8.1.10 [lib.ofstream.members] paragraph 3, under open() effects, add a footnote:

A successful open does not change the error state.

Rationale:

This may seem surprising to some users, but it's just an instance of a general rule: error flags are never cleared by the implementation. The only way error flags are are ever cleared is if the user explicitly clears them by hand.

The LWG believed that preserving this general rule was important enough so that an exception shouldn't be made just for this one case. The resolution of this issue clarifies what the LWG believes to have been the original intent.


24. "do_convert" doesn't exist

Section: 22.2.1.5.2 [lib.locale.codecvt.virtuals]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

The description of codecvt<>::do_out and do_in mentions a symbol "do_convert" which is not defined in the standard. This is a leftover from an edit, and should be "do_in and do_out".

Proposed resolution:

In 22.2.1.5 [lib.locale.codecvt], paragraph 3, change "do_convert" to "do_in or do_out". Also, in 22.2.1.5.2 [lib.locale.codecvt.virtuals], change "do_convert()" to "do_in or do_out".


25. String operator<< uses width() value wrong

Section: 21.3.7.9 [lib.string.io]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

In the description of operator<< applied to strings, the standard says that uses the smaller of os.width() and str.size(), to pad "as described in stage 3" elsewhere; but this is inconsistent, as this allows no possibility of space for padding.

Proposed resolution:

Change 21.3.7.9 [lib.string.io] paragraph 4 from:

    "... where n is the smaller of os.width() and str.size(); ..."

to:

    "... where n is the larger of os.width() and str.size(); ..."


26. Bad sentry example

Section: 27.6.1.1.2 [lib.istream::sentry]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

In paragraph 6, the code in the example:

  template <class charT, class traits = char_traits<charT> >
  basic_istream<charT,traits>::sentry(
           basic_istream<charT,traits>& is, bool noskipws = false) {
      ...
      int_type c;
      typedef ctype<charT> ctype_type;
      const ctype_type& ctype = use_facet<ctype_type>(is.getloc());
      while ((c = is.rdbuf()->snextc()) != traits::eof()) {
        if (ctype.is(ctype.space,c)==0) {
          is.rdbuf()->sputbackc (c);
          break;
        }
      }
      ...
   }

fails to demonstrate correct use of the facilities described. In particular, it fails to use traits operators, and specifies incorrect semantics. (E.g. it specifies skipping over the first character in the sequence without examining it.)

Proposed resolution:

Remove the example above from 27.6.1.1.2 [lib.istream::sentry] paragraph 6.

Rationale:

The originally proposed replacement code for the example was not correct. The LWG tried in Kona and again in Tokyo to correct it without success. In Tokyo, an implementor reported that actual working code ran over one page in length and was quite complicated. The LWG decided that it would be counter-productive to include such a lengthy example, which might well still contain errors.


27. String::erase(range) yields wrong iterator

Section: 21.3.5.5 [lib.string::erase]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

The string::erase(iterator first, iterator last) is specified to return an element one place beyond the next element after the last one erased. E.g. for the string "abcde", erasing the range ['b'..'d') would yield an iterator for element 'e', while 'd' has not been erased.

Proposed resolution:

In 21.3.5.5 [lib.string::erase], paragraph 10, change:

Returns: an iterator which points to the element immediately following _last_ prior to the element being erased.

to read

Returns: an iterator which points to the element pointed to by _last_ prior to the other elements being erased.


28. Ctype<char>is ambiguous

Section: 22.2.1.3.2 [lib.facet.ctype.char.members]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

The description of the vector form of ctype<char>::is can be interpreted to mean something very different from what was intended. Paragraph 4 says

Effects: The second form, for all *p in the range [low, high), assigns vec[p-low] to table()[(unsigned char)*p].

This is intended to copy the value indexed from table()[] into the place identified in vec[].

Proposed resolution:

Change 22.2.1.3.2 [lib.facet.ctype.char.members], paragraph 4, to read

Effects: The second form, for all *p in the range [low, high), assigns into vec[p-low] the value table()[(unsigned char)*p].


29. Ios_base::init doesn't exist

Section: 27.3.1 [lib.narrow.stream.objects]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

Sections 27.3.1 [lib.narrow.stream.objects] and 27.3.2 [lib.wide.stream.objects] mention a function ios_base::init, which is not defined. Probably they mean basic_ios<>::init, defined in 27.4.4.1 [lib.basic.ios.cons], paragraph 3.

Proposed resolution:

[R12: modified to include paragraph 5.]

In 27.3.1 [lib.narrow.stream.objects] paragraph 2 and 5, change

ios_base::init

to

basic_ios<char>::init

Also, make a similar change in 27.3.2 [lib.wide.stream.objects] except it should read

basic_ios<wchar_t>::init


30. Wrong header for LC_*

Section: 22.1.1.1.1 [lib.locale.category]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

Paragraph 2 implies that the C macros LC_CTYPE etc. are defined in <cctype>, where they are in fact defined elsewhere to appear in <clocale>.

Proposed resolution:

In 22.1.1.1.1 [lib.locale.category], paragraph 2, change "<cctype>" to read "<clocale>".


31. Immutable locale values

Section: 22.1.1 [lib.locale]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

Paragraph 6, says "An instance of locale is immutable; once a facet reference is obtained from it, ...". This has caused some confusion, because locale variables are manifestly assignable.

Proposed resolution:

In 22.1.1 [lib.locale] replace paragraph 6

An instance of locale is immutable; once a facet reference is obtained from it, that reference remains usable as long as the locale value itself exists.

with

Once a facet reference is obtained from a locale object by calling use_facet<>, that reference remains usable, and the results from member functions of it may be cached and re-used, as long as some locale object refers to that facet.


32. Pbackfail description inconsistent

Section: 27.5.2.4.4 [lib.streambuf.virt.pback]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

The description of the required state before calling virtual member basic_streambuf<>::pbackfail requirements is inconsistent with the conditions described in 27.5.2.2.4 [lib.streambuf.pub.pback] where member sputbackc calls it. Specifically, the latter says it calls pbackfail if:

    traits::eq(c,gptr()[-1]) is false

where pbackfail claims to require:

    traits::eq(*gptr(),traits::to_char_type(c)) returns false

It appears that the pbackfail description is wrong.

Proposed resolution:

In 27.5.2.4.4 [lib.streambuf.virt.pback], paragraph 1, change:

"traits::eq(*gptr(),traits::to_char_type( c))"

to

"traits::eq(traits::to_char_type(c),gptr()[-1])"

Rationale:

Note deliberate reordering of arguments for clarity in addition to the correction of the argument value.


33. Codecvt<> mentions from_type

Section: 22.2.1.5.2 [lib.locale.codecvt.virtuals]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

In the table defining the results from do_out and do_in, the specification for the result error says

encountered a from_type character it could not convert

but from_type is not defined. This clearly is intended to be an externT for do_in, or an internT for do_out.

Proposed resolution:

In 22.2.1.5.2 [lib.locale.codecvt.virtuals] paragraph 4, replace the definition in the table for the case of _error_ with

encountered a character in [from,from_end) that it could not convert.


34. True/falsename() not in ctype<>

Section: 22.2.2.2.2 [lib.facet.num.put.virtuals]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

In paragraph 19, Effects:, members truename() and falsename are used from facet ctype<charT>, but it has no such members. Note that this is also a problem in 22.2.2.1.2, addressed in (4).

Proposed resolution:

In 22.2.2.2.2 [lib.facet.num.put.virtuals], paragraph 19, in the Effects: clause for member put(...., bool), replace the initialization of the string_type value s as follows:

const numpunct& np = use_facet<numpunct<charT> >(loc);
string_type s = val ? np.truename() : np.falsename(); 

35. No manipulator unitbuf in synopsis

Section: 27.4 [lib.iostreams.base]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

In 27.4.5.1 [lib.fmtflags.manip], we have a definition for a manipulator named "unitbuf". Unlike other manipulators, it's not listed in synopsis. Similarly for "nounitbuf".

Proposed resolution:

Add to the synopsis for <ios> in 27.4 [lib.iostreams.base], after the entry for "nouppercase", the prototypes:

ios_base& unitbuf(ios_base& str);
ios_base& nounitbuf(ios_base& str); 

36. Iword & pword storage lifetime omitted

Section: 27.4.2.5 [lib.ios.base.storage]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

In the definitions for ios_base::iword and pword, the lifetime of the storage is specified badly, so that an implementation which only keeps the last value stored appears to conform. In particular, it says:

The reference returned may become invalid after another call to the object's iword member with a different index ...

This is not idle speculation; at least one implementation was done this way.

Proposed resolution:

Add in 27.4.2.5 [lib.ios.base.storage], in both paragraph 2 and also in paragraph 4, replace the sentence:

The reference returned may become invalid after another call to the object's iword [pword] member with a different index, after a call to its copyfmt member, or when the object is destroyed.

with:

The reference returned is invalid after any other operations on the object. However, the value of the storage referred to is retained, so that until the next call to copyfmt, calling iword [pword] with the same index yields another reference to the same value.

substituting "iword" or "pword" as appropriate.


37. Leftover "global" reference

Section: 22.1.1 [lib.locale]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

In the overview of locale semantics, paragraph 4, is the sentence

If Facet is not present in a locale (or, failing that, in the global locale), it throws the standard exception bad_cast.

This is not supported by the definition of use_facet<>, and represents semantics from an old draft.

Proposed resolution:

In 22.1.1 [lib.locale], paragraph 4, delete the parenthesized expression

(or, failing that, in the global locale)


38. Facet definition incomplete

Section: 22.1.2 [lib.locale.global.templates]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

It has been noticed by Esa Pulkkinen that the definition of "facet" is incomplete. In particular, a class derived from another facet, but which does not define a member id, cannot safely serve as the argument F to use_facet<F>(loc), because there is no guarantee that a reference to the facet instance stored in loc is safely convertible to F.

Proposed resolution:

In the definition of std::use_facet<>(), replace the text in paragraph 1 which reads:

Get a reference to a facet of a locale.

with:

Requires: Facet is a facet class whose definition contains the public static member id as defined in 22.1.1.1.2 [lib.locale.facet].

[ Kona: strike as overspecification the text "(not inherits)" from the original resolution, which read "... whose definition contains (not inherits) the public static member id..." ]


39. istreambuf_iterator<>::operator++(int) definition garbled

Section: 24.5.3.4 [lib.istreambuf.iterator::op++]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

Following the definition of istreambuf_iterator<>::operator++(int) in paragraph 3, the standard contains three lines of garbage text left over from a previous edit.

istreambuf_iterator<charT,traits> tmp = *this;
sbuf_->sbumpc();
return(tmp); 

Proposed resolution:

In 24.5.3.4 [lib.istreambuf.iterator::op++], delete the three lines of code at the end of paragraph 3.


40. Meaningless normative paragraph in examples

Section: 22.2.8 [lib.facets.examples]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

Paragraph 3 of the locale examples is a description of part of an implementation technique that has lost its referent, and doesn't mean anything.

Proposed resolution:

Delete 22.2.8 [lib.facets.examples] paragraph 3 which begins "This initialization/identification system depends...", or (at the editor's option) replace it with a place-holder to keep the paragraph numbering the same.


41. Ios_base needs clear(), exceptions()

Section: 27.4.2 [lib.ios.base]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

The description of ios_base::iword() and pword() in 27.4.2.4 [lib.ios.members.static], say that if they fail, they "set badbit, which may throw an exception". However, ios_base offers no interface to set or to test badbit; those interfaces are defined in basic_ios<>.

Proposed resolution:

Change the description in 27.4.2.5 [lib.ios.base.storage] in paragraph 2, and also in paragraph 4, as follows. Replace

If the function fails it sets badbit, which may throw an exception.

with

If the function fails, and *this is a base sub-object of a basic_ios<> object or sub-object, the effect is equivalent to calling basic_ios<>::setstate(badbit) on the derived object (which may throw failure).

[Kona: LWG reviewed wording; setstate(failbit) changed to setstate(badbit).]


42. String ctors specify wrong default allocator

Section: 21.3 [lib.basic.string]  Status: TC  Submitter: Nathan Myers  Date: 6 Aug 1998

The basic_string<> copy constructor:

basic_string(const basic_string& str, size_type pos = 0,
             size_type n = npos, const Allocator& a = Allocator()); 

specifies an Allocator argument default value that is counter-intuitive. The natural choice for a the allocator to copy from is str.get_allocator(). Though this cannot be expressed in default-argument notation, overloading suffices.

Alternatively, the other containers in Clause 23 (deque, list, vector) do not have this form of constructor, so it is inconsistent, and an evident source of confusion, for basic_string<> to have it, so it might better be removed.

Proposed resolution:

In 21.3 [lib.basic.string], replace the declaration of the copy constructor as follows:

basic_string(const basic_string& str);
basic_string(const basic_string& str, size_type pos, size_type n = npos,
             const Allocator& a = Allocator());

In 21.3.1 [lib.string.cons], replace the copy constructor declaration as above. Add to paragraph 5, Effects:

In the first form, the Allocator value used is copied from str.get_allocator().

Rationale:

The LWG believes the constructor is actually broken, rather than just an unfortunate design choice.

The LWG considered two other possible resolutions:

A. In 21.3 [lib.basic.string], replace the declaration of the copy constructor as follows:

basic_string(const basic_string& str, size_type pos = 0,
             size_type n = npos);
basic_string(const basic_string& str, size_type pos,
             size_type n, const Allocator& a); 

In 21.3.1 [lib.string.cons], replace the copy constructor declaration as above. Add to paragraph 5, Effects:

When no Allocator argument is provided, the string is constructed using the value str.get_allocator().

B. In 21.3 [lib.basic.string], and also in 21.3.1 [lib.string.cons], replace the declaration of the copy constructor as follows:

basic_string(const basic_string& str, size_type pos = 0,
             size_type n = npos); 

The proposed resolution reflects the original intent of the LWG. It was also noted by Pete Becker that this fix "will cause a small amount of existing code to now work correctly."

[ Kona: issue editing snafu fixed - the proposed resolution now correctly reflects the LWG consensus. ]


44. Iostreams use operator== on int_type values

Section: 27 [lib.input.output]  Status: WP  Submitter: Nathan Myers  Date: 6 Aug 1998

Many of the specifications for iostreams specify that character values or their int_type equivalents are compared using operators == or !=, though in other places traits::eq() or traits::eq_int_type is specified to be used throughout. This is an inconsistency; we should change uses of == and != to use the traits members instead.

Proposed resolution:

[Pre-Kona: Dietmar supplied wording]

List of changes to clause 27:

  1. In lib.basic.ios.members paragraph 13 (postcondition clause for 'fill(cT)') change
    fillch == fill()
    to
    traits::eq(fillch, fill())
  2. In lib.istream.unformatted paragraph 7 (effects clause for 'get(cT,streamsize,cT)'), third bullet, change
    c == delim for the next available input character c
    to
    traits::eq(c, delim) for the next available input character c
  3. In lib.istream.unformatted paragraph 12 (effects clause for 'get(basic_streambuf<cT,Tr>&,cT)'), third bullet, change
    c == delim for the next available input character c
    to
    traits::eq(c, delim) for the next available input character c
  4. In lib.istream.unformatted paragraph 17 (effects clause for 'getline(cT,streamsize,cT)'), second bullet, change
    c == delim for the next available input character c
    to
    traits::eq(c, delim) for the next available input character c
  5. In lib.istream.unformatted paragraph 24 (effects clause for 'ignore(int,int_type)'), second bullet, change
    c == delim for the next available input character c
    to
    traits::eq_int_type(c, delim) for the next available input character c
  6. In lib.istream.unformatted paragraph 25 (notes clause for 'ignore(int,int_type)'), second bullet, change
    The last condition will never occur if delim == traits::eof()
    to
    The last condition will never occur if traits::eq_int_type(delim, traits::eof()).
  7. In lib.istream.sentry paragraph 6 (example implementation for the sentry constructor) change
    while ((c = is.rdbuf()->snextc()) != traits::eof()) {
    to
    while (!traits::eq_int_type(c = is.rdbuf()->snextc(), traits::eof())) {

List of changes to Chapter 21:

  1. In lib.string::find paragraph 1 (effects clause for find()), second bullet, change
    at(xpos+I) == str.at(I) for all elements ...
    to
    traits::eq(at(xpos+I), str.at(I)) for all elements ...
  2. In lib.string::rfind paragraph 1 (effects clause for rfind()), second bullet, change
    at(xpos+I) == str.at(I) for all elements ...
    to
    traits::eq(at(xpos+I), str.at(I)) for all elements ...
  3. In lib.string::find.first.of paragraph 1 (effects clause for find_first_of()), second bullet, change
    at(xpos+I) == str.at(I) for all elements ...
    to
    traits::eq(at(xpos+I), str.at(I)) for all elements ...
  4. In lib.string::find.last.of paragraph 1 (effects clause for find_last_of()), second bullet, change
    at(xpos+I) == str.at(I) for all elements ...
    to
    traits::eq(at(xpos+I), str.at(I)) for all elements ...
  5. In lib.string::find.first.not.of paragraph 1 (effects clause for find_first_not_of()), second bullet, change
    at(xpos+I) == str.at(I) for all elements ...
    to
    traits::eq(at(xpos+I), str.at(I)) for all elements ...
  6. In lib.string::find.last.not.of paragraph 1 (effects clause for find_last_not_of()), second bullet, change
    at(xpos+I) == str.at(I) for all elements ...
    to
    traits::eq(at(xpos+I), str.at(I)) for all elements ...
  7. In lib.string.ios paragraph 5 (effects clause for getline()), second bullet, change
    c == delim for the next available input character c
    to
    traits::eq(c, delim) for the next available input character c

Notes:


46. Minor Annex D errors

Section: D.7 [depr.str.strstreams]  Status: TC  Submitter: Brendan Kehoe  Date:  1 Jun 1998

See lib-6522 and edit-814.

Proposed resolution:

Change D.7.1 [depr.strstreambuf] (since streambuf is a typedef of basic_streambuf<char>) from:

         virtual streambuf<char>* setbuf(char* s, streamsize n);

to:

         virtual streambuf* setbuf(char* s, streamsize n);

In D.7.4 [depr.strstream] insert the semicolon now missing after int_type:

     namespace std {
       class strstream
         : public basic_iostream<char> {
       public:
         // Types
         typedef char                                char_type;
         typedef typename char_traits<char>::int_type int_type
         typedef typename char_traits<char>::pos_type pos_type;

47. Imbue() and getloc() Returns clauses swapped

Section: 27.4.2.3 [lib.ios.base.locales]  Status: TC  Submitter: Matt Austern  Date: 21 Jun 1998

Section 27.4.2.3 specifies how imbue() and getloc() work. That section has two RETURNS clauses, and they make no sense as stated. They make perfect sense, though, if you swap them. Am I correct in thinking that paragraphs 2 and 4 just got mixed up by accident?

Proposed resolution:

In 27.4.2.3 [lib.ios.base.locales] swap paragraphs 2 and 4.


48. Use of non-existent exception constructor

Section: 27.4.2.1.1 [lib.ios::failure]  Status: TC  Submitter: Matt Austern  Date: 21 Jun 1998

27.4.2.1.1, paragraph 2, says that class failure initializes the base class, exception, with exception(msg). Class exception (see 18.6.1) has no such constructor.

Proposed resolution:

Replace 27.4.2.1.1 [lib.ios::failure], paragraph 2, with

EFFECTS: Constructs an object of class failure.


49. Underspecification of ios_base::sync_with_stdio

Section: 27.4.2.4 [lib.ios.members.static]  Status: WP  Submitter: Matt Austern  Date: 21 Jun 1998

Two problems

(1) 27.4.2.4 doesn't say what ios_base::sync_with_stdio(f) returns. Does it return f, or does it return the previous synchronization state? My guess is the latter, but the standard doesn't say so.

(2) 27.4.2.4 doesn't say what it means for streams to be synchronized with stdio. Again, of course, I can make some guesses. (And I'm unhappy about the performance implications of those guesses, but that's another matter.)

Proposed resolution:

Change the following sentence in 27.4.2.4 [lib.ios.members.static] returns clause from:

true if the standard iostream objects (27.3) are synchronized and otherwise returns false.

to:

true if the previous state of the standard iostream objects (27.3) was synchronized and otherwise returns false.

Add the following immediately after 27.4.2.4 [lib.ios.members.static], paragraph 2:

When a standard iostream object str is synchronized with a standard stdio stream f, the effect of inserting a character c by

  fputc(f, c);

is the same as the effect of

  str.rdbuf()->sputc(c);

for any sequence of characters; the effect of extracting a character c by

  c = fgetc(f);

is the same as the effect of:

  c = str.rdbuf()->sbumpc(c);

for any sequences of characters; and the effect of pushing back a character c by

  ungetc(c, f);

is the same as the effect of

  str.rdbuf()->sputbackc(c);

for any sequence of characters. [Footnote: This implies that operations on a standard iostream object can be mixed arbitrarily with operations on the corresponding stdio stream. In practical terms, synchronization usually means that a standard iostream object and a standard stdio object share a buffer. --End Footnote]

[pre-Copenhagen: PJP and Matt contributed the definition of "synchronization"]

[post-Copenhagen: proposed resolution was revised slightly: text was added in the non-normative footnote to say that operations on the two streams can be mixed arbitrarily.]


50. Copy constructor and assignment operator of ios_base

Section: 27.4.2 [lib.ios.base]  Status: TC  Submitter: Matt Austern  Date: 21 Jun 1998

As written, ios_base has a copy constructor and an assignment operator. (Nothing in the standard says it doesn't have one, and all classes have copy constructors and assignment operators unless you take specific steps to avoid them.) However, nothing in 27.4.2 says what the copy constructor and assignment operator do.

My guess is that this was an oversight, that ios_base is, like basic_ios, not supposed to have a copy constructor or an assignment operator.

Jerry Schwarz comments: Yes, its an oversight, but in the opposite sense to what you're suggesting. At one point there was a definite intention that you could copy ios_base. It's an easy way to save the entire state of a stream for future use. As you note, to carry out that intention would have required a explicit description of the semantics (e.g. what happens to the iarray and parray stuff).

Proposed resolution:

In 27.4.2 [lib.ios.base], class ios_base, specify the copy constructor and operator= members as being private.

Rationale:

The LWG believes the difficulty of specifying correct semantics outweighs any benefit of allowing ios_base objects to be copyable.


51. Requirement to not invalidate iterators missing

Section: 23.1 [lib.container.requirements]  Status: TC  Submitter: David Vandevoorde  Date: 23 Jun 1998

The std::sort algorithm can in general only sort a given sequence by moving around values. The list<>::sort() member on the other hand could move around values or just update internal pointers. Either method can leave iterators into the list<> dereferencable, but they would point to different things.

Does the FDIS mandate anywhere which method should be used for list<>::sort()?

Matt Austern comments:

I think you've found an omission in the standard.

The library working group discussed this point, and there was supposed to be a general requirement saying that list, set, map, multiset, and multimap may not invalidate iterators, or change the values that iterators point to, except when an operation does it explicitly. So, for example, insert() doesn't invalidate any iterators and erase() and remove() only invalidate iterators pointing to the elements that are being erased.

I looked for that general requirement in the FDIS, and, while I found a limited form of it for the sorted associative containers, I didn't find it for list. It looks like it just got omitted.

The intention, though, is that list<>::sort does not invalidate any iterators and does not change the values that any iterator points to. There would be no reason to have the member function otherwise.

Proposed resolution:

Add a new paragraph at the end of 23.1:

Unless otherwise specified (either explicitly or by defining a function in terms of other functions), invoking a container member function or passing a container as an argument to a library function shall not invalidate iterators to, or change the values of, objects within that container.

Rationale:

This was US issue CD2-23-011; it was accepted in London but the change was not made due to an editing oversight. The wording in the proposed resolution below is somewhat updated from CD2-23-011, particularly the addition of the phrase "or change the values of"


52. Small I/O problems

Section: 27.4.3.2 [lib.fpos.operations]  Status: TC  Submitter: Matt Austern  Date: 23 Jun 1998

First, 27.4.4.1 [lib.basic.ios.cons], table 89. This is pretty obvious: it should be titled "basic_ios<>() effects", not "ios_base() effects".

[The second item is a duplicate; see issue 6 for resolution.]

Second, 27.4.3.2 [lib.fpos.operations] table 88 . There are a couple different things wrong with it, some of which I've already discussed with Jerry, but the most obvious mechanical sort of error is that it uses expressions like P(i) and p(i), without ever defining what sort of thing "i" is.

(The other problem is that it requires support for streampos arithmetic. This is impossible on some systems, i.e. ones where file position is a complicated structure rather than just a number. Jerry tells me that the intention was to require syntactic support for streampos arithmetic, but that it wasn't actually supposed to do anything meaningful except on platforms, like Unix, where genuine arithmetic is possible.)

Proposed resolution:

Change 27.4.4.1 [lib.basic.ios.cons] table 89 title from "ios_base() effects" to "basic_ios<>() effects".


53. Basic_ios destructor unspecified

Section: 27.4.4.1 [lib.basic.ios.cons]  Status: TC  Submitter: Matt Austern  Date: 23 Jun 1998

There's nothing in 27.4.4 saying what basic_ios's destructor does. The important question is whether basic_ios::~basic_ios() destroys rdbuf().

Proposed resolution:

Add after 27.4.4.1 [lib.basic.ios.cons] paragraph 2:

virtual ~basic_ios();

Notes: The destructor does not destroy rdbuf().

Rationale:

The LWG reviewed the additional question of whether or not rdbuf(0) may set badbit. The answer is clearly yes; it may be set via clear(). See 27.4.4.2 [lib.basic.ios.members], paragraph 6. This issue was reviewed at length by the LWG, which removed from the original proposed resolution a footnote which incorrectly said "rdbuf(0) does not set badbit".


54. Basic_streambuf's destructor

Section: 27.5.2.1 [lib.streambuf.cons]  Status: TC  Submitter: Matt Austern  Date: 25 Jun 1998

The class synopsis for basic_streambuf shows a (virtual) destructor, but the standard doesn't say what that destructor does. My assumption is that it does nothing, but the standard should say so explicitly.

Proposed resolution:

Add after 27.5.2.1 [lib.streambuf.cons] paragraph 2:

virtual  ~basic_streambuf();

Effects: None.


55. Invalid stream position is undefined

Section: 27 [lib.input.output]  Status: TC  Submitter: Matt Austern  Date: 26 Jun 1998

Several member functions in clause 27 are defined in certain circumstances to return an "invalid stream position", a term that is defined nowhere in the standard. Two places (27.5.2.4.2, paragraph 4, and 27.8.1.4, paragraph 15) contain a cross-reference to a definition in _lib.iostreams.definitions_, a nonexistent section.

I suspect that the invalid stream position is just supposed to be pos_type(-1). Probably best to say explicitly in (for example) 27.5.2.4.2 that the return value is pos_type(-1), rather than to use the term "invalid stream position", define that term somewhere, and then put in a cross-reference.

The phrase "invalid stream position" appears ten times in the C++ Standard. In seven places it refers to a return value, and it should be changed. In three places it refers to an argument, and it should not be changed. Here are the three places where "invalid stream position" should not be changed:

27.7.1.3 [lib.stringbuf.virtuals], paragraph 14
27.8.1.4 [lib.filebuf.virtuals], paragraph 14
D.7.1.3 [depr.strstreambuf.virtuals], paragraph 17

Proposed resolution:

In 27.5.2.4.2 [lib.streambuf.virt.buffer], paragraph 4, change "Returns an object of class pos_type that stores an invalid stream position (_lib.iostreams.definitions_)" to "Returns pos_type(off_type(-1))".

In 27.5.2.4.2 [lib.streambuf.virt.buffer], paragraph 6, change "Returns an object of class pos_type that stores an invalid stream position" to "Returns pos_type(off_type(-1))".

In 27.7.1.3 [lib.stringbuf.virtuals], paragraph 13, change "the object stores an invalid stream position" to "the return value is pos_type(off_type(-1))".

In 27.8.1.4 [lib.filebuf.virtuals], paragraph 13, change "returns an invalid stream position (27.4.3)" to "returns pos_type(off_type(-1))"

In 27.8.1.4 [lib.filebuf.virtuals], paragraph 15, change "Otherwise returns an invalid stream position (_lib.iostreams.definitions_)" to "Otherwise returns pos_type(off_type(-1))"

In D.7.1.3 [depr.strstreambuf.virtuals], paragraph 15, change "the object stores an invalid stream position" to "the return value is pos_type(off_type(-1))"

In D.7.1.3 [depr.strstreambuf.virtuals], paragraph 18, change "the object stores an invalid stream position" to "the return value is pos_type(off_type(-1))"


56. Showmanyc's return type

Section: 27.5.2 [lib.streambuf]  Status: TC  Submitter: Matt Austern  Date: 29 Jun 1998

The class summary for basic_streambuf<>, in 27.5.2, says that showmanyc has return type int. However, 27.5.2.4.3 says that its return type is streamsize.

Proposed resolution:

Change showmanyc's return type in the 27.5.2 [lib.streambuf] class summary to streamsize.


57. Mistake in char_traits

Section: 21.1.3.2 [lib.char.traits.specializations.wchar.t]  Status: TC  Submitter: Matt Austern  Date: 1 Jul 1998

21.1.3.2, paragraph 3, says "The types streampos and wstreampos may be different if the implementation supports no shift encoding in narrow-oriented iostreams but supports one or more shift encodings in wide-oriented streams".

That's wrong: the two are the same type. The <iosfwd> summary in 27.2 says that streampos and wstreampos are, respectively, synonyms for fpos<char_traits<char>::state_type> and fpos<char_traits<wchar_t>::state_type>, and, flipping back to clause 21, we see in 21.1.3.1 and 21.1.3.2 that char_traits<char>::state_type and char_traits<wchar_t>::state_type must both be mbstate_t.

Proposed resolution:

Remove the sentence in 21.1.3.2 [lib.char.traits.specializations.wchar.t] paragraph 3 which begins "The types streampos and wstreampos may be different..." .


59. Ambiguity in specification of gbump

Section: 27.5.2.3.1 [lib.streambuf.get.area]  Status: TC  Submitter: Matt Austern  Date: 28 Jul 1998

27.5.2.3.1 says that basic_streambuf::gbump() "Advances the next pointer for the input sequence by n."

The straightforward interpretation is that it is just gptr() += n. An alternative interpretation, though, is that it behaves as if it calls sbumpc n times. (The issue, of course, is whether it might ever call underflow.) There is a similar ambiguity in the case of pbump.

(The "classic" AT&T implementation used the former interpretation.)

Proposed resolution:

Change 27.5.2.3.1 [lib.streambuf.get.area] paragraph 4 gbump effects from:

Effects: Advances the next pointer for the input sequence by n.

to:

Effects: Adds n to the next pointer for the input sequence.

Make the same change to 27.5.2.3.2 [lib.streambuf.put.area] paragraph 4 pbump effects.


60. What is a formatted input function?

Section: 27.6.1.2.1 [lib.istream.formatted.reqmts]  Status: TC  Submitter: Matt Austern  Date: 3 Aug 1998

Paragraph 1 of 27.6.1.2.1 contains general requirements for all formatted input functions. Some of the functions defined in section 27.6.1.2 explicitly say that those requirements apply ("Behaves like a formatted input member (as described in 27.6.1.2.1)"), but others don't. The question: is 27.6.1.2.1 supposed to apply to everything in 27.6.1.2, or only to those member functions that explicitly say "behaves like a formatted input member"? Or to put it differently: are we to assume that everything that appears in a section called "Formatted input functions" really is a formatted input function? I assume that 27.6.1.2.1 is intended to apply to the arithmetic extractors (27.6.1.2.2), but I assume that it is not intended to apply to extractors like

    basic_istream& operator>>(basic_istream& (*pf)(basic_istream&));

and

    basic_istream& operator>>(basic_streammbuf*);

There is a similar ambiguity for unformatted input, formatted output, and unformatted output.

Comments from Judy Ward: It seems like the problem is that the basic_istream and basic_ostream operator <<()'s that are used for the manipulators and streambuf* are in the wrong section and should have their own separate section or be modified to make it clear that the "Common requirements" listed in section 27.6.1.2.1 (for basic_istream) and section 27.6.2.5.1 (for basic_ostream) do not apply to them.

Additional comments from Dietmar Kühl: It appears to be somewhat nonsensical to consider the functions defined in 27.6.1.2.3 [lib.istream::extractors] paragraphs 1 to 5 to be "Formatted input function" but since these functions are defined in a section labeled "Formatted input functions" it is unclear to me whether these operators are considered formatted input functions which have to conform to the "common requirements" from 27.6.1.2.1 [lib.istream.formatted.reqmts]: If this is the case, all manipulators, not just ws, would skip whitespace unless noskipws is set (... but setting noskipws using the manipulator syntax would also skip whitespace :-)

It is not clear which functions are to be considered unformatted input functions. As written, it seems that all functions in 27.6.1.3 [lib.istream.unformatted] are unformatted input functions. However, it does not really make much sense to construct a sentry object for gcount(), sync(), ... Also it is unclear what happens to the gcount() if eg. gcount(), putback(), unget(), or sync() is called: These functions don't extract characters, some of them even "unextract" a character. Should this still be reflected in gcount()? Of course, it could be read as if after a call to gcount() gcount() return 0 (the last unformatted input function, gcount(), didn't extract any character) and after a call to putback() gcount() returns -1 (the last unformatted input function putback() did "extract" back into the stream). Correspondingly for unget(). Is this what is intended? If so, this should be clarified. Otherwise, a corresponding clarification should be used.

Proposed resolution:

In 27.6.1.2.2 [lib.istream.formatted.arithmetic], paragraph 1. Change the beginning of the second sentence from "The conversion occurs" to "These extractors behave as formatted input functions (as described in 27.6.1.2.1). After a sentry object is constructed, the conversion occurs"

In 27.6.1.2.3, [lib.istream::extractors], before paragraph 1. Add an effects clause. "Effects: None. This extractor does not behave as a formatted input function (as described in 27.6.1.2.1).

In 27.6.1.2.3, [lib.istream::extractors], paragraph 2. Change the effects clause to "Effects: Calls pf(*this). This extractor does not behave as a formatted input function (as described in 27.6.1.2.1).

In 27.6.1.2.3, [lib.istream::extractors], paragraph 4. Change the effects clause to "Effects: Calls pf(*this). This extractor does not behave as a formatted input function (as described in 27.6.1.2.1).

In 27.6.1.2.3, [lib.istream::extractors], paragraph 12. Change the first two sentences from "If sb is null, calls setstate(failbit), which may throw ios_base::failure (27.4.4.3). Extracts characters from *this..." to "Behaves as a formatted input function (as described in 27.6.1.2.1). If sb is null, calls setstate(failbit), which may throw ios_base::failure (27.4.4.3). After a sentry object is constructed, extracts characters from *this...".

In 27.6.1.3, [lib.istream.unformatted], before paragraph 2. Add an effects clause. "Effects: none. This member function does not behave as an unformatted input function (as described in 27.6.1.3, paragraph 1)."

In 27.6.1.3, [lib.istream.unformatted], paragraph 3. Change the beginning of the first sentence of the effects clause from "Extracts a character" to "Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1). After constructing a sentry object, extracts a character"

In 27.6.1.3, [lib.istream.unformatted], paragraph 5. Change the beginning of the first sentence of the effects clause from "Extracts a character" to "Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1). After constructing a sentry object, extracts a character"

In 27.6.1.3, [lib.istream.unformatted], paragraph 5. Change the beginning of the first sentence of the effects clause from "Extracts characters" to "Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1). After constructing a sentry object, extracts characters"

[No change needed in paragraph 10, because it refers to paragraph 7.]

In 27.6.1.3, [lib.istream.unformatted], paragraph 12. Change the beginning of the first sentence of the effects clause from "Extracts characters" to "Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1). After constructing a sentry object, extracts characters"

[No change needed in paragraph 15.]

In 27.6.1.3, [lib.istream.unformatted], paragraph 17. Change the beginning of the first sentence of the effects clause from "Extracts characters" to "Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1). After constructing a sentry object, extracts characters"

[No change needed in paragraph 23.]

In 27.6.1.3, [lib.istream.unformatted], paragraph 24. Change the beginning of the first sentence of the effects clause from "Extracts characters" to "Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1). After constructing a sentry object, extracts characters"

In 27.6.1.3, [lib.istream.unformatted], before paragraph 27. Add an Effects clause: "Effects: Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1). After constructing a sentry object, reads but does not extract the current input character."

In 27.6.1.3, [lib.istream.unformatted], paragraph 28. Change the first sentence of the Effects clause from "If !good() calls" to Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1). After constructing a sentry object, if !good() calls"

In 27.6.1.3, [lib.istream.unformatted], paragraph 30. Change the first sentence of the Effects clause from "If !good() calls" to "Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1). After constructing a sentry object, if !good() calls"

In 27.6.1.3, [lib.istream.unformatted], paragraph 32. Change the first sentence of the Effects clause from "If !good() calls..." to "Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1). After constructing a sentry object, if !good() calls..." Add a new sentence to the end of the Effects clause: "[Note: this function extracts no characters, so the value returned by the next call to gcount() is 0.]"

In 27.6.1.3, [lib.istream.unformatted], paragraph 34. Change the first sentence of the Effects clause from "If !good() calls" to "Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1). After constructing a sentry object, if !good() calls". Add a new sentence to the end of the Effects clause: "[Note: this function extracts no characters, so the value returned by the next call to gcount() is 0.]"

In 27.6.1.3, [lib.istream.unformatted], paragraph 36. Change the first sentence of the Effects clause from "If !rdbuf() is" to "Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1), except that it does not count the number of characters extracted and does not affect the value returned by subsequent calls to gcount(). After constructing a sentry object, if rdbuf() is"

In 27.6.1.3, [lib.istream.unformatted], before paragraph 37. Add an Effects clause: "Effects: Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1), except that it does not count the number of characters extracted and does not affect the value returned by subsequent calls to gcount()." Change the first sentence of paragraph 37 from "if fail()" to "after constructing a sentry object, if fail()".

In 27.6.1.3, [lib.istream.unformatted], paragraph 38. Change the first sentence of the Effects clause from "If fail()" to "Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1), except that it does not count the number of characters extracted and does not affect the value returned by subsequent calls to gcount(). After constructing a sentry object, if fail()

In 27.6.1.3, [lib.istream.unformatted], paragraph 40. Change the first sentence of the Effects clause from "If fail()" to "Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1), except that it does not count the number of characters extracted and does not affect the value returned by subsequent calls to gcount(). After constructing a sentry object, if fail()

In 27.6.2.5.2 [lib.ostream.inserters.arithmetic], paragraph 1. Change the beginning of the third sentence from "The formatting conversion" to "These extractors behave as formatted output functions (as described in 27.6.2.5.1). After the sentry object is constructed, the conversion occurs".

In 27.6.2.5.3 [lib.ostream.inserters], before paragraph 1. Add an effects clause: "Effects: None. Does not behave as a formatted output function (as described in 27.6.2.5.1).".

In 27.6.2.5.3 [lib.ostream.inserters], paragraph 2. Change the effects clause to "Effects: calls pf(*this). This extractor does not behave as a formatted output function (as described in 27.6.2.5.1).".

In 27.6.2.5.3 [lib.ostream.inserters], paragraph 4. Change the effects clause to "Effects: calls pf(*this). This extractor does not behave as a formatted output function (as described in 27.6.2.5.1).".

In 27.6.2.5.3 [lib.ostream.inserters], paragraph 6. Change the first sentence from "If sb" to "Behaves as a formatted output function (as described in 27.6.2.5.1). After the sentry object is constructed, if sb".

In 27.6.2.6 [lib.ostream.unformatted], paragraph 2. Change the first sentence from "Inserts the character" to "Behaves as an unformatted output function (as described in 27.6.2.6, paragraph 1). After constructing a sentry object, inserts the character".

In 27.6.2.6 [lib.ostream.unformatted], paragraph 5. Change the first sentence from "Obtains characters" to "Behaves as an unformatted output function (as described in 27.6.2.6, paragraph 1). After constructing a sentry object, obtains characters".

In 27.6.2.6 [lib.ostream.unformatted], paragraph 7. Add a new sentence at the end of the paragraph: "Does not behave as an unformatted output function (as described in 27.6.2.6, paragraph 1)."

Rationale:

See J16/99-0043==WG21/N1219, Proposed Resolution to Library Issue 60, by Judy Ward and Matt Austern. This proposed resolution is section VI of that paper.


61. Ambiguity in iostreams exception policy

Section: 27.6.1.3 [lib.istream.unformatted]  Status: TC  Submitter: Matt Austern  Date: 6 Aug 1998

The introduction to the section on unformatted input (27.6.1.3) says that every unformatted input function catches all exceptions that were thrown during input, sets badbit, and then conditionally rethrows the exception. That seems clear enough. Several of the specific functions, however, such as get() and read(), are documented in some circumstances as setting eofbit and/or failbit. (The standard notes, correctly, that setting eofbit or failbit can sometimes result in an exception being thrown.) The question: if one of these functions throws an exception triggered by setting failbit, is this an exception "thrown during input" and hence covered by 27.6.1.3, or does 27.6.1.3 only refer to a limited class of exceptions? Just to make this concrete, suppose you have the following snippet.

  
  char buffer[N];
  istream is;
  ...
  is.exceptions(istream::failbit); // Throw on failbit but not on badbit.
  is.read(buffer, N);

Now suppose we reach EOF before we've read N characters. What iostate bits can we expect to be set, and what exception (if any) will be thrown?

Proposed resolution:

In 27.6.1.3, paragraph 1, after the sentence that begins "If an exception is thrown...", add the following parenthetical comment: "(Exceptions thrown from basic_ios<>::clear() are not caught or rethrown.)"

Rationale:

The LWG looked to two alternative wordings, and choose the proposed resolution as better standardese.


62. Sync's return value

Section: 27.6.1.3 [lib.istream.unformatted]  Status: TC  Submitter: Matt Austern  Date: 6 Aug 1998

The Effects clause for sync() (27.6.1.3, paragraph 36) says that it "calls rdbuf()->pubsync() and, if that function returns -1 ... returns traits::eof()."

That looks suspicious, because traits::eof() is of type traits::int_type while the return type of sync() is int.

Proposed resolution:

In 27.6.1.3 [lib.istream.unformatted], paragraph 36, change "returns traits::eof()" to "returns -1".


63. Exception-handling policy for unformatted output

Section: 27.6.2.6 [lib.ostream.unformatted]  Status: TC  Submitter: Matt Austern  Date: 11 Aug 1998

Clause 27 details an exception-handling policy for formatted input, unformatted input, and formatted output. It says nothing for unformatted output (27.6.2.6). 27.6.2.6 should either include the same kind of exception-handling policy as in the other three places, or else it should have a footnote saying that the omission is deliberate.

Proposed resolution:

In 27.6.2.6, paragraph 1, replace the last sentence ("In any case, the unformatted output function ends by destroying the sentry object, then returning the value specified for the formatted output function.") with the following text:

If an exception is thrown during output, then ios::badbit is turned on [Footnote: without causing an ios::failure to be thrown.] in *this's error state. If (exceptions() & badbit) != 0 then the exception is rethrown. In any case, the unformatted output function ends by destroying the sentry object, then, if no exception was thrown, returning the value specified for the formatted output function.

Rationale:

This exception-handling policy is consistent with that of formatted input, unformatted input, and formatted output.


64. Exception handling in basic_istream::operator>>(basic_streambuf*)

Section: 27.6.1.2.3 [lib.istream::extractors]  Status: TC  Submitter: Matt Austern  Date: 11 Aug 1998

27.6.1.2.3, paragraph 13, is ambiguous. It can be interpreted two different ways, depending on whether the second sentence is read as an elaboration of the first.

Proposed resolution:

Replace 27.6.1.2.3 [lib.istream::extractors], paragraph 13, which begins "If the function inserts no characters ..." with:

If the function inserts no characters, it calls setstate(failbit), which may throw ios_base::failure (27.4.4.3). If it inserted no characters because it caught an exception thrown while extracting characters from sb and failbit is on in exceptions() (27.4.4.3), then the caught exception is rethrown.


66. Strstreambuf::setbuf

Section: D.7.1.3 [depr.strstreambuf.virtuals]  Status: TC  Submitter: Matt Austern  Date: 18 Aug 1998

D.7.1.3, paragraph 19, says that strstreambuf::setbuf "Performs an operation that is defined separately for each class derived from strstreambuf". This is obviously an incorrect cut-and-paste from basic_streambuf. There are no classes derived from strstreambuf.

Proposed resolution:

D.7.1.3 [depr.strstreambuf.virtuals], paragraph 19, replace the setbuf effects clause which currently says "Performs an operation that is defined separately for each class derived from strstreambuf" with:

Effects: implementation defined, except that setbuf(0,0) has no effect.


68. Extractors for char* should store null at end

Section: 27.6.1.2.3 [lib.istream::extractors]  Status: TC  Submitter: Angelika Langer  Date: 14 Jul 1998

Extractors for char* (27.6.1.2.3) do not store a null character after the extracted character sequence whereas the unformatted functions like get() do. Why is this?

Comment from Jerry Schwarz: There is apparently an editing glitch. You'll notice that the last item of the list of what stops extraction doesn't make any sense. It was supposed to be the line that said a null is stored.

Proposed resolution:

27.6.1.2.3 [lib.istream::extractors], paragraph 7, change the last list item from:

A null byte (charT()) in the next position, which may be the first position if no characters were extracted.

to become a new paragraph which reads:

Operator>> then stores a null byte (charT()) in the next position, which may be the first position if no characters were extracted.

69. Must elements of a vector be contiguous?

Section: 23.2.4 [lib.vector]  Status: TC  Submitter: Andrew Koenig  Date: 29 Jul 1998

The issue is this: Must the elements of a vector be in contiguous memory?

(Please note that this is entirely separate from the question of whether a vector iterator is required to be a pointer; the answer to that question is clearly "no," as it would rule out debugging implementations)

Proposed resolution:

Add the following text to the end of 23.2.4 [lib.vector], paragraph 1.

The elements of a vector are stored contiguously, meaning that if v is a vector<T, Allocator> where T is some type other than bool, then it obeys the identity &v[n] == &v[0] + n for all 0 <= n < v.size().

Rationale:

The LWG feels that as a practical matter the answer is clearly "yes". There was considerable discussion as to the best way to express the concept of "contiguous", which is not directly defined in the standard. Discussion included:


70. Uncaught_exception() missing throw() specification

Section: 18.6 [lib.support.exception], 18.6.4 [lib.uncaught]  Status: TC  Submitter: Steve Clamage  Date: Unknown

In article 3E04@pratique.fr, Valentin Bonnard writes:

uncaught_exception() doesn't have a throw specification.

It is intentional ? Does it means that one should be prepared to handle exceptions thrown from uncaught_exception() ?

uncaught_exception() is called in exception handling contexts where exception safety is very important.

Proposed resolution:

In 15.5.3 [except.uncaught], paragraph 1, 18.6 [lib.support.exception], and 18.6.4 [lib.uncaught], add "throw()" to uncaught_exception().


71. Do_get_monthname synopsis missing argument

Section: 22.2.5.1 [lib.locale.time.get]  Status: TC  Submitter: Nathan Myers  Date: 13 Aug 1998

The locale facet member time_get<>::do_get_monthname is described in 22.2.5.1.2 [lib.locale.time.get.virtuals] with five arguments, consistent with do_get_weekday and with its specified use by member get_monthname. However, in the synopsis, it is specified instead with four arguments. The missing argument is the "end" iterator value.

Proposed resolution:

In 22.2.5.1 [lib.locale.time.get], add an "end" argument to the declaration of member do_monthname as follows:

  virtual iter_type do_get_monthname(iter_type s, iter_type end, ios_base&,
                                     ios_base::iostate& err, tm* t) const;

74. Garbled text for codecvt::do_max_length

Section: 22.2.1.5.2 [lib.locale.codecvt.virtuals]  Status: TC  Submitter: Matt Austern  Date: 8 Sep 1998

The text of codecvt::do_max_length's "Returns" clause (22.2.1.5.2, paragraph 11) is garbled. It has unbalanced parentheses and a spurious n.

Proposed resolution:

Replace 22.2.1.5.2 [lib.locale.codecvt.virtuals] paragraph 11 with the following:

Returns: The maximum value that do_length(state, from, from_end, 1) can return for any valid range [from, from_end) and stateT value state. The specialization codecvt<char, char, mbstate_t>::do_max_length() returns 1.

75. Contradiction in codecvt::length's argument types

Section: 22.2.1.5 [lib.locale.codecvt]  Status: TC  Submitter:  Matt Austern  Date:  18 Sep 1998

The class synopses for classes codecvt<> (22.2.1.5) and codecvt_byname<> (22.2.1.6) say that the first parameter of the member functions length and do_length is of type const stateT&. The member function descriptions, however (22.2.1.5.1, paragraph 6; 22.2.1.5.2, paragraph 9) say that the type is stateT&. Either the synopsis or the summary must be changed.

If (as I believe) the member function descriptions are correct, then we must also add text saying how do_length changes its stateT argument.

Proposed resolution:

In 22.2.1.5 [lib.locale.codecvt], and also in 22.2.1.6 [lib.locale.codecvt.byname], change the stateT argument type on both member length() and member do_length() from

const stateT&

to

stateT&

In 22.2.1.5.2 [lib.locale.codecvt.virtuals], add to the definition for member do_length a paragraph:

Effects: The effect on the state argument is ``as if'' it called do_in(state, from, from_end, from, to, to+max, to) for to pointing to a buffer of at least max elements.


76. Can a codecvt facet always convert one internal character at a time?

Section: 22.2.1.5 [lib.locale.codecvt]  Status: WP  Submitter: Matt Austern  Date: 25 Sep 1998

This issue concerns the requirements on classes derived from codecvt, including user-defined classes. What are the restrictions on the conversion from external characters (e.g. char) to internal characters (e.g. wchar_t)? Or, alternatively, what assumptions about codecvt facets can the I/O library make?

The question is whether it's possible to convert from internal characters to external characters one internal character at a time, and whether, given a valid sequence of external characters, it's possible to pick off internal characters one at a time. Or, to put it differently: given a sequence of external characters and the corresponding sequence of internal characters, does a position in the internal sequence correspond to some position in the external sequence?

To make this concrete, suppose that [first, last) is a sequence of M external characters and that [ifirst, ilast) is the corresponding sequence of N internal characters, where N > 1. That is, my_encoding.in(), applied to [first, last), yields [ifirst, ilast). Now the question: does there necessarily exist a subsequence of external characters, [first, last_1), such that the corresponding sequence of internal characters is the single character *ifirst?

(What a "no" answer would mean is that my_encoding translates sequences only as blocks. There's a sequence of M external characters that maps to a sequence of N internal characters, but that external sequence has no subsequence that maps to N-1 internal characters.)

Some of the wording in the standard, such as the description of codecvt::do_max_length (22.2.1.5.2 [lib.locale.codecvt.virtuals], paragraph 11) and basic_filebuf::underflow (27.8.1.4 [lib.filebuf.virtuals], paragraph 3) suggests that it must always be possible to pick off internal characters one at a time from a sequence of external characters. However, this is never explicitly stated one way or the other.

This issue seems (and is) quite technical, but it is important if we expect users to provide their own encoding facets. This is an area where the standard library calls user-supplied code, so a well-defined set of requirements for the user-supplied code is crucial. Users must be aware of the assumptions that the library makes. This issue affects positioning operations on basic_filebuf, unbuffered input, and several of codecvt's member functions.

Proposed resolution:

Add the following text as a new paragraph, following 22.2.1.5.2 [lib.locale.codecvt.virtuals] paragraph 2:

A codecvt facet that is used by basic_filebuf (27.8 [lib.file.streams]) must have the property that if

    do_out(state, from, from_end, from_next, to, to_lim, to_next)
would return ok, where from != from_end, then
    do_out(state, from, from + 1, from_next, to, to_end, to_next)
must also return ok, and that if
    do_in(state, from, from_end, from_next, to, to_lim, to_next)
would return ok, where to != to_lim, then
    do_in(state, from, from_end, from_next, to, to + 1, to_next)

must also return ok. [Footnote: Informally, this means that basic_filebuf assumes that the mapping from internal to external characters is 1 to N: a codecvt that is used by basic_filebuf must be able to translate characters one internal character at a time. --End Footnote]

[Redmond: Minor change in proposed resolution. Original proposed resolution talked about "success", with a parenthetical comment that success meant returning ok. New wording removes all talk about "success", and just talks about the return value.]

Rationale:

The proposed resoluion says that conversions can be performed one internal character at a time. This rules out some encodings that would otherwise be legal. The alternative answer would mean there would be some internal positions that do not correspond to any external file position.

An example of an encoding that this rules out is one where the internT and externT are of the same type, and where the internal sequence c1 c2 corresponds to the external sequence c2 c1.

It was generally agreed that basic_filebuf relies on this property: it was designed under the assumption that the external-to-internal mapping is N-to-1, and it is not clear that basic_filebuf is implementable without that restriction.

The proposed resolution is expressed as a restriction on codecvt when used by basic_filebuf, rather than a blanket restriction on all codecvt facets, because basic_filebuf is the only other part of the library that uses codecvt. If a user wants to define a codecvt facet that implements a more general N-to-M mapping, there is no reason to prohibit it, so long as the user does not expect basic_filebuf to be able to use it.


78. Typo: event_call_back

Section: 27.4.2 [lib.ios.base]  Status: TC  Submitter: Nico Josuttis  Date: 29 Sep 1998

typo: event_call_back should be event_callback  

Proposed resolution:

In the 27.4.2 [lib.ios.base] synopsis change "event_call_back" to "event_callback".


79. Inconsistent declaration of polar()

Section: 26.2.1 [lib.complex.synopsis], 26.2.7 [lib.complex.value.ops]  Status: TC  Submitter: Nico Josuttis  Date: 29 Sep 1998

In 26.2.1 [lib.complex.synopsis] polar is declared as follows:

   template<class T> complex<T> polar(const T&, const T&); 

In 26.2.7 [lib.complex.value.ops] it is declared as follows:

   template<class T> complex<T> polar(const T& rho, const T& theta = 0); 

Thus whether the second parameter is optional is not clear.

Proposed resolution:

In 26.2.1 [lib.complex.synopsis] change:

   template<class T> complex<T> polar(const T&, const T&);

to:

   template<class T> complex<T> polar(const T& rho, const T& theta = 0); 

80. Global Operators of complex declared twice

Section: 26.2.1 [lib.complex.synopsis], 26.2.2 [lib.complex]  Status: TC  Submitter: Nico Josuttis  Date: 29 Sep 1998

Both 26.2.1 and 26.2.2 contain declarations of global operators for class complex. This redundancy should be removed.

Proposed resolution:

Reduce redundancy according to the general style of the standard.


83. String::npos vs. string::max_size()

Section: 21.3 [lib.basic.string]  Status: TC  Submitter: Nico Josuttis  Date: 29 Sep 1998

Many string member functions throw if size is getting or exceeding npos. However, I wonder why they don't throw if size is getting or exceeding max_size() instead of npos. May be npos is known at compile time, while max_size() is known at runtime. However, what happens if size exceeds max_size() but not npos, then? It seems the standard lacks some clarifications here.

Proposed resolution:

After 21.3 [lib.basic.string] paragraph 4 ("The functions described in this clause...") add a new paragraph:

For any string operation, if as a result of the operation, size() would exceed max_size() then the operation throws length_error.

Rationale:

The LWG believes length_error is the correct exception to throw.


86. String constructors don't describe exceptions

Section: 21.3.1 [lib.string.cons]  Status: TC  Submitter: Nico Josuttis  Date: 29 Sep 1998

The constructor from a range:

template<class InputIterator> 
         basic_string(InputIterator begin, InputIterator end, 
                      const Allocator& a = Allocator());

lacks a throws clause. However, I would expect that it throws according to the other constructors if the numbers of characters in the range equals npos (or exceeds max_size(), see above).

Proposed resolution:

In 21.3.1 [lib.string.cons], Strike throws paragraphs for constructors which say "Throws: length_error if n == npos."

Rationale:

Throws clauses for length_error if n == npos are no longer needed because they are subsumed by the general wording added by the resolution for issue 83.


90. Incorrect description of operator >> for strings

Section: 21.3.7.9 [lib.string.io]  Status: TC  Submitter: Nico Josuttis  Date: 29 Sep 1998

The effect of operator >> for strings contain the following item:

    isspace(c,getloc()) is true for the next available input character c.

Here getloc() has to be replaced by is.getloc().

Proposed resolution:

In 21.3.7.9 [lib.string.io] paragraph 1 Effects clause replace:

isspace(c,getloc()) is true for the next available input character c.

with:

isspace(c,is.getloc()) is true for the next available input character c.


91. Description of operator>> and getline() for string<> might cause endless loop

Section: 21.3.7.9 [lib.string.io]  Status: WP  Submitter: Nico Josuttis  Date: 29 Sep 1998

Operator >> and getline() for strings read until eof() in the input stream is true. However, this might never happen, if the stream can't read anymore without reaching EOF. So shouldn't it be changed into that it reads until !good() ?

Proposed resolution:

In 21.3.7.9 [lib.string.io], paragraph 1, replace:

Effects: Begins by constructing a sentry object k as if k were constructed by typename basic_istream<charT,traits>::sentry k( is). If bool( k) is true, it calls str.erase() and then extracts characters from is and appends them to str as if by calling str.append(1, c). If is.width() is greater than zero, the maximum number n of characters appended is is.width(); otherwise n is str.max_size(). Characters are extracted and appended until any of the following occurs:

with:

Effects: Behaves as a formatted input function (27.6.1.2.1 [lib.istream.formatted.reqmts]). After constructing a sentry object, if the sentry converts to true, calls str.erase() and then extracts characters from is and appends them to str as if by calling str.append(1,c). If is.width() is greater than zero, the maximum number n of characters appended is is.width(); otherwise n is str.max_size(). Characters are extracted and appended until any of the following occurs:

In 21.3.7.9 [lib.string.io], paragraph 6, replace

Effects: Begins by constructing a sentry object k as if by typename basic_istream<charT,traits>::sentry k( is, true). If bool( k) is true, it calls str.erase() and then extracts characters from is and appends them to str as if by calling str.append(1, c) until any of the following occurs:

with:

Effects: Behaves as an unformatted input function (27.6.1.3 [lib.istream.unformatted]), except that it does not affect the value returned by subsequent calls to basic_istream<>::gcount(). After constructing a sentry object, if the sentry converts to true, calls str.erase() and then extracts characters from is and appends them to str as if by calling str.append(1,c) until any of the following occurs:

[Redmond: Made changes in proposed resolution. operator>> should be a formatted input function, not an unformatted input function. getline should not be required to set gcount, since there is no mechanism for gcount to be set except by one of basic_istream's member functions.]

[Curaçao: Nico agrees with proposed resolution.]

Rationale:

The real issue here is whether or not these string input functions get their characters from a streambuf, rather than by calling an istream's member functions, a streambuf signals failure either by returning eof or by throwing an exception; there are no other possibilities. The proposed resolution makes it clear that these two functions do get characters from a streambuf.


92. Incomplete Algorithm Requirements

Section: 25 [lib.algorithms]  Status: WP  Submitter: Nico Josuttis  Date: 29 Sep 1998

The standard does not state, how often a function object is copied, called, or the order of calls inside an algorithm. This may lead to surprising/buggy behavior. Consider the following example:

class Nth {    // function object that returns true for the nth element 
  private: 
    int nth;     // element to return true for 
    int count;   // element counter 
  public: 
    Nth (int n) : nth(n), count(0) { 
    } 
    bool operator() (int) { 
        return ++count == nth; 
    } 
}; 
.... 
// remove third element 
    list<int>::iterator pos; 
    pos = remove_if(coll.begin(),coll.end(),  // range 
                    Nth(3)),                  // remove criterion 
    coll.erase(pos,coll.end()); 

This call, in fact removes the 3rd AND the 6th element. This happens because the usual implementation of the algorithm copies the function object internally:

template <class ForwIter, class Predicate> 
ForwIter std::remove_if(ForwIter beg, ForwIter end, Predicate op) 
{ 
    beg = find_if(beg, end, op); 
    if (beg == end) { 
        return beg; 
    } 
    else { 
        ForwIter next = beg; 
        return remove_copy_if(++next, end, beg, op); 
    } 
} 

The algorithm uses find_if() to find the first element that should be removed. However, it then uses a copy of the passed function object to process the resulting elements (if any). Here, Nth is used again and removes also the sixth element. This behavior compromises the advantage of function objects being able to have a state. Without any cost it could be avoided (just implement it directly instead of calling find_if()).

Proposed resolution:

Add a new paragraph following 25 [lib.algorithms] paragraph 8:

[Note: Unless otherwise specified, algorithms that take function objects as arguments are permitted to copy those function objects freely. Programmers for whom object identity is important should consider using a wrapper class that points to a noncopied implementation object, or some equivalent solution.]

[Dublin: Pete Becker felt that this may not be a defect, but rather something that programmers need to be educated about. There was discussion of adding wording to the effect that the number and order of calls to function objects, including predicates, not affect the behavior of the function object.]

[Pre-Kona: Nico comments: It seems the problem is that we don't have a clear statement of "predicate" in the standard. People including me seemed to think "a function returning a Boolean value and being able to be called by an STL algorithm or be used as sorting criterion or ... is a predicate". But a predicate has more requirements: It should never change its behavior due to a call or being copied. IMHO we have to state this in the standard. If you like, see section 8.1.4 of my library book for a detailed discussion.]

[Kona: Nico will provide wording to the effect that "unless otherwise specified, the number of copies of and calls to function objects by algorithms is unspecified".  Consider placing in 25 [lib.algorithms] after paragraph 9.]

[Santa Cruz: The standard doesn't currently guarantee that functions object won't be copied, and what isn't forbidden is allowed. It is believed (especially since implementations that were written in concert with the standard do make copies of function objects) that this was intentional. Thus, no normative change is needed. What we should put in is a non-normative note suggesting to programmers that if they want to guarantee the lack of copying they should use something like the ref wrapper.]

[Oxford: Matt provided wording.]


98. Input iterator requirements are badly written

Section: 24.1.1 [lib.input.iterators]  Status: WP  Submitter: AFNOR  Date: 7 Oct 1998

Table 72 in 24.1.1 [lib.input.iterators] specifies semantics for *r++ of:

   { T tmp = *r; ++r; return tmp; }

There are two problems with this. First, the return type is specified to be "T", as opposed to something like "convertible to T". This is too specific: we want to allow *r++ to return an lvalue.

Second, writing the semantics in terms of code misleadingly suggests that the effects *r++ should precisely replicate the behavior of this code, including side effects. (Does this mean that *r++ should invoke the copy constructor exactly as many times as the sample code above would?) See issue 334 for a similar problem.

Proposed resolution:

In Table 72 in 24.1.1 [lib.input.iterators], change the return type for *r++ from T to "convertible to T".

Rationale:

This issue has two parts: the return type, and the number of times the copy constructor is invoked.

The LWG believes the the first part is a real issue. It's inappropriate for the return type to be specified so much more precisely for *r++ than it is for *r. In particular, if r is of (say) type int*, then *r++ isn't int, but int&.

The LWG does not believe that the number of times the copy constructor is invoked is a real issue. This can vary in any case, because of language rules on copy constructor elision. That's too much to read into these semantics clauses.

Additionally, as Dave Abrahams pointed out (c++std-lib-13703): since we're told (24.1/3) that forward iterators satisfy all the requirements of input iterators, we can't impose any requirements in the Input Iterator requirements table that forward iterators don't satisfy.


103. set::iterator is required to be modifiable, but this allows modification of keys

Section: 23.1.2 [lib.associative.reqmts]  Status: WP  Submitter: AFNOR  Date: 7 Oct 1998

Set::iterator is described as implementation-defined with a reference to the container requirement; the container requirement says that const_iterator is an iterator pointing to const T and iterator an iterator pointing to T.

23.1.2 paragraph 2 implies that the keys should not be modified to break the ordering of elements. But that is not clearly specified. Especially considering that the current standard requires that iterator for associative containers be different from const_iterator. Set, for example, has the following:

typedef implementation defined iterator;
       // See _lib.container.requirements_

23.1 [lib.container.requirements] actually requires that iterator type pointing to T (table 65). Disallowing user modification of keys by changing the standard to require an iterator for associative container to be the same as const_iterator would be overkill since that will unnecessarily significantly restrict the usage of associative container. A class to be used as elements of set, for example, can no longer be modified easily without either redesigning the class (using mutable on fields that have nothing to do with ordering), or using const_cast, which defeats requiring iterator to be const_iterator. The proposed solution goes in line with trusting user knows what he is doing.

Other Options Evaluated:

Option A.   In 23.1.2 [lib.associative.reqmts], paragraph 2, after first sentence, and before "In addition,...", add one line:

Modification of keys shall not change their strict weak ordering.

Option B. Add three new sentences to 23.1.2 [lib.associative.reqmts]:

At the end of paragraph 5: "Keys in an associative container are immutable." At the end of paragraph 6: "For associative containers where the value type is the same as the key type, both iterator and const_iterator are constant iterators. It is unspecified whether or not iterator and const_iterator are the same type."

Option C. To 23.1.2 [lib.associative.reqmts], paragraph 3, which currently reads:

The phrase ``equivalence of keys'' means the equivalence relation imposed by the comparison and not the operator== on keys. That is, two keys k1 and k2 in the same container are considered to be equivalent if for the comparison object comp, comp(k1, k2) == false && comp(k2, k1) == false.

  add the following:

For any two keys k1 and k2 in the same container, comp(k1, k2) shall return the same value whenever it is evaluated. [Note: If k2 is removed from the container and later reinserted, comp(k1, k2) must still return a consistent value but this value may be different than it was the first time k1 and k2 were in the same container. This is intended to allow usage like a string key that contains a filename, where comp compares file contents; if k2 is removed, the file is changed, and the same k2 (filename) is reinserted, comp(k1, k2) must again return a consistent value but this value may be different than it was the previous time k2 was in the container.]

Proposed resolution:

Add the following to 23.1.2 [lib.associative.reqmts] at the indicated location:

At the end of paragraph 3: "For any two keys k1 and k2 in the same container, calling comp(k1, k2) shall always return the same value."

At the end of paragraph 5: "Keys in an associative container are immutable."

At the end of paragraph 6: "For associative containers where the value type is the same as the key type, both iterator and const_iterator are constant iterators. It is unspecified whether or not iterator and const_iterator are the same type."

Rationale:

Several arguments were advanced for and against allowing set elements to be mutable as long as the ordering was not effected. The argument which swayed the LWG was one of safety; if elements were mutable, there would be no compile-time way to detect of a simple user oversight which caused ordering to be modified. There was a report that this had actually happened in practice, and had been painful to diagnose. If users need to modify elements, it is possible to use mutable members or const_cast.

Simply requiring that keys be immutable is not sufficient, because the comparison object may indirectly (via pointers) operate on values outside of the keys.

The types iterator and const_iterator are permitted to be different types to allow for potential future work in which some member functions might be overloaded between the two types. No such member functions exist now, and the LWG believes that user functionality will not be impaired by permitting the two types to be the same. A function that operates on both iterator types can be defined for const_iterator alone, and can rely on the automatic conversion from iterator to const_iterator.

[Tokyo: The LWG crafted the proposed resolution and rationale.]


106. Numeric library private members are implementation defined

Section: 26.3.5 [lib.template.slice.array]  Status: TC  Submitter: AFNOR  Date: 7 Oct 1998

This is the only place in the whole standard where the implementation has to document something private.

Proposed resolution:

Remove the comment which says "// remainder implementation defined" from:


108. Lifetime of exception::what() return unspecified

Section: 18.6.1 [lib.exception]  Status: TC  Submitter: AFNOR  Date: 7 Oct 1998

In 18.6.1, paragraphs 8-9, the lifetime of the return value of exception::what() is left unspecified. This issue has implications with exception safety of exception handling: some exceptions should not throw bad_alloc.

Proposed resolution:

Add to 18.6.1 [lib.exception] paragraph 9 (exception::what notes clause) the sentence:

The return value remains valid until the exception object from which it is obtained is destroyed or a non-const member function of the exception object is called.

Rationale:

If an exception object has non-const members, they may be used to set internal state that should affect the contents of the string returned by what().


109. Missing binders for non-const sequence elements

Section: 20.3.6 [lib.binders]  Status: WP  Submitter: Bjarne Stroustrup  Date: 7 Oct 1998

There are no versions of binders that apply to non-const elements of a sequence. This makes examples like for_each() using bind2nd() on page 521 of "The C++ Programming Language (3rd)" non-conforming. Suitable versions of the binders need to be added.

Further discussion from Nico:

What is probably meant here is shown in the following example:

class Elem { 
  public: 
    void print (int i) const { } 
    void modify (int i) { } 
}; 
int main() 
{ 
    vector<Elem> coll(2); 
    for_each (coll.begin(), coll.end(), bind2nd(mem_fun_ref(&Elem::print),42));    // OK 
    for_each (coll.begin(), coll.end(), bind2nd(mem_fun_ref(&Elem::modify),42));   // ERROR 
}

The error results from the fact that bind2nd() passes its first argument (the argument of the sequence) as constant reference. See the following typical implementation:

template <class Operation> 
class binder2nd 
  : public unary_function<typename Operation::first_argument_type, 
                          typename Operation::result_type> { 
protected: 
  Operation op; 
  typename Operation::second_argument_type value; 
public: 
  binder2nd(const Operation& o, 
            const typename Operation::second_argument_type& v) 
      : op(o), value(v) {} 
 typename Operation::result_type 
  operator()(const typename Operation::first_argument_type& x) const { 
    return op(x, value); 
  } 
};

The solution is to overload operator () of bind2nd for non-constant arguments:

template <class Operation> 
class binder2nd 
  : public unary_function<typename Operation::first_argument_type, 
                          typename Operation::result_type> { 
protected: 
  Operation op; 
  typename Operation::second_argument_type value; 
public: 
  binder2nd(const Operation& o, 
            const typename Operation::second_argument_type& v) 
      : op(o), value(v) {} 
 typename Operation::result_type 
  operator()(const typename Operation::first_argument_type& x) const { 
    return op(x, value); 
  } 
  typename Operation::result_type 
  operator()(typename Operation::first_argument_type& x) const { 
    return op(x, value); 
  } 
};

Proposed resolution:

Howard believes there is a flaw in this resolution. See c++std-lib-9127. We may need to reopen this issue.

In 20.3.6.1 [lib.binder.1st] in the declaration of binder1st after:

typename Operation::result_type
 operator()(const typename Operation::second_argument_type& x) const;

insert:

typename Operation::result_type
 operator()(typename Operation::second_argument_type& x) const;

In 20.3.6.3 [lib.binder.2nd] in the declaration of binder2nd after:

typename Operation::result_type
 operator()(const typename Operation::first_argument_type& x) const;

insert:

typename Operation::result_type
 operator()(typename Operation::first_argument_type& x) const;

[Kona: The LWG discussed this at some length.It was agreed that this is a mistake in the design, but there was no consensus on whether it was a defect in the Standard. Straw vote: NAD - 5. Accept proposed resolution - 3. Leave open - 6.]

[Copenhagen: It was generally agreed that this was a defect. Strap poll: NAD - 0. Accept proposed resolution - 10. Leave open - 1.]


110. istreambuf_iterator::equal not const

Section: 24.5.3 [lib.istreambuf.iterator], 24.5.3.5 [lib.istreambuf.iterator::equal]  Status: TC  Submitter: Nathan Myers  Date: 15 Oct 1998

Member istreambuf_iterator<>::equal is not declared "const", yet 24.5.3.6 [lib.istreambuf.iterator::op==] says that operator==, which is const, calls it. This is contradictory.

Proposed resolution:

In 24.5.3 [lib.istreambuf.iterator] and also in 24.5.3.5 [lib.istreambuf.iterator::equal], replace:

bool equal(istreambuf_iterator& b);

with:

bool equal(const istreambuf_iterator& b) const;

112. Minor typo in ostreambuf_iterator constructor

Section: 24.5.4.1 [lib.ostreambuf.iter.cons]  Status: TC  Submitter: Matt Austern  Date: 20 Oct 1998

The requires clause for ostreambuf_iterator's constructor from an ostream_type (24.5.4.1, paragraph 1) reads "s is not null". However, s is a reference, and references can't be null.

Proposed resolution:

In 24.5.4.1 [lib.ostreambuf.iter.cons]:

Move the current paragraph 1, which reads "Requires: s is not null.", from the first constructor to the second constructor.

Insert a new paragraph 1 Requires clause for the first constructor reading:

Requires: s.rdbuf() is not null.


114. Placement forms example in error twice

Section: 18.4.1.3 [lib.new.delete.placement]  Status: TC  Submitter: Steve Clamage  Date: 28 Oct 1998

Section 18.4.1.3 contains the following example:

[Example: This can be useful for constructing an object at a known address:
        char place[sizeof(Something)];
        Something* p = new (place) Something();
 -end example]

First code line: "place" need not have any special alignment, and the following constructor could fail due to misaligned data.

Second code line: Aren't the parens on Something() incorrect?  [Dublin: the LWG believes the () are correct.]

Examples are not normative, but nevertheless should not show code that is invalid or likely to fail.

Proposed resolution:

Replace the first line of code in the example in 18.4.1.3 [lib.new.delete.placement] with:

void* place = operator new(sizeof(Something));

115. Typo in strstream constructors

Section: D.7.4.1 [depr.strstream.cons]  Status: TC  Submitter: Steve Clamage  Date: 2 Nov 1998

D.7.4.1 strstream constructors paragraph 2 says:

Effects: Constructs an object of class strstream, initializing the base class with iostream(& sb) and initializing sb with one of the two constructors:

- If mode&app==0, then s shall designate the first element of an array of n elements. The constructor is strstreambuf(s, n, s).

- If mode&app==0, then s shall designate the first element of an array of n elements that contains an NTBS whose first element is designated by s. The constructor is strstreambuf(s, n, s+std::strlen(s)).

Notice the second condition is the same as the first. I think the second condition should be "If mode&app==app", or "mode&app!=0", meaning that the append bit is set.

Proposed resolution:

In D.7.3.1 [depr.ostrstream.cons] paragraph 2 and D.7.4.1 [depr.strstream.cons] paragraph 2, change the first condition to (mode&app)==0 and the second condition to (mode&app)!=0.


117. basic_ostream uses nonexistent num_put member functions

Section: 27.6.2.5.2 [lib.ostream.inserters.arithmetic]  Status: WP  Submitter: Matt Austern  Date: 20 Nov 1998

The effects clause for numeric inserters says that insertion of a value x, whose type is either bool, short, unsigned short, int, unsigned int, long, unsigned long, float, double, long double, or const void*, is delegated to num_put, and that insertion is performed as if through the following code fragment:

bool failed = use_facet<
   num_put<charT,ostreambuf_iterator<charT,traits> > 
   >(getloc()).put(*this, *this, fill(), val). failed();

This doesn't work, because num_put<>::put is only overloaded for the types bool, long, unsigned long, double, long double, and const void*. That is, the code fragment in the standard is incorrect (it is diagnosed as ambiguous at compile time) for the types short, unsigned short, int, unsigned int, and float.

We must either add new member functions to num_put, or else change the description in ostream so that it only calls functions that are actually there. I prefer the latter.

Proposed resolution:

Replace 27.6.2.5.2, paragraph 1 with the following:

The classes num_get<> and num_put<> handle locale­dependent numeric formatting and parsing. These inserter functions use the imbued locale value to perform numeric formatting. When val is of type bool, long, unsigned long, double, long double, or const void*, the formatting conversion occurs as if it performed the following code fragment:

bool failed = use_facet<
   num_put<charT,ostreambuf_iterator<charT,traits> >
   >(getloc()).put(*this, *this, fill(), val). failed();

When val is of type short the formatting conversion occurs as if it performed the following code fragment:

ios_base::fmtflags baseflags = ios_base::flags() & ios_base::basefield;
bool failed = use_facet<
   num_put<charT,ostreambuf_iterator<charT,traits> >
   >(getloc()).put(*this, *this, fill(),
      baseflags == ios_base::oct || baseflags == ios_base::hex
         ? static_cast<long>(static_cast<unsigned short>(val))
         : static_cast<long>(val)). failed();

When val is of type int the formatting conversion occurs as if it performed the following code fragment:

ios_base::fmtflags baseflags = ios_base::flags() & ios_base::basefield;
bool failed = use_facet<
   num_put<charT,ostreambuf_iterator<charT,traits> >
   >(getloc()).put(*this, *this, fill(),
      baseflags == ios_base::oct || baseflags == ios_base::hex
         ? static_cast<long>(static_cast<unsigned int>(val))
         : static_cast<long>(val)). failed();

When val is of type unsigned short or unsigned int the formatting conversion occurs as if it performed the following code fragment:

bool failed = use_facet<
   num_put<charT,ostreambuf_iterator<charT,traits> >
   >(getloc()).put(*this, *this, fill(), static_cast<unsigned long>(val)).
failed();

When val is of type float the formatting conversion occurs as if it performed the following code fragment:

bool failed = use_facet<
   num_put<charT,ostreambuf_iterator<charT,traits> >
   >(getloc()).put(*this, *this, fill(), static_cast<double>(val)).
failed();

[post-Toronto: This differs from the previous proposed resolution; PJP provided the new wording. The differences are in signed short and int output.]

Rationale:

The original proposed resolution was to cast int and short to long, unsigned int and unsigned short to unsigned long, and float to double, thus ensuring that we don't try to use nonexistent num_put<> member functions. The current proposed resolution is more complicated, but gives more expected results for hex and octal output of signed short and signed int. (On a system with 16-bit short, for example, printing short(-1) in hex format should yield 0xffff.)


118. basic_istream uses nonexistent num_get member functions

Section: 27.6.1.2.2 [lib.istream.formatted.arithmetic]  Status: WP  Submitter: Matt Austern  Date: 20 Nov 1998

Formatted input is defined for the types short, unsigned short, int, unsigned int, long, unsigned long, float, double, long double, bool, and void*. According to section 27.6.1.2.2, formatted input of a value x is done as if by the following code fragment:

typedef num_get< charT,istreambuf_iterator<charT,traits> > numget; 
iostate err = 0; 
use_facet< numget >(loc).get(*this, 0, *this, err, val); 
setstate(err);

According to section 22.2.2.1.1 [lib.facet.num.get.members], however, num_get<>::get() is only overloaded for the types bool, long, unsigned short, unsigned int, unsigned long, unsigned long, float, double, long double, and void*. Comparing the lists from the two sections, we find that 27.6.1.2.2 is using a nonexistent function for types short and int.

Proposed resolution:

In 27.6.1.2.2 [lib.istream.formatted.arithmetic] Arithmetic Extractors, remove the two lines (1st and 3rd) which read:

operator>>(short& val);
...
operator>>(int& val);

And add the following at the end of that section (27.6.1.2.2) :

operator>>(short& val);

The conversion occurs as if performed by the following code fragment (using the same notation as for the preceding code fragment):

  typedef num_get< charT,istreambuf_iterator<charT,traits> > numget;
  iostate err = 0;
  long lval;
  use_facet< numget >(loc).get(*this, 0, *this, err, lval);
        if (err == 0
                && (lval < numeric_limits<short>::min() || numeric_limits<short>::max() < lval))
                err = ios_base::failbit;
  setstate(err);
operator>>(int& val);

The conversion occurs as if performed by the following code fragment (using the same notation as for the preceding code fragment):

  typedef num_get< charT,istreambuf_iterator<charT,traits> > numget;
  iostate err = 0;
  long lval;
  use_facet< numget >(loc).get(*this, 0, *this, err, lval);
        if (err == 0
                && (lval < numeric_limits<int>::min() || numeric_limits<int>::max() < lval))
                err = ios_base::failbit;
  setstate(err);

[Post-Tokyo: PJP provided the above wording.]


119. Should virtual functions be allowed to strengthen the exception specification?

Section: 17.4.4.8 [lib.res.on.exception.handling]  Status: TC  Submitter: Judy Ward  Date: 15 Dec 1998

Section 17.4.4.8 [lib.res.on.exception.handling] states:

"An implementation may strengthen the exception-specification for a function by removing listed exceptions."

The problem is that if an implementation is allowed to do this for virtual functions, then a library user cannot write a class that portably derives from that class.

For example, this would not compile if ios_base::failure::~failure had an empty exception specification:

#include <ios>
#include <string>

class D : public std::ios_base::failure {
public:
        D(const std::string&);
        ~D(); // error - exception specification must be compatible with 
              // overridden virtual function ios_base::failure::~failure()
};

Proposed resolution:

Change Section 17.4.4.8 [lib.res.on.exception.handling] from:

     "may strengthen the exception-specification for a function"

to:

     "may strengthen the exception-specification for a non-virtual function".


120. Can an implementor add specializations?

Section: 17.4.3.1 [lib.reserved.names]  Status: WP  Submitter: Judy Ward  Date: 15 Dec 1998

The original issue asked whether a library implementor could specialize standard library templates for built-in types. (This was an issue because users are permitted to explicitly instantiate standard library templates.)

Specializations are no longer a problem, because of the resolution to core issue 259. Under the proposed resolution, it will be legal for a translation unit to contain both a specialization and an explicit instantiation of the same template, provided that the specialization comes first. In such a case, the explicit instantiation will be ignored. Further discussion of library issue 120 assumes that the core 259 resolution will be adopted.

However, as noted in lib-7047, one piece of this issue still remains: what happens if a standard library implementor explicitly instantiates a standard library templates? It's illegal for a program to contain two different explicit instantiations of the same template for the same type in two different translation units (ODR violation), and the core working group doesn't believe it is practical to relax that restriction.

The issue, then, is: are users allowed to explicitly instantiate standard library templates for non-user defined types? The status quo answer is 'yes'. Changing it to 'no' would give library implementors more freedom.

This is an issue because, for performance reasons, library implementors often need to explicitly instantiate standard library templates. (for example, std::basic_string<char>) Does giving users freedom to explicitly instantiate standard library templates for non-user defined types make it impossible or painfully difficult for library implementors to do this?

John Spicer suggests, in lib-8957, that library implementors have a mechanism they can use for explicit instantiations that doesn't prevent users from performing their own explicit instantiations: put each explicit instantiation in its own object file. (Different solutions might be necessary for Unix DSOs or MS-Windows DLLs.) On some platforms, library implementors might not need to do anything special: the "undefined behavior" that results from having two different explicit instantiations might be harmless.

Proposed resolution:

Append to 17.4.3.1 [lib.reserved.names] paragraph 1:

A program may explicitly instantiate any templates in the standard library only if the declaration depends on the name of a user-defined type of external linkage and the instantiation meets the standard library requirements for the original template.

[Kona: changed the wording from "a user-defined name" to "the name of a user-defined type"]

Rationale:

The LWG considered another possible resolution:

In light of the resolution to core issue 259, no normative changes in the library clauses are necessary. Add the following non-normative note to the end of 17.4.3.1 [lib.reserved.names] paragraph 1:

[Note: A program may explicitly instantiate standard library templates, even when an explicit instantiation does not depend on a user-defined name. --end note]

The LWG rejected this because it was believed that it would make it unnecessarily difficult for library implementors to write high-quality implementations. A program may not include an explicit instantiation of the same template, for the same template arguments, in two different translation units. If users are allowed to provide explicit instantiations of Standard Library templates for built-in types, then library implementors aren't, at least not without nonportable tricks.

The most serious problem is a class template that has writeable static member variables. Unfortunately, such class templates are important and, in existing Standard Library implementations, are often explicitly specialized by library implementors: locale facets, which have a writeable static member variable id. If a user's explicit instantiation collided with the implementations explicit instantiation, iostream initialization could cause locales to be constructed in an inconsistent state.

One proposed implementation technique was for Standard Library implementors to provide explicit instantiations in separate object files, so that they would not be picked up by the linker when the user also provides an explicit instantiation. However, this technique only applies for Standard Library implementations that are packaged as static archives. Most Standard Library implementations nowadays are packaged as dynamic libraries, so this technique would not apply.

The Committee is now considering standardization of dynamic linking. If there are such changes in the future, it may be appropriate to revisit this issue later.


122. streambuf/wstreambuf description should not say they are specializations

Section: 27.5.2 [lib.streambuf]  Status: TC  Submitter: Judy Ward  Date: 15 Dec 1998

Section 27.5.2 describes the streambuf classes this way:

The class streambuf is a specialization of the template class basic_streambuf specialized for the type char.

The class wstreambuf is a specialization of the template class basic_streambuf specialized for the type wchar_t.

This implies that these classes must be template specializations, not typedefs.

It doesn't seem this was intended, since Section 27.5 has them declared as typedefs.

Proposed resolution:

Remove 27.5.2 [lib.streambuf] paragraphs 2 and 3 (the above two sentences).

Rationale:

The streambuf synopsis already has a declaration for the typedefs and that is sufficient.


123. Should valarray helper arrays fill functions be const?

Section: 26.3.5.4 [lib.slice.arr.fill], 26.3.7.4 [lib.gslice.array.fill], 26.3.8.4 [lib.mask.array.fill], 26.3.9.4 [lib.indirect.array.fill]  Status: WP  Submitter: Judy Ward  Date: 15 Dec 1998

One of the operator= in the valarray helper arrays is const and one is not. For example, look at slice_array. This operator= in Section 26.3.5.2 [lib.slice.arr.assign] is const:

    void operator=(const valarray<T>&) const;

but this one in Section 26.3.5.4 [lib.slice.arr.fill] is not:

    void operator=(const T&);

The description of the semantics for these two functions is similar.

Proposed resolution:

26.3.5 [lib.template.slice.array] Template class slice_array

In the class template definition for slice_array, replace the member function declaration

      void operator=(const T&);
    

with

      void operator=(const T&) const;
    

26.3.5.4 [lib.slice.arr.fill] slice_array fill function

Change the function declaration

      void operator=(const T&);
    

to

      void operator=(const T&) const;
    

26.3.7 [lib.template.gslice.array] Template class gslice_array

In the class template definition for gslice_array, replace the member function declaration

      void operator=(const T&);
    

with

      void operator=(const T&) const;
    

26.3.7.4 [lib.gslice.array.fill] gslice_array fill function

Change the function declaration

      void operator=(const T&);
    

to

      void operator=(const T&) const;
    

26.3.8 [lib.template.mask.array] Template class mask_array

In the class template definition for mask_array, replace the member function declaration

      void operator=(const T&);
    

with

      void operator=(const T&) const;
    

26.3.8.4 [lib.mask.array.fill] mask_array fill function

Change the function declaration

      void operator=(const T&);
    

to

      void operator=(const T&) const;
    

26.3.9 [lib.template.indirect.array] Template class indirect_array

In the class template definition for indirect_array, replace the member function declaration

      void operator=(const T&);
    

with

      void operator=(const T&) const;
    

26.3.9.4 [lib.indirect.array.fill] indirect_array fill function

Change the function declaration

      void operator=(const T&);
    

to

      void operator=(const T&) const;
    

[Redmond: Robert provided wording.]

Rationale:

There's no good reason for one version of operator= being const and another one not. Because of issue 253, this now matters: these functions are now callable in more circumstances. In many existing implementations, both versions are already const.


124. ctype_byname<charT>::do_scan_is & do_scan_not return type should be const charT*

Section: 22.2.1.2 [lib.locale.ctype.byname]  Status: TC  Submitter: Judy Ward  Date: 15 Dec 1998

In Section 22.2.1.2 [lib.locale.ctype.byname] ctype_byname<charT>::do_scan_is() and do_scan_not() are declared to return a const char* not a const charT*.

Proposed resolution:

Change Section 22.2.1.2 [lib.locale.ctype.byname] do_scan_is() and do_scan_not() to return a const charT*.


125. valarray<T>::operator!() return type is inconsistent

Section: 26.3.2 [lib.template.valarray]  Status: TC  Submitter: Judy Ward  Date: 15 Dec 1998

In Section 26.3.2 [lib.template.valarray] valarray<T>::operator!() is declared to return a valarray<T>, but in Section 26.3.2.5 [lib.valarray.unary] it is declared to return a valarray<bool>. The latter appears to be correct.

Proposed resolution:

Change in Section 26.3.2 [lib.template.valarray] the declaration of operator!() so that the return type is valarray<bool>.


126. typos in Effects clause of ctype::do_narrow()

Section: 22.2.1.1.2 [lib.locale.ctype.virtuals]  Status: TC  Submitter: Judy Ward  Date: 15 Dec 1998

Typos in 22.2.1.1.2 need to be fixed.

Proposed resolution:

In Section 22.2.1.1.2 [lib.locale.ctype.virtuals] change:

   do_widen(do_narrow(c),0) == c

to:

   do_widen(do_narrow(c,0)) == c

and change:

   (is(M,c) || !ctc.is(M, do_narrow(c),dfault) )

to:

   (is(M,c) || !ctc.is(M, do_narrow(c,dfault)) )

127. auto_ptr<> conversion issues

Section: 20.4.5 [lib.auto.ptr]  Status: TC  Submitter: Greg Colvin  Date: 17 Feb 1999

There are two problems with the current auto_ptr wording in the standard:

First, the auto_ptr_ref definition cannot be nested because auto_ptr<Derived>::auto_ptr_ref is unrelated to auto_ptr<Base>::auto_ptr_ref. Also submitted by Nathan Myers, with the same proposed resolution.

Second, there is no auto_ptr assignment operator taking an auto_ptr_ref argument.

I have discussed these problems with my proposal coauthor, Bill Gibbons, and with some compiler and library implementors, and we believe that these problems are not desired or desirable implications of the standard.

25 Aug 1999: The proposed resolution now reflects changes suggested by Dave Abrahams, with Greg Colvin's concurrence; 1) changed "assignment operator" to "public assignment operator", 2) changed effects to specify use of release(), 3) made the conversion to auto_ptr_ref const.

2 Feb 2000: Lisa Lippincott comments: [The resolution of] this issue states that the conversion from auto_ptr to auto_ptr_ref should be const. This is not acceptable, because it would allow initialization and assignment from _any_ const auto_ptr! It also introduces an implementation difficulty in writing this conversion function -- namely, somewhere along the line, a const_cast will be necessary to remove that const so that release() may be called. This may result in undefined behavior [7.1.5.1/4]. The conversion operator does not have to be const, because a non-const implicit object parameter may be bound to an rvalue [13.3.3.1.4/3] [13.3.1/5].

Tokyo: The LWG removed the following from the proposed resolution:

In 20.4.5 [lib.auto.ptr], paragraph 2, and 20.4.5.3 [lib.auto.ptr.conv], paragraph 2, make the conversion to auto_ptr_ref const:

template<class Y> operator auto_ptr_ref<Y>() const throw();

Proposed resolution:

In 20.4.5 [lib.auto.ptr], paragraph 2, move the auto_ptr_ref definition to namespace scope.

In 20.4.5 [lib.auto.ptr], paragraph 2, add a public assignment operator to the auto_ptr definition:

auto_ptr& operator=(auto_ptr_ref<X> r) throw();

Also add the assignment operator to 20.4.5.3 [lib.auto.ptr.conv]:

auto_ptr& operator=(auto_ptr_ref<X> r) throw()
Effects: Calls reset(p.release()) for the auto_ptr p that r holds a reference to.
Returns: *this.

129. Need error indication from seekp() and seekg()

Section: 27.6.1.3 [lib.istream.unformatted], 27.6.2.4 [lib.ostream.seeks]  Status: TC  Submitter: Angelika Langer  Date: 22 Feb 1999

Currently, the standard does not specify how seekg() and seekp() indicate failure. They are not required to set failbit, and they can't return an error indication because they must return *this, i.e. the stream. Hence, it is undefined what happens if they fail. And they can fail, for instance, when a file stream is disconnected from the underlying file (is_open()==false) or when a wide character file stream must perform a state-dependent code conversion, etc.

The stream functions seekg() and seekp() should set failbit in the stream state in case of failure.

Proposed resolution:

Add to the Effects: clause of  seekg() in 27.6.1.3 [lib.istream.unformatted] and to the Effects: clause of seekp() in 27.6.2.4 [lib.ostream.seeks]:

In case of failure, the function calls setstate(failbit) (which may throw ios_base::failure).

Rationale:

Setting failbit is the usual error reporting mechanism for streams


130. Return type of container::erase(iterator) differs for associative containers

Section: 23.1.2 [lib.associative.reqmts], 23.1.1 [lib.sequence.reqmts]  Status: DR  Submitter: Andrew Koenig  Date: 2 Mar 1999

Table 67 (23.1.1) says that container::erase(iterator) returns an iterator. Table 69 (23.1.2) says that in addition to this requirement, associative containers also say that container::erase(iterator) returns void. That's not an addition; it's a change to the requirements, which has the effect of making associative containers fail to meet the requirements for containers.

Proposed resolution:

In 23.1.2 [lib.associative.reqmts], in Table 69 Associative container requirements, change the return type of a.erase(q) from void to iterator. Change the assertion/not/pre/post-condition from "erases the element pointed to by q" to "erases the element pointed to by q. Returns an iterator pointing to the element immediately following q prior to the element being erased. If no such element exists, a.end() is returned."

In 23.1.2 [lib.associative.reqmts], in Table 69 Associative container requirements, change the return type of a.erase(q1, q2) from void to iterator. Change the assertion/not/pre/post-condition from "erases the elements in the range [q1, q2)" to "erases the elements in the range [q1, q2). Returns q2."

In 23.3.1 [lib.map], in the map class synopsis; and in 23.3.2 [lib.multimap], in the multimap class synopsis; and in 23.3.3 [lib.set], in the set class synopsis; and in 23.3.4 [lib.multiset], in the multiset class synopsis: change the signature of the first erase overload to

   iterator erase(iterator position);

and change the signature of the third erase overload to

  iterator erase(iterator first, iterator last); 

[Pre-Kona: reopened at the request of Howard Hinnant]

[Post-Kona: the LWG agrees the return type should be iterator, not void. (Alex Stepanov agrees too.) Matt provided wording.]

[ Sydney: the proposed wording went in the right direction, but it wasn't good enough. We want to return an iterator from the range form of erase as well as the single-iterator form. Also, the wording is slightly different from the wording we have for sequences; there's no good reason for having a difference. Matt provided new wording, which we will review at the next meeting. ]


132. list::resize description uses random access iterators

Section: 23.2.2.2 [lib.list.capacity]  Status: TC  Submitter: Howard Hinnant  Date: 6 Mar 1999

The description reads:

-1- Effects:

         if (sz > size())
           insert(end(), sz-size(), c);
         else if (sz < size())
           erase(begin()+sz, end());
         else
           ;                           //  do nothing

Obviously list::resize should not be specified in terms of random access iterators.

Proposed resolution:

Change 23.2.2.2 paragraph 1 to:

Effects:

         if (sz > size())
           insert(end(), sz-size(), c);
         else if (sz < size())
         {
           iterator i = begin();
           advance(i, sz);
           erase(i, end());
         }

[Dublin: The LWG asked Howard to discuss exception safety offline with David Abrahams. They had a discussion and believe there is no issue of exception safety with the proposed resolution.]


133. map missing get_allocator()

Section: 23.3.1 [lib.map]  Status: TC  Submitter: Howard Hinnant  Date: 6 Mar 1999

The title says it all.

Proposed resolution:

Insert in 23.3.1 [lib.map], paragraph 2, after operator= in the map declaration:

    allocator_type get_allocator() const;

134. vector constructors over specified

Section: 23.2.4.1 [lib.vector.cons]  Status: TC  Submitter: Howard Hinnant  Date: 6 Mar 1999

The complexity description says: "It does at most 2N calls to the copy constructor of T and logN reallocations if they are just input iterators ...".

This appears to be overly restrictive, dictating the precise memory/performance tradeoff for the implementor.

Proposed resolution:

Change 23.2.4.1 [lib.vector.cons], paragraph 1 to:

-1- Complexity: The constructor template <class InputIterator> vector(InputIterator first, InputIterator last) makes only N calls to the copy constructor of T (where N is the distance between first and last) and no reallocations if iterators first and last are of forward, bidirectional, or random access categories. It makes order N calls to the copy constructor of T and order logN reallocations if they are just input iterators.

Rationale:

"at most 2N calls" is correct only if the growth factor is greater than or equal to 2.


136. seekp, seekg setting wrong streams?

Section: 27.6.1.3 [lib.istream.unformatted]  Status: WP  Submitter: Howard Hinnant  Date: 6 Mar 1999

I may be misunderstanding the intent, but should not seekg set only the input stream and seekp set only the output stream? The description seems to say that each should set both input and output streams. If that's really the intent, I withdraw this proposal.

Proposed resolution:

In section 27.6.1.3 change:

basic_istream<charT,traits>& seekg(pos_type pos);
Effects: If fail() != true, executes rdbuf()->pubseekpos(pos). 

To:

basic_istream<charT,traits>& seekg(pos_type pos);
Effects: If fail() != true, executes rdbuf()->pubseekpos(pos, ios_base::in). 

In section 27.6.1.3 change:

basic_istream<charT,traits>& seekg(off_type& off, ios_base::seekdir dir);
Effects: If fail() != true, executes rdbuf()->pubseekoff(off, dir). 

To:

basic_istream<charT,traits>& seekg(off_type& off, ios_base::seekdir dir);
Effects: If fail() != true, executes rdbuf()->pubseekoff(off, dir, ios_base::in). 

In section 27.6.2.4, paragraph 2 change:

-2- Effects: If fail() != true, executes rdbuf()->pubseekpos(pos). 

To:

-2- Effects: If fail() != true, executes rdbuf()->pubseekpos(pos, ios_base::out). 

In section 27.6.2.4, paragraph 4 change:

-4- Effects: If fail() != true, executes rdbuf()->pubseekoff(off, dir). 

To:

-4- Effects: If fail() != true, executes rdbuf()->pubseekoff(off, dir, ios_base::out). 

[Dublin: Dietmar Kühl thinks this is probably correct, but would like the opinion of more iostream experts before taking action.]

[Tokyo: Reviewed by the LWG. PJP noted that although his docs are incorrect, his implementation already implements the Proposed Resolution.]

[Post-Tokyo: Matt Austern comments:
Is it a problem with basic_istream and basic_ostream, or is it a problem with basic_stringbuf? We could resolve the issue either by changing basic_istream and basic_ostream, or by changing basic_stringbuf. I prefer the latter change (or maybe both changes): I don't see any reason for the standard to require that std::stringbuf s(std::string("foo"), std::ios_base::in); s.pubseekoff(0, std::ios_base::beg); must fail.
This requirement is a bit weird. There's no similar requirement for basic_streambuf<>::seekpos, or for basic_filebuf<>::seekoff or basic_filebuf<>::seekpos.]


137. Do use_facet and has_facet look in the global locale?

Section: 22.1.1 [lib.locale]  Status: TC  Submitter: Angelika Langer  Date: 17 Mar 1999

Section 22.1.1 [lib.locale] says:

-4- In the call to use_facet<Facet>(loc), the type argument chooses a facet, making available all members of the named type. If Facet is not present in a locale (or, failing that, in the global locale), it throws the standard exception bad_cast. A C++ program can check if a locale implements a particular facet with the template function has_facet<Facet>().

This contradicts the specification given in section 22.1.2 [lib.locale.global.templates]:

template <class  Facet> const  Facet& use_facet(const locale&  loc);

-1- Get a reference to a facet of a locale.
-2- Returns: a reference to the corresponding facet of loc, if present.
-3- Throws: bad_cast if has_facet<Facet>(loc) is false.
-4- Notes: The reference returned remains valid at least as long as any copy of loc exists

Proposed resolution:

Remove the phrase "(or, failing that, in the global locale)" from section 22.1.1.

Rationale:

Needed for consistency with the way locales are handled elsewhere in the standard.


139. Optional sequence operation table description unclear

Section: 23.1.1 [lib.sequence.reqmts]  Status: TC  Submitter: Andrew Koenig  Date: 30 Mar 1999

The sentence introducing the Optional sequence operation table (23.1.1 paragraph 12) has two problems:

A. It says ``The operations in table 68 are provided only for the containers for which they take constant time.''

That could be interpreted in two ways, one of them being ``Even though table 68 shows particular operations as being provided, implementations are free to omit them if they cannot implement them in constant time.''

B. That paragraph says nothing about amortized constant time, and it should. 

Proposed resolution:

Replace the wording in 23.1.1 paragraph 12  which begins ``The operations in table 68 are provided only..." with:

Table 68 lists sequence operations that are provided for some types of sequential containers but not others. An implementation shall provide these operations for all container types shown in the ``container'' column, and shall implement them so as to take amortized constant time.


141. basic_string::find_last_of, find_last_not_of say pos instead of xpos

Section: 21.3.6.4 [lib.string::find.last.of], 21.3.6.6 [lib.string::find.last.not.of]  Status: TC  Submitter: Arch Robison  Date: 28 Apr 1999

Sections 21.3.6.4 paragraph 1 and 21.3.6.6 paragraph 1 surely have misprints where they say:

xpos <= pos and pos < size();

Surely the document meant to say ``xpos < size()'' in both places.

[Judy Ward also sent in this issue for 21.3.6.4 with the same proposed resolution.]

Proposed resolution:

Change Sections 21.3.6.4 paragraph 1 and 21.3.6.6 paragraph 1, the line which says:

xpos <= pos and pos < size();

to:

xpos <= pos and xpos < size();


142. lexicographical_compare complexity wrong

Section: 25.3.8 [lib.alg.lex.comparison]  Status: TC  Submitter: Howard Hinnant  Date: 20 Jun 1999

The lexicographical_compare complexity is specified as:

     "At most min((last1 - first1), (last2 - first2)) applications of the corresponding comparison."

The best I can do is twice that expensive.

Nicolai Josuttis comments in lib-6862: You mean, to check for equality you have to check both < and >? Yes, IMO you are right! (and Matt states this complexity in his book)

Proposed resolution:

Change 25.3.8 [lib.alg.lex.comparison] complexity to:

At most 2*min((last1 - first1), (last2 - first2)) applications of the corresponding comparison.

Change the example at the end of paragraph 3 to read:

[Example:

    for ( ; first1 != last1 && first2 != last2 ; ++first1, ++first2) {
      if (*first1 < *first2) return true;
      if (*first2 < *first1) return false;
    }
    return first1 == last1 && first2 != last2;
   
--end example]

144. Deque constructor complexity wrong

Section: 23.2.1.1 [lib.deque.cons]  Status: TC  Submitter: Herb Sutter  Date: 9 May 1999

In 23.2.1.1 paragraph 6, the deque ctor that takes an iterator range appears to have complexity requirements which are incorrect, and which contradict the complexity requirements for insert(). I suspect that the text in question, below, was taken from vector:

Complexity: If the iterators first and last are forward iterators, bidirectional iterators, or random access iterators the constructor makes only N calls to the copy constructor, and performs no reallocations, where N is last - first.

The word "reallocations" does not really apply to deque. Further, all of the following appears to be spurious:

It makes at most 2N calls to the copy constructor of T and log N reallocations if they are input iterators.1)

1) The complexity is greater in the case of input iterators because each element must be added individually: it is impossible to determine the distance between first abd last before doing the copying.

This makes perfect sense for vector, but not for deque. Why should deque gain an efficiency advantage from knowing in advance the number of elements to insert?

Proposed resolution:

In 23.2.1.1 paragraph 6, replace the Complexity description, including the footnote, with the following text (which also corrects the "abd" typo):

Complexity: Makes last - first calls to the copy constructor of T.


146. complex<T> Inserter and Extractor need sentries

Section: 26.2.6 [lib.complex.ops]  Status: TC  Submitter: Angelika Langer  Date: 12 May 1999

The extractor for complex numbers is specified as: 

template<class T, class charT, class traits> 
basic_istream<charT, traits>& 
operator>>(basic_istream<charT, traits>& is, complex<T>& x);
 
Effects: Extracts a complex number x of the form: u, (u), or (u,v), where u is the real part and v is the imaginary part (lib.istream.formatted). 
Requires: The input values be convertible to T. If bad input is encountered, calls is.setstate(ios::failbit) (which may throw ios::failure (lib.iostate.flags). 
Returns: is .

Is it intended that the extractor for complex numbers does not skip whitespace, unlike all other extractors in the standard library do? Shouldn't a sentry be used? 

The inserter for complex numbers is specified as:

template<class T, class charT, class traits> 
basic_ostream<charT, traits>& 
operator<<(basic_ostream<charT, traits>& o, const complex<T>& x);

Effects: inserts the complex number x onto the stream o as if it were implemented as follows:

template<class T, class charT, class traits> 
basic_ostream<charT, traits>& 
operator<<(basic_ostream<charT, traits>& o, const complex<T>& x) 

basic_ostringstream<charT, traits> s; 
s.flags(o.flags()); 
s.imbue(o.getloc()); 
s.precision(o.precision()); 
s << '(' << x.real() << "," << x.imag() << ')'; 
return o << s.str(); 
}

Is it intended that the inserter for complex numbers ignores the field width and does not do any padding? If, with the suggested implementation above, the field width were set in the stream then the opening parentheses would be adjusted, but the rest not, because the field width is reset to zero after each insertion.

I think that both operations should use sentries, for sake of consistency with the other inserters and extractors in the library. Regarding the issue of padding in the inserter, I don't know what the intent was. 

Proposed resolution:

After 26.2.6 [lib.complex.ops] paragraph 14 (operator>>), add a Notes clause:

Notes: This extraction is performed as a series of simpler extractions. Therefore, the skipping of whitespace is specified to be the same for each of the simpler extractions.

Rationale:

For extractors, the note is added to make it clear that skipping whitespace follows an "all-or-none" rule.

For inserters, the LWG believes there is no defect; the standard is correct as written.


147. Library Intro refers to global functions that aren't global

Section: 17.4.4.3 [lib.global.functions]  Status: TC  Submitter: Lois Goldthwaite  Date: 4 Jun 1999

The library had many global functions until 17.4.1.1 [lib.contents] paragraph 2 was added:

All library entities except macros, operator new and operator delete are defined within the namespace std or namespaces nested within namespace std.

It appears "global function" was never updated in the following:

17.4.4.3 - Global functions [lib.global.functions]

-1- It is unspecified whether any global functions in the C++ Standard Library are defined as inline (dcl.fct.spec).

-2- A call to a global function signature described in Clauses lib.language.support through lib.input.output behaves the same as if the implementation declares no additional global function signatures.*

[Footnote: A valid C++ program always calls the expected library global function. An implementation may also define additional global functions that would otherwise not be called by a valid C++ program. --- end footnote]

-3- A global function cannot be declared by the implementation as taking additional default arguments. 

17.4.4.4 - Member functions [lib.member.functions]

-2- An implementation can declare additional non-virtual member function signatures within a class:

-- by adding arguments with default values to a member function signature; The same latitude does not extend to the implementation of virtual or global functions, however.

Proposed resolution:

Change "global" to "global or non-member" in:

17.4.4.3 [lib.global.functions] section title,
17.4.4.3 [lib.global.functions] para 1,
17.4.4.3 [lib.global.functions] para 2 in 2 places plus 2 places in the footnote,
17.4.4.3 [lib.global.functions] para 3,
17.4.4.4 [lib.member.functions] para 2

Rationale:

Because operator new and delete are global, the proposed resolution was changed from "non-member" to "global or non-member.


148. Functions in the example facet BoolNames should be const

Section: 22.2.8 [lib.facets.examples]  Status: TC  Submitter: Jeremy Siek  Date: 3 Jun 1999

In 22.2.8 [lib.facets.examples] paragraph 13, the do_truename() and do_falsename() functions in the example facet BoolNames should be const. The functions they are overriding in numpunct_byname<char> are const.

Proposed resolution:

In 22.2.8 [lib.facets.examples] paragraph 13, insert "const" in two places:

string do_truename() const { return "Oui Oui!"; }
string do_falsename() const { return "Mais Non!"; }


150. Find_first_of says integer instead of iterator

Section: 25.1.4 [lib.alg.find.first.of]  Status: TC  Submitter: Matt McClure  Date: 30 Jun 1999

Proposed resolution:

Change 25.1.4 [lib.alg.find.first.of] paragraph 2 from:

Returns: The first iterator i in the range [first1, last1) such that for some integer j in the range [first2, last2) ...

to:

Returns: The first iterator i in the range [first1, last1) such that for some iterator j in the range [first2, last2) ...


151. Can't currently clear() empty container

Section: 23.1.1 [lib.sequence.reqmts]  Status: TC  Submitter: Ed Brey  Date: 21 Jun 1999

For both sequences and associative containers, a.clear() has the semantics of erase(a.begin(),a.end()), which is undefined for an empty container since erase(q1,q2) requires that q1 be dereferenceable (23.1.1,3 and 23.1.2,7). When the container is empty, a.begin() is not dereferenceable.

The requirement that q1 be unconditionally dereferenceable causes many operations to be intuitively undefined, of which clearing an empty container is probably the most dire.

Since q1 and q2 are only referenced in the range [q1, q2), and [q1, q2) is required to be a valid range, stating that q1 and q2 must be iterators or certain kinds of iterators is unnecessary.

Proposed resolution:

In 23.1.1, paragraph 3, change:

p and q2 denote valid iterators to a, q and q1 denote valid dereferenceable iterators to a, [q1, q2) denotes a valid range

to:

p denotes a valid iterator to a, q denotes a valid dereferenceable iterator to a, [q1, q2) denotes a valid range in a

In 23.1.2, paragraph 7, change:

p and q2 are valid iterators to a, q and q1 are valid dereferenceable iterators to a, [q1, q2) is a valid range

to

p is a valid iterator to a, q is a valid dereferenceable iterator to a, [q1, q2) is a valid range into a


152. Typo in scan_is() semantics

Section: 22.2.1.1.2 [lib.locale.ctype.virtuals]  Status: TC  Submitter: Dietmar Kühl  Date: 20 Jul 1999

The semantics of scan_is() (paragraphs 4 and 6) is not exactly described because there is no function is() which only takes a character as argument. Also, in the effects clause (paragraph 3), the semantic is also kept vague.

Proposed resolution:

In 22.2.1.1.2 [lib.locale.ctype.virtuals] paragraphs 4 and 6, change the returns clause from:

"... such that is(*p) would..."

to:  "... such that is(m, *p) would...."


153. Typo in narrow() semantics

Section: 22.2.1.3.2 [lib.facet.ctype.char.members]  Status: WP  Submitter: Dietmar Kühl  Date: 20 Jul 1999

The description of the array version of narrow() (in paragraph 11) is flawed: There is no member do_narrow() which takes only three arguments because in addition to the range a default character is needed.

Additionally, for both widen and narrow we have two signatures followed by a Returns clause that only addresses one of them.

Proposed resolution:

Change the returns clause in 22.2.1.3.2 [lib.facet.ctype.char.members] paragraph 10 from:

    Returns: do_widen(low, high, to).

to:

    Returns: do_widen(c) or do_widen(low, high, to), respectively.

Change 22.2.1.3.2 [lib.facet.ctype.char.members] paragraph 10 and 11 from:

        char        narrow(char c, char /*dfault*/) const;
        const char* narrow(const char* low, const char* high,
                           char /*dfault*/, char* to) const;
        Returns: do_narrow(low, high, to).

to:

        char        narrow(char c, char dfault) const;
        const char* narrow(const char* low, const char* high,
                           char dfault, char* to) const;
        Returns: do_narrow(c, dfault) or
                 do_narrow(low, high, dfault, to), respectively.

[Kona: 1) the problem occurs in additional places, 2) a user defined version could be different.]

[Post-Tokyo: Dietmar provided the above wording at the request of the LWG. He could find no other places the problem occurred. He asks for clarification of the Kona "a user defined version..." comment above. Perhaps it was a circuitous way of saying "dfault" needed to be uncommented?]

[Post-Toronto: the issues list maintainer has merged in the proposed resolution from issue 207, which addresses the same paragraphs.]


154. Missing double specifier for do_get()

Section: 22.2.2.1.2 [lib.facet.num.get.virtuals]  Status: TC  Submitter: Dietmar Kühl  Date: 20 Jul 1999

The table in paragraph 7 for the length modifier does not list the length modifier l to be applied if the type is double. Thus, the standard asks the implementation to do undefined things when using scanf() (the missing length modifier for scanf() when scanning doubles is actually a problem I found quite often in production code, too).

Proposed resolution:

In 22.2.2.1.2 [lib.facet.num.get.virtuals], paragraph 7, add a row in the Length Modifier table to say that for double a length modifier l is to be used.

Rationale:

The standard makes an embarrassing beginner's mistake.


155. Typo in naming the class defining the class Init

Section: 27.3 [lib.iostream.objects]  Status: TC  Submitter: Dietmar Kühl  Date: 20 Jul 1999

There are conflicting statements about where the class Init is defined. According to 27.3 [lib.iostream.objects] paragraph 2 it is defined as basic_ios::Init, according to 27.4.2 [lib.ios.base] it is defined as ios_base::Init.

Proposed resolution:

Change 27.3 [lib.iostream.objects] paragraph 2 from "basic_ios::Init" to "ios_base::Init".

Rationale:

Although not strictly wrong, the standard was misleading enough to warrant the change.


156. Typo in imbue() description

Section: 27.4.2.3 [lib.ios.base.locales]  Status: TC  Submitter: Dietmar Kühl  Date: 20 Jul 1999

There is a small discrepancy between the declarations of imbue(): in 27.4.2 [lib.ios.base] the argument is passed as locale const& (correct), in 27.4.2.3 [lib.ios.base.locales] it is passed as locale const (wrong).

Proposed resolution:

In 27.4.2.3 [lib.ios.base.locales] change the imbue argument from "locale const" to "locale const&".


158. Underspecified semantics for setbuf()

Section: 27.5.2.4.2 [lib.streambuf.virt.buffer]  Status: TC  Submitter: Dietmar Kühl  Date: 20 Jul 1999

The default behavior of setbuf() is described only for the situation that gptr() != 0 && gptr() != egptr(): namely to do nothing. What has to be done in other situations  is not described although there is actually only one reasonable approach, namely to do nothing, too.

Since changing the buffer would almost certainly mess up most buffer management of derived classes unless these classes do it themselves, the default behavior of setbuf() should always be to do nothing.

Proposed resolution:

Change 27.5.2.4.2 [lib.streambuf.virt.buffer], paragraph 3, Default behavior, to: "Default behavior: Does nothing. Returns this."


159. Strange use of underflow()

Section: 27.5.2.4.3 [lib.streambuf.virt.get]  Status: TC  Submitter: Dietmar Kühl  Date: 20 Jul 1999

The description of the meaning of the result of showmanyc() seems to be rather strange: It uses calls to underflow(). Using underflow() is strange because this function only reads the current character but does not extract it, uflow() would extract the current character. This should be fixed to use sbumpc() instead.

Proposed resolution:

Change 27.5.2.4.3 [lib.streambuf.virt.get] paragraph 1, showmanyc()returns clause, by replacing the word "supplied" with the words "extracted from the stream".


160. Typo: Use of non-existing function exception()

Section: 27.6.1.1 [lib.istream]  Status: TC  Submitter: Dietmar Kühl  Date: 20 Jul 1999

The paragraph 4 refers to the function exception() which is not defined. Probably, the referred function is basic_ios<>::exceptions().

Proposed resolution:

In 27.6.1.1 [lib.istream], 27.6.1.3 [lib.istream.unformatted], paragraph 1, 27.6.2.1 [lib.ostream], paragraph 3, and 27.6.2.5.1 [lib.ostream.formatted.reqmts], paragraph 1, change "exception()" to "exceptions()".

[Note to Editor: "exceptions" with an "s" is the correct spelling.]


161. Typo: istream_iterator vs. istreambuf_iterator

Section: 27.6.1.2.2 [lib.istream.formatted.arithmetic]  Status: TC  Submitter: Dietmar Kühl  Date: 20 Jul 1999

The note in the second paragraph pretends that the first argument is an object of type istream_iterator. This is wrong: It is an object of type istreambuf_iterator.

Proposed resolution:

Change 27.6.1.2.2 [lib.istream.formatted.arithmetic] from:

The first argument provides an object of the istream_iterator class...

to

The first argument provides an object of the istreambuf_iterator class...


164. do_put() has apparently unused fill argument

Section: 22.2.5.3.2 [lib.locale.time.put.virtuals]  Status: TC  Submitter: Angelika Langer  Date: 23 Jul 1999

In 22.2.5.3.2 [lib.locale.time.put.virtuals] the do_put() function is specified as taking a fill character as an argument, but the description of the function does not say whether the character is used at all and, if so, in which way. The same holds for any format control parameters that are accessible through the ios_base& argument, such as the adjustment or the field width. Is strftime() supposed to use the fill character in any way? In any case, the specification of time_put.do_put() looks inconsistent to me.

Is the signature of do_put() wrong, or is the effects clause incomplete?

Proposed resolution:

Add the following note after 22.2.5.3.2 [lib.locale.time.put.virtuals] paragraph 2:

[Note: the fill argument may be used in the implementation-defined formats, or by derivations. A space character is a reasonable default for this argument. --end Note]

Rationale:

The LWG felt that while the normative text was correct, users need some guidance on what to pass for the fill argument since the standard doesn't say how it's used.


165. xsputn(), pubsync() never called by basic_ostream members?

Section: 27.6.2.1 [lib.ostream]  Status: WP  Submitter: Dietmar Kühl  Date: 20 Jul 1999

Paragraph 2 explicitly states that none of the basic_ostream functions falling into one of the groups "formatted output functions" and "unformatted output functions" calls any stream buffer function which might call a virtual function other than overflow(). Basically this is fine but this implies that sputn() (this function would call the virtual function xsputn()) is never called by any of the standard output functions. Is this really intended? At minimum it would be convenient to call xsputn() for strings... Also, the statement that overflow() is the only virtual member of basic_streambuf called is in conflict with the definition of flush() which calls rdbuf()->pubsync() and thereby the virtual function sync() (flush() is listed under "unformatted output functions").

In addition, I guess that the sentence starting with "They may use other public members of basic_ostream ..." probably was intended to start with "They may use other public members of basic_streamuf..." although the problem with the virtual members exists in both cases.

I see two obvious resolutions:

  1. state in a footnote that this means that xsputn() will never be called by any ostream member and that this is intended.
  2. relax the restriction and allow calling overflow() and xsputn(). Of course, the problem with flush() has to be resolved in some way.

Proposed resolution:

Change the last sentence of 27.6.2.1 (lib.ostream) paragraph 2 from:

They may use other public members of basic_ostream except that they do not invoke any virtual members of rdbuf() except overflow().

to:

They may use other public members of basic_ostream except that they shall not invoke any virtual members of rdbuf() except overflow(), xsputn(), and sync().

[Kona: the LWG believes this is a problem. Wish to ask Jerry or PJP why the standard is written this way.]

[Post-Tokyo: Dietmar supplied wording at the request of the LWG. He comments: The rules can be made a little bit more specific if necessary be explicitly spelling out what virtuals are allowed to be called from what functions and eg to state specifically that flush() is allowed to call sync() while other functions are not.]


167. Improper use of traits_type::length()

Section: 27.6.2.5.4 [lib.ostream.inserters.character]  Status: WP  Submitter: Dietmar Kühl  Date: 20 Jul 1999

Paragraph 4 states that the length is determined using traits::length(s). Unfortunately, this function is not defined for example if the character type is wchar_t and the type of s is char const*. Similar problems exist if the character type is char and the type of s is either signed char const* or unsigned char const*.

Proposed resolution:

Change 27.6.2.5.4 [lib.ostream.inserters.character] paragraph 4 from:

Effects: Behaves like an formatted inserter (as described in lib.ostream.formatted.reqmts) of out. After a sentry object is constructed it inserts characters. The number of characters starting at s to be inserted is traits::length(s). Padding is determined as described in lib.facet.num.put.virtuals. The traits::length(s) characters starting at s are widened using out.widen (lib.basic.ios.members). The widened characters and any required padding are inserted into out. Calls width(0).

to:

Effects: Behaves like a formatted inserter (as described in lib.ostream.formatted.reqmts) of out. After a sentry object is constructed it inserts n characters starting at s, where n is the number that would be computed as if by:

Padding is determined as described in lib.facet.num.put.virtuals. The n characters starting at s are widened using out.widen (lib.basic.ios.members). The widened characters and any required padding are inserted into out. Calls width(0).

[Santa Cruz: Matt supplied new wording]

[Kona: changed "where n is" to " where n is the number that would be computed as if by"]

Rationale:

We have five separate cases. In two of them we can use the user-supplied traits class without any fuss. In the other three we try to use something as close to that user-supplied class as possible. In two cases we've got a traits class that's appropriate for char and what we've got is a const signed char* or a const unsigned char*; that's close enough so we can just use a reinterpret cast, and continue to use the user-supplied traits class. Finally, there's one case where we just have to give up: where we've got a traits class for some arbitrary charT type, and we somehow have to deal with a const char*. There's nothing better to do but fall back to char_traits<char>


168. Typo: formatted vs. unformatted

Section: 27.6.2.6 [lib.ostream.unformatted]  Status: TC  Submitter: Dietmar Kühl  Date: 20 Jul 1999

The first paragraph begins with a descriptions what has to be done in formatted output functions. Probably this is a typo and the paragraph really want to describe unformatted output functions...

Proposed resolution:

In 27.6.2.6 [lib.ostream.unformatted] paragraph 1, the first and last sentences, change the word "formatted" to "unformatted":

"Each unformatted output function begins ..."
"... value specified for the unformatted output function."


169. Bad efficiency of overflow() mandated

Section: 27.7.1.3 [lib.stringbuf.virtuals]  Status: TC  Submitter: Dietmar Kühl  Date: 20 Jul 1999

Paragraph 8, Notes, of this section seems to mandate an extremely inefficient way of buffer handling for basic_stringbuf, especially in view of the restriction that basic_ostream member functions are not allowed to use xsputn() (see 27.6.2.1 [lib.ostream]): For each character to be inserted, a new buffer is to be created.

Of course, the resolution below requires some handling of simultaneous input and output since it is no longer possible to update egptr() whenever epptr() is changed. A possible solution is to handle this in underflow().

Proposed resolution:

In 27.7.1.3 [lib.stringbuf.virtuals] paragraph 8, Notes, insert the words "at least" as in the following:

To make a write position available, the function reallocates (or initially allocates) an array object with a sufficient number of elements to hold the current array object (if any), plus at least one additional write position.


170. Inconsistent definition of traits_type

Section: 27.7.4 [lib.stringstream]  Status: TC  Submitter: Dietmar Kühl  Date: 20 Jul 1999

The classes basic_stringstream (27.7.4 [lib.stringstream]), basic_istringstream (27.7.2 [lib.istringstream]), and basic_ostringstream (27.7.3 [lib.ostringstream]) are inconsistent in their definition of the type traits_type: For istringstream, this type is defined, for the other two it is not. This should be consistent.

Proposed resolution:

Proposed resolution:

To the declarations of basic_ostringstream (27.7.3 [lib.ostringstream]) and basic_stringstream (27.7.4 [lib.stringstream]) add:

typedef traits traits_type;

171. Strange seekpos() semantics due to joint position

Section: 27.8.1.4 [lib.filebuf.virtuals]  Status: WP  Submitter: Dietmar Kühl  Date: 20 Jul 1999

Overridden virtual functions, seekpos()

In 27.8.1.1 [lib.filebuf] paragraph 3, it is stated that a joint input and output position is maintained by basic_filebuf. Still, the description of seekpos() seems to talk about different file positions. In particular, it is unclear (at least to me) what is supposed to happen to the output buffer (if there is one) if only the input position is changed. The standard seems to mandate that the output buffer is kept and processed as if there was no positioning of the output position (by changing the input position). Of course, this can be exactly what you want if the flag ios_base::ate is set. However, I think, the standard should say something like this:

Plus the appropriate error handling, that is...

Proposed resolution:

Change the unnumbered paragraph in 27.8.1.4 (lib.filebuf.virtuals) before paragraph 14 from:

pos_type seekpos(pos_type sp, ios_base::openmode = ios_base::in | ios_base::out);

Alters the file position, if possible, to correspond to the position stored in sp (as described below).

- if (which&ios_base::in)!=0, set the file position to sp, then update the input sequence

- if (which&ios_base::out)!=0, then update the output sequence, write any unshift sequence, and set the file position to sp.

to:

pos_type seekpos(pos_type sp, ios_base::openmode = ios_base::in | ios_base::out);

Alters the file position, if possible, to correspond to the position stored in sp (as described below). Altering the file position performs as follows:

1. if (om & ios_base::out)!=0, then update the output sequence and write any unshift sequence;

2. set the file position to sp;

3. if (om & ios_base::in)!=0, then update the input sequence;

where om is the open mode passed to the last call to open(). The operation fails if is_open() returns false.

[Kona: Dietmar is working on a proposed resolution.]

[Post-Tokyo: Dietmar supplied the above wording.]


172. Inconsistent types for basic_istream::ignore()

Section: 27.6.1.3 [lib.istream.unformatted]  Status: TC  Submitter: Greg Comeau, Dietmar Kühl  Date: 23 Jul 1999

In 27.6.1.1 [lib.istream] the function ignore() gets an object of type streamsize as first argument. However, in 27.6.1.3 [lib.istream.unformatted] paragraph 23 the first argument is of type int.

As far as I can see this is not really a contradiction because everything is consistent if streamsize is typedef to be int. However, this is almost certainly not what was intended. The same thing happened to basic_filebuf::setbuf(), as described in issue 173.

Darin Adler also submitted this issue, commenting: Either 27.6.1.1 should be modified to show a first parameter of type int, or 27.6.1.3 should be modified to show a first parameter of type streamsize and use numeric_limits<streamsize>::max.

Proposed resolution:

In 27.6.1.3 [lib.istream.unformatted] paragraph 23 and 24, change both uses of int in the description of ignore() to streamsize.


173. Inconsistent types for basic_filebuf::setbuf()

Section: 27.8.1.4 [lib.filebuf.virtuals]  Status: TC  Submitter: Greg Comeau, Dietmar Kühl  Date: 23 Jul 1999

In 27.8.1.1 [lib.filebuf] the function setbuf() gets an object of type streamsize as second argument. However, in 27.8.1.4 [lib.filebuf.virtuals] paragraph 9 the second argument is of type int.

As far as I can see this is not really a contradiction because everything is consistent if streamsize is typedef to be int. However, this is almost certainly not what was intended. The same thing happened to basic_istream::ignore(), as described in issue 172.

Proposed resolution:

In 27.8.1.4 [lib.filebuf.virtuals] paragraph 9, change all uses of int in the description of setbuf() to streamsize.


174. Typo: OFF_T vs. POS_T

Section: D.6 [depr.ios.members]  Status: TC  Submitter: Dietmar Kühl  Date: 23 Jul 1999

According to paragraph 1 of this section, streampos is the type OFF_T, the same type as streamoff. However, in paragraph 6 the streampos gets the type POS_T

Proposed resolution:

Change D.6 [depr.ios.members] paragraph 1 from "typedef OFF_T streampos;" to "typedef POS_T streampos;"


175. Ambiguity for basic_streambuf::pubseekpos() and a few other functions.

Section: D.6 [depr.ios.members]  Status: TC  Submitter: Dietmar Kühl  Date: 23 Jul 1999

According to paragraph 8 of this section, the methods basic_streambuf::pubseekpos(), basic_ifstream::open(), and basic_ofstream::open "may" be overloaded by a version of this function taking the type ios_base::open_mode as last argument argument instead of ios_base::openmode (ios_base::open_mode is defined in this section to be an alias for one of the integral types). The clause specifies, that the last argument has a default argument in three cases. However, this generates an ambiguity with the overloaded version because now the arguments are absolutely identical if the last argument is not specified.

Proposed resolution:

In D.6 [depr.ios.members] paragraph 8, remove the default arguments for basic_streambuf::pubseekpos(), basic_ifstream::open(), and basic_ofstream::open().


176. exceptions() in ios_base...?

Section: D.6 [depr.ios.members]  Status: TC  Submitter: Dietmar Kühl  Date: 23 Jul 1999

The "overload" for the function exceptions() in paragraph 8 gives the impression that there is another function of this function defined in class ios_base. However, this is not the case. Thus, it is hard to tell how the semantics (paragraph 9) can be implemented: "Call the corresponding member function specified in clause 27 [lib.input.output]."

Proposed resolution:

In D.6 [depr.ios.members] paragraph 8, move the declaration of the function exceptions()into class basic_ios.


179. Comparison of const_iterators to iterators doesn't work

Section: 23.1 [lib.container.requirements]  Status: WP  Submitter: Judy Ward  Date: 2 Jul 1998

Currently the following will not compile on two well-known standard library implementations:

#include <set>
using namespace std;

void f(const set<int> &s)
{
  set<int>::iterator i;
  if (i==s.end()); // s.end() returns a const_iterator
}

The reason this doesn't compile is because operator== was implemented as a member function of the nested classes set:iterator and set::const_iterator, and there is no conversion from const_iterator to iterator. Surprisingly, (s.end() == i) does work, though, because of the conversion from iterator to const_iterator.

I don't see a requirement anywhere in the standard that this must work. Should there be one? If so, I think the requirement would need to be added to the tables in section 24.1.1. I'm not sure about the wording. If this requirement existed in the standard, I would think that implementors would have to make the comparison operators non-member functions.

This issues was also raised on comp.std.c++ by Darin Adler.  The example given was:

bool check_equal(std::deque<int>::iterator i,
std::deque<int>::const_iterator ci)
{
return i == ci;
}

Comment from John Potter:

In case nobody has noticed, accepting it will break reverse_iterator.

The fix is to make the comparison operators templated on two types.

    template <class Iterator1, class Iterator2>
    bool operator== (reverse_iterator<Iterator1> const& x,
                     reverse_iterator<Iterator2> const& y);
    

Obviously: return x.base() == y.base();

Currently, no reverse_iterator to const_reverse_iterator compares are valid.

BTW, I think the issue is in support of bad code. Compares should be between two iterators of the same type. All std::algorithms require the begin and end iterators to be of the same type.

Proposed resolution:

Insert this paragraph after 23.1 [lib.container.requirements] paragraph 7:

In the expressions

    i == j
    i != j
    i < j
    i <= j
    i >= j
    i > j
    i - j
  

Where i and j denote objects of a container's iterator type, either or both may be replaced by an object of the container's const_iterator type referring to the same element with no change in semantics.

[post-Toronto: Judy supplied a proposed resolution saying that iterator and const_iterator could be freely mixed in iterator comparison and difference operations.]

[Redmond: Dave and Howard supplied a new proposed resolution which explicitly listed expressions; there was concern that the previous proposed resolution was too informal.]

Rationale:

The LWG believes it is clear that the above wording applies only to the nested types X::iterator and X::const_iterator, where X is a container. There is no requirement that X::reverse_iterator and X::const_reverse_iterator can be mixed. If mixing them is considered important, that's a separate issue. (Issue 280.)


181. make_pair() unintended behavior

Section: 20.2.2 [lib.pairs]  Status: TC  Submitter: Andrew Koenig  Date: 3 Aug 1999

The claim has surfaced in Usenet that expressions such as

       make_pair("abc", 3)

are illegal, notwithstanding their use in examples, because template instantiation tries to bind the first template parameter to const char (&)[4], which type is uncopyable.

I doubt anyone intended that behavior...

Proposed resolution:

In 20.2 [lib.utility], paragraph 1 change the following declaration of make_pair():

template <class T1, class T2> pair<T1,T2> make_pair(const T1&, const T2&);

to:

template <class T1, class T2> pair<T1,T2> make_pair(T1, T2);

In 20.2.2 [lib.pairs] paragraph 7 and the line before, change:

template <class T1, class T2>
pair<T1, T2> make_pair(const T1& x, const T2& y);

to:

template <class T1, class T2>
pair<T1, T2> make_pair(T1 x, T2 y);

and add the following footnote to the effects clause:

According to 12.8 [class.copy], an implementation is permitted to not perform a copy of an argument, thus avoiding unnecessary copies.

Rationale:

Two potential fixes were suggested by Matt Austern and Dietmar Kühl, respectively, 1) overloading with array arguments, and 2) use of a reference_traits class with a specialization for arrays. Andy Koenig suggested changing to pass by value. In discussion, it appeared that this was a much smaller change to the standard that the other two suggestions, and any efficiency concerns were more than offset by the advantages of the solution. Two implementors reported that the proposed resolution passed their test suites.


182. Ambiguous references to size_t

Section: 17 [lib.library]  Status: WP  Submitter: Al Stevens  Date: 15 Aug 1999

Many references to size_t throughout the document omit the std:: namespace qualification.

For example, 17.4.3.4 [lib.replacement.functions] paragraph 2:

— operator new(size_t)
— operator new(size_t, const std::nothrow_t&)
— operator new[](size_t)
— operator new[](size_t, const std::nothrow_t&)

Proposed resolution:

In 17.4.3.4 [lib.replacement.functions] paragraph 2: replace:

- operator new(size_t)
- operator new(size_t, const std::nothrow_t&)
- operator new[](size_t)
- operator new[](size_t, const std::nothrow_t&)

by:

- operator new(std::size_t)
- operator new(std::size_t, const std::nothrow_t&)
- operator new[](std::size_t)
- operator new[](std::size_t, const std::nothrow_t&)

In [lib.allocator.requirements] 20.1.5, paragraph 4: replace:

The typedef members pointer, const_pointer, size_type, and difference_type are required to be T*, T const*, size_t, and ptrdiff_t, respectively.

 by:

The typedef members pointer, const_pointer, size_type, and difference_type are required to be T*, T const*, std::size_t, and std::ptrdiff_t, respectively.

In [lib.allocator.members] 20.4.1.1, paragraphs 3 and 6: replace:

3 Notes: Uses ::operator new(size_t) (18.4.1).

6 Note: the storage is obtained by calling ::operator new(size_t), but it is unspecified when or how often this function is called. The use of hint is unspecified, but intended as an aid to locality if an implementation so desires.

by:

3 Notes: Uses ::operator new(std::size_t) (18.4.1).

6 Note: the storage is obtained by calling ::operator new(std::size_t), but it is unspecified when or how often this function is called. The use of hint is unspecified, but intended as an aid to locality if an implementation so desires.

In [lib.char.traits.require] 21.1.1, paragraph 1: replace:

In Table 37, X denotes a Traits class defining types and functions for the character container type CharT; c and d denote values of type CharT; p and q denote values of type const CharT*; s denotes a value of type CharT*; n, i and j denote values of type size_t; e and f denote values of type X::int_type; pos denotes a value of type X::pos_type; and state denotes a value of type X::state_type.

by:

In Table 37, X denotes a Traits class defining types and functions for the character container type CharT; c and d denote values of type CharT; p and q denote values of type const CharT*; s denotes a value of type CharT*; n, i and j denote values of type std::size_t; e and f denote values of type X::int_type; pos denotes a value of type X::pos_type; and state denotes a value of type X::state_type.

In [lib.char.traits.require] 21.1.1, table 37: replace the return type of X::length(p): "size_t" by "std::size_t".

In [lib.std.iterator.tags] 24.3.3, paragraph 2: replace:
    typedef ptrdiff_t difference_type;
by:
    typedef std::ptrdiff_t difference_type;

In [lib.locale.ctype] 22.2.1.1 put namespace std { ...} around the declaration of template <class charT> class ctype.

In [lib.iterator.traits] 24.3.1, paragraph 2 put namespace std { ...} around the declaration of:

    template<class Iterator> struct iterator_traits
    template<class T> struct iterator_traits<T*>
    template<class T> struct iterator_traits<const T*>

Rationale:

The LWG believes correcting names like size_t and ptrdiff_t to std::size_t and std::ptrdiff_t to be essentially editorial. There there can't be another size_t or ptrdiff_t meant anyway because, according to 17.4.3.1.4 [lib.extern.types],

For each type T from the Standard C library, the types ::T and std::T are reserved to the implementation and, when defined, ::T shall be identical to std::T.

The issue is treated as a Defect Report to make explicit the Project Editor's authority to make this change.

[Post-Tokyo: Nico Josuttis provided the above wording at the request of the LWG.]

[Toronto: This is tangentially related to issue 229, but only tangentially: the intent of this issue is to address use of the name size_t in contexts outside of namespace std, such as in the description of ::operator new. The proposed changes should be reviewed to make sure they are correct.]

[pre-Copenhagen: Nico has reviewed the changes and believes them to be correct.]


183. I/O stream manipulators don't work for wide character streams

Section: 27.6.3 [lib.std.manip]  Status: WP  Submitter: Andy Sawyer  Date: 7 Jul 1999

27.6.3 [lib.std.manip] paragraph 3 says (clause numbering added for exposition):

Returns: An object s of unspecified type such that if [1] out is an (instance of) basic_ostream then the expression out<<s behaves as if f(s) were called, and if [2] in is an (instance of) basic_istream then the expression in>>s behaves as if f(s) were called. Where f can be defined as: ios_base& f(ios_base& str, ios_base::fmtflags mask) { // reset specified flags str.setf(ios_base::fmtflags(0), mask); return str; } [3] The expression out<<s has type ostream& and value out. [4] The expression in>>s has type istream& and value in.

Given the definitions [1] and [2] for out and in, surely [3] should read: "The expression out << s has type basic_ostream& ..." and [4] should read: "The expression in >> s has type basic_istream& ..."

If the wording in the standard is correct, I can see no way of implementing any of the manipulators so that they will work with wide character streams.

e.g. wcout << setbase( 16 );

Must have value 'wcout' (which makes sense) and type 'ostream&' (which doesn't).

The same "cut'n'paste" type also seems to occur in Paras 4,5,7 and 8. In addition, Para 6 [setfill] has a similar error, but relates only to ostreams.

I'd be happier if there was a better way of saying this, to make it clear that the value of the expression is "the same specialization of basic_ostream as out"&

Proposed resolution:

Replace section 27.6.3 [lib.std.manip] except paragraph 1 with the following:

2- The type designated smanip in each of the following function descriptions is implementation-specified and may be different for each function.

smanip resetiosflags(ios_base::fmtflags mask);

-3- Returns: An object s of unspecified type such that if out is an instance of basic_ostream<charT,traits> then the expression out<<s behaves as if f(s, mask) were called, or if in is an instance of basic_istream<charT,traits> then the expression in>>s behaves as if f(s, mask) were called. The function f can be defined as:*

[Footnote: The expression cin >> resetiosflags(ios_base::skipws) clears ios_base::skipws in the format flags stored in the basic_istream<charT,traits> object cin (the same as cin >> noskipws), and the expression cout << resetiosflags(ios_base::showbase) clears ios_base::showbase in the format flags stored in the basic_ostream<charT,traits> object cout (the same as cout << noshowbase). --- end footnote]

     ios_base& f(ios_base& str, ios_base::fmtflags mask)
   {
   // reset specified flags
   str.setf(ios_base::fmtflags(0), mask);
   return str;
   }

The expression out<<s has type basic_ostream<charT,traits>& and value out. The expression in>>s has type basic_istream<charT,traits>& and value in.

 smanip setiosflags(ios_base::fmtflags mask);

-4- Returns: An object s of unspecified type such that if out is an instance of basic_ostream<charT,traits> then the expression out<<s behaves as if f(s, mask) were called, or if in is an instance of basic_istream<charT,traits> then the expression in>>s behaves as if f(s, mask) were called. The function f can be defined as:

     ios_base& f(ios_base& str, ios_base::fmtflags mask)
   {
   // set specified flags
   str.setf(mask);
   return str;
   }

The expression out<<s has type basic_ostream<charT,traits>& and value out. The expression in>>s has type basic_istream<charT,traits>& and value in.

smanip setbase(int base);

-5- Returns: An object s of unspecified type such that if out is an instance of basic_ostream<charT,traits> then the expression out<<s behaves as if f(s, base) were called, or if in is an instance of basic_istream<charT,traits> then the expression in>>s behaves as if f(s, base) were called. The function f can be defined as:

     ios_base& f(ios_base& str, int base)
   {
   // set basefield
   str.setf(base == 8 ? ios_base::oct :
   base == 10 ? ios_base::dec :
   base == 16 ? ios_base::hex :
   ios_base::fmtflags(0), ios_base::basefield);
   return str;
   }

The expression out<<s has type basic_ostream<charT,traits>& and value out. The expression in>>s has type basic_istream<charT,traits>& and value in.

smanip setfill(char_type c);

-6- Returns: An object s of unspecified type such that if out is (or is derived from) basic_ostream<charT,traits> and c has type charT then the expression out<<s behaves as if f(s, c) were called. The function f can be defined as:

      template<class charT, class traits>
   basic_ios<charT,traits>& f(basic_ios<charT,traits>& str, charT c)
   {
   // set fill character
   str.fill(c);
   return str;
   }

The expression out<<s has type basic_ostream<charT,traits>& and value out.

smanip setprecision(int n);

-7- Returns: An object s of unspecified type such that if out is an instance of basic_ostream<charT,traits> then the expression out<<s behaves as if f(s, n) were called, or if in is an instance of basic_istream<charT,traits> then the expression in>>s behaves as if f(s, n) were called. The function f can be defined as:

      ios_base& f(ios_base& str, int n)
   {
   // set precision
   str.precision(n);
   return str;
   }

The expression out<<s has type basic_ostream<charT,traits>& and value out. The expression in>>s has type basic_istream<charT,traits>& and value in
.
smanip setw(int n);

-8- Returns: An object s of unspecified type such that if out is an instance of basic_ostream<charT,traits> then the expression out<<s behaves as if f(s, n) were called, or if in is an instance of basic_istream<charT,traits> then the expression in>>s behaves as if f(s, n) were called. The function f can be defined as:

      ios_base& f(ios_base& str, int n)
   {
   // set width
   str.width(n);
   return str;
   }

The expression out<<s has type basic_ostream<charT,traits>& and value out. The expression in>>s has type basic_istream<charT,traits>& and value in.

[Kona: Andy Sawyer and Beman Dawes will work to improve the wording of the proposed resolution.]

[Tokyo - The LWG noted that issue 216 involves the same paragraphs.]

[Post-Tokyo: The issues list maintainer combined the proposed resolution of this issue with the proposed resolution for issue 216 as they both involved the same paragraphs, and were so intertwined that dealing with them separately appear fraught with error. The full text was supplied by Bill Plauger; it was cross checked against changes supplied by Andy Sawyer. It should be further checked by the LWG.]


184. numeric_limits<bool> wording problems

Section: 18.2.1.5 [lib.numeric.special]  Status: WP  Submitter: Gabriel Dos Reis  Date: 21 Jul 1999

bools are defined by the standard to be of integer types, as per 3.9.1 [basic.fundamental] paragraph 7. However "integer types" seems to have a special meaning for the author of 18.2. The net effect is an unclear and confusing specification for numeric_limits<bool> as evidenced below.

18.2.1.2/7 says numeric_limits<>::digits is, for built-in integer types, the number of non-sign bits in the representation.

4.5/4 states that a bool promotes to int ; whereas 4.12/1 says any non zero arithmetical value converts to true.

I don't think it makes sense at all to require numeric_limits<bool>::digits and numeric_limits<bool>::digits10 to be meaningful.

The standard defines what constitutes a signed (resp. unsigned) integer types. It doesn't categorize bool as being signed or unsigned. And the set of values of bool type has only two elements.

I don't think it makes sense to require numeric_limits<bool>::is_signed to be meaningful.

18.2.1.2/18 for numeric_limits<integer_type>::radix  says:

For integer types, specifies the base of the representation.186)

This disposition is at best misleading and confusing for the standard requires a "pure binary numeration system" for integer types as per 3.9.1/7

The footnote 186) says: "Distinguishes types with base other than 2 (e.g BCD)."  This also erroneous as the standard never defines any integer types with base representation other than 2.

Furthermore, numeric_limits<bool>::is_modulo and numeric_limits<bool>::is_signed have similar problems.

Proposed resolution:

Append to the end of 18.2.1.5 [lib.numeric.special]:

The specialization for bool shall be provided as follows:

    namespace std {
       template<> class numeric_limits<bool> {
       public:
         static const bool is_specialized = true;
         static bool min() throw() { return false; }
         static bool max() throw() { return true; }

         static const int  digits = 1;
         static const int  digits10 = 0;
         static const bool is_signed = false;
         static const bool is_integer = true;
         static const bool is_exact = true;
         static const int  radix = 2;
         static bool epsilon() throw() { return 0; }
         static bool round_error() throw() { return 0; }

         static const int  min_exponent = 0;
         static const int  min_exponent10 = 0;
         static const int  max_exponent = 0;
         static const int  max_exponent10 = 0;

         static const bool has_infinity = false;
         static const bool has_quiet_NaN = false;
         static const bool has_signaling_NaN = false;
         static const float_denorm_style has_denorm = denorm_absent;
         static const bool has_denorm_loss = false;
         static bool infinity() throw() { return 0; }
         static bool quiet_NaN() throw() { return 0; }
         static bool signaling_NaN() throw() { return 0; }
         static bool denorm_min() throw() { return 0; }

         static const bool is_iec559 = false;
         static const bool is_bounded = true;
         static const bool is_modulo = false;

         static const bool traps = false;
         static const bool tinyness_before = false;
         static const float_round_style round_style = round_toward_zero;
       };
     }

[Tokyo:  The LWG desires wording that specifies exact values rather than more general wording in the original proposed resolution.]

[Post-Tokyo:  At the request of the LWG in Tokyo, Nico Josuttis provided the above wording.]


185. Questionable use of term "inline"

Section: 20.3 [lib.function.objects]  Status: WP  Submitter: UK Panel  Date: 26 Jul 1999

Paragraph 4 of 20.3 [lib.function.objects] says:

 [Example: To negate every element of a: transform(a.begin(), a.end(), a.begin(), negate<double>()); The corresponding functions will inline the addition and the negation. end example]

(Note: The "addition" referred to in the above is in para 3) we can find no other wording, except this (non-normative) example which suggests that any "inlining" will take place in this case.

Indeed both:

17.4.4.3 Global Functions [lib.global.functions] 1 It is unspecified whether any global functions in the C++ Standard Library are defined as inline (7.1.2).

and

17.4.4.4 Member Functions [lib.member.functions] 1 It is unspecified whether any member functions in the C++ Standard Library are defined as inline (7.1.2).

take care to state that this may indeed NOT be the case.

Thus the example "mandates" behavior that is explicitly not required elsewhere.

Proposed resolution:

In 20.3 [lib.function.objects] paragraph 1, remove the sentence:

They are important for the effective use of the library.

Remove 20.3 [lib.function.objects] paragraph 2, which reads:

Using function objects together with function templates increases the expressive power of the library as well as making the resulting code much more efficient.

In 20.3 [lib.function.objects] paragraph 4, remove the sentence:

The corresponding functions will inline the addition and the negation.

[Kona: The LWG agreed there was a defect.]

[Tokyo: The LWG crafted the proposed resolution.]


186. bitset::set() second parameter should be bool

Section: 23.3.5.2 [lib.bitset.members]  Status: WP  Submitter: Darin Adler  Date: 13 Aug 1999

In section 23.3.5.2 [lib.bitset.members], paragraph 13 defines the bitset::set operation to take a second parameter of type int. The function tests whether this value is non-zero to determine whether to set the bit to true or false. The type of this second parameter should be bool. For one thing, the intent is to specify a Boolean value. For another, the result type from test() is bool. In addition, it's possible to slice an integer that's larger than an int. This can't happen with bool, since conversion to bool has the semantic of translating 0 to false and any non-zero value to true.

Proposed resolution:

In 23.3.5 [lib.template.bitset] Para 1 Replace:

bitset<N>& set(size_t pos, int val = true ); 

With:

bitset<N>& set(size_t pos, bool val = true );

In 23.3.5.2 [lib.bitset.members] Para 12(.5) Replace:

bitset<N>& set(size_t pos, int val = 1 );

With:

bitset<N>& set(size_t pos, bool val = true );

[Kona: The LWG agrees with the description.  Andy Sawyer will work on better P/R wording.]

[Post-Tokyo: Andy provided the above wording.]

Rationale:

bool is a better choice. It is believed that binary compatibility is not an issue, because this member function is usually implemented as inline, and because it is already the case that users cannot rely on the type of a pointer to a nonvirtual member of a standard library class.


187. iter_swap underspecified

Section: 25.2.2 [lib.alg.swap]  Status: WP  Submitter: Andrew Koenig  Date: 14 Aug 1999

The description of iter_swap in 25.2.2 paragraph 7,says that it ``exchanges the values'' of the objects to which two iterators refer.

What it doesn't say is whether it does so using swap or using the assignment operator and copy constructor.

This question is an important one to answer, because swap is specialized to work efficiently for standard containers.
For example:

vector<int> v1, v2;
iter_swap(&v1, &v2);

Is this call to iter_swap equivalent to calling swap(v1, v2)?  Or is it equivalent to

{
vector<int> temp = v1;
v1 = v2;
v2 = temp;
}

The first alternative is O(1); the second is O(n).

A LWG member, Dave Abrahams, comments:

Not an objection necessarily, but I want to point out the cost of that requirement:

iter_swap(list<T>::iterator, list<T>::iterator)

can currently be specialized to be more efficient than iter_swap(T*,T*) for many T (by using splicing). Your proposal would make that optimization illegal. 

[Kona: The LWG notes the original need for iter_swap was proxy iterators which are no longer permitted.]

Proposed resolution:

Change the effect clause of iter_swap in 25.2.2 paragraph 7 from:

Exchanges the values pointed to by the two iterators a and b.

to

swap(*a, *b).

Rationale:

It's useful to say just what iter_swap does. There may be some iterators for which we want to specialize iter_swap, but the fully general version should have a general specification.

Note that in the specific case of list<T>::iterator, iter_swap should not be specialized as suggested above. That would do much more than exchanging the two iterators' values: it would change predecessor/successor relationships, possibly moving the iterator from one list to another. That would surely be inappropriate.


189. setprecision() not specified correctly

Section: 27.4.2.2 [lib.fmtflags.state]  Status: TC  Submitter: Andrew Koenig  Date: 25 Aug 1999

27.4.2.2 paragraph 9 claims that setprecision() sets the precision, and includes a parenthetical note saying that it is the number of digits after the decimal point.

This claim is not strictly correct. For example, in the default floating-point output format, setprecision sets the number of significant digits printed, not the number of digits after the decimal point.

I would like the committee to look at the definition carefully and correct the statement in 27.4.2.2

Proposed resolution:

Remove from 27.4.2.2 [lib.fmtflags.state], paragraph 9, the text "(number of digits after the decimal point)".


193. Heap operations description incorrect

Section: 25.3.6 [lib.alg.heap.operations]  Status: TC  Submitter: Markus Mauhart  Date: 24 Sep 1999

25.3.6 [lib.alg.heap.operations] states two key properties of a heap [a,b), the first of them is

    `"(1) *a is the largest element"

I think this is incorrect and should be changed to the wording in the proposed resolution.

Actually there are two independent changes:

A-"part of largest equivalence class" instead of "largest", cause 25.3 [lib.alg.sorting] asserts "strict weak ordering" for all its sub clauses.

B-Take 'an oldest' from that equivalence class, otherwise the heap functions could not be used for a priority queue as explained in 23.2.3.2.2 [lib.priqueue.members] (where I assume that a "priority queue" respects priority AND time).

Proposed resolution:

Change 25.3.6 [lib.alg.heap.operations] property (1) from:

(1) *a is the largest element

to:

(1) There is no element greater than *a


195. Should basic_istream::sentry's constructor ever set eofbit?

Section: 27.6.1.1.2 [lib.istream::sentry]  Status: TC  Submitter: Matt Austern  Date: 13 Oct 1999

Suppose that is.flags() & ios_base::skipws is nonzero. What should basic_istream<>::sentry's constructor do if it reaches eof while skipping whitespace? 27.6.1.1.2/5 suggests it should set failbit. Should it set eofbit as well? The standard doesn't seem to answer that question.

On the one hand, nothing in 27.6.1.1.2 [lib.istream::sentry] says that basic_istream<>::sentry should ever set eofbit. On the other hand, 27.6.1.1 [lib.istream] paragraph 4 says that if extraction from a streambuf "returns traits::eof(), then the input function, except as explicitly noted otherwise, completes its actions and does setstate(eofbit)". So the question comes down to whether basic_istream<>::sentry's constructor is an input function.

Comments from Jerry Schwarz:

It was always my intention that eofbit should be set any time that a virtual returned something to indicate eof, no matter what reason iostream code had for calling the virtual.

The motivation for this is that I did not want to require streambufs to behave consistently if their virtuals are called after they have signaled eof.

The classic case is a streambuf reading from a UNIX file. EOF isn't really a state for UNIX file descriptors. The convention is that a read on UNIX returns 0 bytes to indicate "EOF", but the file descriptor isn't shut down in any way and future reads do not necessarily also return 0 bytes. In particular, you can read from tty's on UNIX even after they have signaled "EOF". (It isn't always understood that a ^D on UNIX is not an EOF indicator, but an EOL indicator. By typing a "line" consisting solely of ^D you cause a read to return 0 bytes, and by convention this is interpreted as end of file.)

Proposed resolution:

Add a sentence to the end of 27.6.1.1.2 paragraph 2:

If is.rdbuf()->sbumpc() or is.rdbuf()->sgetc() returns traits::eof(), the function calls setstate(failbit | eofbit) (which may throw ios_base::failure).


198. Validity of pointers and references unspecified after iterator destruction

Section: 24.1 [lib.iterator.requirements]  Status: WP  Submitter: Beman Dawes  Date: 3 Nov 1999

Is a pointer or reference obtained from an iterator still valid after destruction of the iterator?

Is a pointer or reference obtained from an iterator still valid after the value of the iterator changes?

#include <iostream>
#include <vector>
#include <iterator>

int main()
{
    typedef std::vector<int> vec_t;
    vec_t v;
    v.push_back( 1 );

    // Is a pointer or reference obtained from an iterator still
    // valid after destruction of the iterator?
    int * p = &*v.begin();
    std::cout << *p << '\n';  // OK?

    // Is a pointer or reference obtained from an iterator still
    // valid after the value of the iterator changes?
    vec_t::iterator iter( v.begin() );
    p = &*iter++;
    std::cout << *p << '\n';  // OK?

    return 0;
}

The standard doesn't appear to directly address these questions. The standard needs to be clarified. At least two real-world cases have been reported where library implementors wasted considerable effort because of the lack of clarity in the standard. The question is important because requiring pointers and references to remain valid has the effect for practical purposes of prohibiting iterators from pointing to cached rather than actual elements of containers.

The standard itself assumes that pointers and references obtained from an iterator are still valid after iterator destruction or change. The definition of reverse_iterator::operator*(), 24.4.1.3.3 [lib.reverse.iter.op.star], which returns a reference, defines effects:

Iterator tmp = current;
return *--tmp;

The definition of reverse_iterator::operator->(), 24.4.1.3.4 [lib.reverse.iter.opref], which returns a pointer, defines effects:

return &(operator*());

Because the standard itself assumes pointers and references remain valid after iterator destruction or change, the standard should say so explicitly. This will also reduce the chance of user code breaking unexpectedly when porting to a different standard library implementation.

Proposed resolution:

Add a new paragraph to 24.1 [lib.iterator.requirements]:

Destruction of an iterator may invalidate pointers and references previously obtained from that iterator.

Replace paragraph 1 of 24.4.1.3.3 [lib.reverse.iter.op.star] with:

Effects:

  this->tmp = current;
  --this->tmp;
  return *this->tmp;

[Note: This operation must use an auxiliary member variable, rather than a temporary variable, to avoid returning a reference that persists beyond the lifetime of its associated iterator. (See 24.1 [lib.iterator.requirements].) The name of this member variable is shown for exposition only. --end note]

[Post-Tokyo: The issue has been reformulated purely in terms of iterators.]

[Pre-Toronto: Steve Cleary pointed out the no-invalidation assumption by reverse_iterator. The issue and proposed resolution was reformulated yet again to reflect this reality.]

[Copenhagen: Steve Cleary pointed out that reverse_iterator assumes its underlying iterator has persistent pointers and references. Andy Koenig pointed out that it is possible to rewrite reverse_iterator so that it no longer makes such an assupmption. However, this issue is related to issue 299. If we decide it is intentional that p[n] may return by value instead of reference when p is a Random Access Iterator, other changes in reverse_iterator will be necessary.]

Rationale:

This issue has been discussed extensively. Note that it is not an issue about the behavior of predefined iterators. It is asking whether or not user-defined iterators are permitted to have transient pointers and references. Several people presented examples of useful user-defined iterators that have such a property; examples include a B-tree iterator, and an "iota iterator" that doesn't point to memory. Library implementors already seem to be able to cope with such iterators: they take pains to avoid forming references to memory that gets iterated past. The only place where this is a problem is reverse_iterator, so this issue changes reverse_iterator to make it work.

This resolution does not weaken any guarantees provided by predefined iterators like list<int>::iterator. Clause 23 should be reviewed to make sure that guarantees for predefined iterators are as strong as users expect.


199. What does allocate(0) return?

Section: 20.1.5 [lib.allocator.requirements]  Status: TC  Submitter: Matt Austern  Date: 19 Nov 1999

Suppose that A is a class that conforms to the Allocator requirements of Table 32, and a is an object of class A What should be the return value of a.allocate(0)? Three reasonable possibilities: forbid the argument 0, return a null pointer, or require that the return value be a unique non-null pointer.

Proposed resolution:

Add a note to the allocate row of Table 32: "[Note: If n == 0, the return value is unspecified. --end note]"

Rationale:

A key to understanding this issue is that the ultimate use of allocate() is to construct an iterator, and that iterator for zero length sequences must be the container's past-the-end representation. Since this already implies special case code, it would be over-specification to mandate the return value.


200. Forward iterator requirements don't allow constant iterators

Section: 24.1.3 [lib.forward.iterators]  Status: WP  Submitter: Matt Austern  Date: 19 Nov 1999

In table 74, the return type of the expression *a is given as T&, where T is the iterator's value type. For constant iterators, however, this is wrong. ("Value type" is never defined very precisely, but it is clear that the value type of, say, std::list<int>::const_iterator is supposed to be int, not const int.)

Proposed resolution:

In table 74, in the *a and *r++ rows, change the return type from "T&" to "T& if X is mutable, otherwise const T&". In the a->m row, change the return type from "U&" to "U& if X is mutable, otherwise const U&".

[Tokyo: The LWG believes this is the tip of a larger iceberg; there are multiple const problems with the STL portion of the library and that these should be addressed as a single package.  Note that issue 180 has already been declared NAD Future for that very reason.]

[Redmond: the LWG thinks this is separable from other constness issues. This issue is just cleanup; it clarifies language that was written before we had iterator_traits. Proposed resolution was modified: the original version only discussed *a. It was pointed out that we also need to worry about *r++ and a->m.]


202. unique() effects unclear when predicate not an equivalence relation

Section: 25.2.8 [lib.alg.unique]  Status: WP  Submitter: Andrew Koenig  Date: 13 Jan 2000

What should unique() do if you give it a predicate that is not an equivalence relation? There are at least two plausible answers:

1. You can't, because 25.2.8 says that it it "eliminates all but the first element from every consecutive group of equal elements..." and it wouldn't make sense to interpret "equal" as meaning anything but an equivalence relation. [It also doesn't make sense to interpret "equal" as meaning ==, because then there would never be any sense in giving a predicate as an argument at all.]

2. The word "equal" should be interpreted to mean whatever the predicate says, even if it is not an equivalence relation (and in particular, even if it is not transitive).

The example that raised this question is from Usenet:

int f[] = { 1, 3, 7, 1, 2 };
int* z = unique(f, f+5, greater<int>());

If one blindly applies the definition using the predicate greater<int>, and ignore the word "equal", you get:

Eliminates all but the first element from every consecutive group of elements referred to by the iterator i in the range [first, last) for which *i > *(i - 1).

The first surprise is the order of the comparison. If we wanted to allow for the predicate not being an equivalence relation, then we should surely compare elements the other way: pred(*(i - 1), *i). If we do that, then the description would seem to say: "Break the sequence into subsequences whose elements are in strictly increasing order, and keep only the first element of each subsequence". So the result would be 1, 1, 2. If we take the description at its word, it would seem to call for strictly DEcreasing order, in which case the result should be 1, 3, 7, 2.

In fact, the SGI implementation of unique() does neither: It yields 1, 3, 7.

Proposed resolution:

Change 25.2.8 [lib.alg.unique] paragraph 1 to:

For a nonempty range, eliminates all but the first element from every consecutive group of equivalent elements referred to by the iterator i in the range [first+1, last) for which the following conditions hold: *(i-1) == *i or pred(*(i-1), *i) != false.

Also insert a new paragraph, paragraph 2a, that reads: "Requires: The comparison function must be an equivalence relation."

[Redmond: discussed arguments for and against requiring the comparison function to be an equivalence relation. Straw poll: 14-2-5. First number is to require that it be an equivalence relation, second number is to explicitly not require that it be an equivalence relation, third number is people who believe they need more time to consider the issue. A separate issue: Andy Sawyer pointed out that "i-1" is incorrect, since "i" can refer to the first iterator in the range. Matt provided wording to address this problem.]

[Curaçao: The LWG changed "... the range (first, last)..." to "... the range [first+1, last)..." for clarity. They considered this change close enough to editorial to not require another round of review.]

Rationale:

The LWG also considered an alternative resolution: change 25.2.8 [lib.alg.unique] paragraph 1 to:

For a nonempty range, eliminates all but the first element from every consecutive group of elements referred to by the iterator i in the range (first, last) for which the following conditions hold: *(i-1) == *i or pred(*(i-1), *i) != false.

Also insert a new paragraph, paragraph 1a, that reads: "Notes: The comparison function need not be an equivalence relation."

Informally: the proposed resolution imposes an explicit requirement that the comparison function be an equivalence relation. The alternative resolution does not, and it gives enough information so that the behavior of unique() for a non-equivalence relation is specified. Both resolutions are consistent with the behavior of existing implementations.


208. Unnecessary restriction on past-the-end iterators

Section: 24.1 [lib.iterator.requirements]  Status: TC  Submitter: Stephen Cleary  Date: 02 Feb 2000

In 24.1 paragraph 5, it is stated ". . . Dereferenceable and past-the-end values are always non-singular."

This places an unnecessary restriction on past-the-end iterators for containers with forward iterators (for example, a singly-linked list). If the past-the-end value on such a container was a well-known singular value, it would still satisfy all forward iterator requirements.

Removing this restriction would allow, for example, a singly-linked list without a "footer" node.

This would have an impact on existing code that expects past-the-end iterators obtained from different (generic) containers being not equal.

Proposed resolution:

Change 24.1 [lib.iterator.requirements] paragraph 5, the last sentence, from:

Dereferenceable and past-the-end values are always non-singular.

to:

Dereferenceable values are always non-singular. 

Rationale:

For some kinds of containers, including singly linked lists and zero-length vectors, null pointers are perfectly reasonable past-the-end iterators. Null pointers are singular.


209. basic_string declarations inconsistent

Section: 21.3 [lib.basic.string]  Status: TC  Submitter: Igor Stauder  Date: 11 Feb 2000

In Section 21.3 [lib.basic.string] the basic_string member function declarations use a consistent style except for the following functions:

void push_back(const charT);
basic_string& assign(const basic_string&);
void swap(basic_string<charT,traits,Allocator>&);

- push_back, assign, swap: missing argument name 
- push_back: use of const with charT (i.e. POD type passed by value not by reference - should be charT or const charT& )
- swap: redundant use of template parameters in argument basic_string<charT,traits,Allocator>&

Proposed resolution:

In Section 21.3 [lib.basic.string] change the basic_string member function declarations push_back, assign, and swap to:

void push_back(charT c); 

basic_string& assign(const basic_string& str);
void swap(basic_string& str);

Rationale:

Although the standard is in general not consistent in declaration style, the basic_string declarations are consistent other than the above. The LWG felt that this was sufficient reason to merit the change.


210. distance first and last confused

Section: 25 [lib.algorithms]  Status: TC  Submitter: Lisa Lippincott  Date: 15 Feb 2000

In paragraph 9 of section 25 [lib.algorithms], it is written:

In the description of the algorithms operators + and - are used for some of the iterator categories for which they do not have to be defined. In these cases the semantics of [...] a-b is the same as of

     return distance(a, b);

Proposed resolution:

On the last line of paragraph 9 of section 25 [lib.algorithms] change "a-b" to "b-a".

Rationale:

There are two ways to fix the defect; change the description to b-a or change the return to distance(b,a). The LWG preferred the former for consistency.


211. operator>>(istream&, string&) doesn't set failbit

Section: 21.3.7.9 [lib.string.io]  Status: TC  Submitter: Scott Snyder  Date: 4 Feb 2000

The description of the stream extraction operator for std::string (section 21.3.7.9 [lib.string.io]) does not contain a requirement that failbit be set in the case that the operator fails to extract any characters from the input stream.

This implies that the typical construction

std::istream is;
std::string str;
...
while (is >> str) ... ;

(which tests failbit) is not required to terminate at EOF.

Furthermore, this is inconsistent with other extraction operators, which do include this requirement. (See sections 27.6.1.2 [lib.istream.formatted] and 27.6.1.3 [lib.istream.unformatted]), where this requirement is present, either explicitly or implicitly, for the extraction operators. It is also present explicitly in the description of getline (istream&, string&, charT) in section 21.3.7.9 [lib.string.io] paragraph 8.)

Proposed resolution:

Insert new paragraph after paragraph 2 in section 21.3.7.9 [lib.string.io]:

If the function extracts no characters, it calls is.setstate(ios::failbit) which may throw ios_base::failure (27.4.4.3).


212. Empty range behavior unclear for several algorithms

Section: 25.3.7 [lib.alg.min.max]  Status: TC  Submitter: Nico Josuttis  Date: 26 Feb 2000

The standard doesn't specify what min_element() and max_element() shall return if the range is empty (first equals last). The usual implementations return last. This problem seems also apply to partition(), stable_partition(), next_permutation(), and prev_permutation().

Proposed resolution:

In 25.3.7 [lib.alg.min.max] - Minimum and maximum, paragraphs 7 and 9, append: Returns last if first==last.

Rationale:

The LWG looked in some detail at all of the above mentioned algorithms, but believes that except for min_element() and max_element() it is already clear that last is returned if first == last.


214. set::find() missing const overload

Section: 23.3.3 [lib.set], 23.3.4 [lib.multiset]  Status: WP  Submitter: Judy Ward  Date: 28 Feb 2000

The specification for the associative container requirements in Table 69 state that the find member function should "return iterator; const_iterator for constant a". The map and multimap container descriptions have two overloaded versions of find, but set and multiset do not, all they have is:

iterator find(const key_type & x) const;

Proposed resolution:

Change the prototypes for find(), lower_bound(), upper_bound(), and equal_range() in section 23.3.3 [lib.set] and section 23.3.4 [lib.multiset] to each have two overloads:

iterator find(const key_type & x);
const_iterator find(const key_type & x) const;
iterator lower_bound(const key_type & x);
const_iterator lower_bound(const key_type & x) const;
iterator upper_bound(const key_type & x);
const_iterator upper_bound(const key_type & x) const;
pair<iterator, iterator> equal_range(const key_type & x);
pair<const_iterator, const_iterator> equal_range(const key_type & x) const;

[Tokyo: At the request of the LWG, Judy Ward provided wording extending the proposed resolution to lower_bound, upper_bound, and equal_range.]


217. Facets example (Classifying Japanese characters) contains errors

Section: 22.2.8 [lib.facets.examples]  Status: TC  Submitter: Martin Sebor  Date: 29 Feb 2000

The example in 22.2.8, paragraph 11 contains the following errors:

1) The member function `My::JCtype::is_kanji()' is non-const; the function must be const in order for it to be callable on a const object (a reference to which which is what std::use_facet<>() returns).

2) In file filt.C, the definition of `JCtype::id' must be qualified with the name of the namespace `My'.

3) In the definition of `loc' and subsequently in the call to use_facet<>() in main(), the name of the facet is misspelled: it should read `My::JCtype' rather than `My::JCType'.

Proposed resolution:

Replace the "Classifying Japanese characters" example in 22.2.8, paragraph 11 with the following:

#include <locale>
namespace My {
    using namespace std;
    class JCtype : public locale::facet {
    public:
        static locale::id id;     //  required for use as a new locale facet
        bool is_kanji (wchar_t c) const;
        JCtype() {}
    protected:
        ~JCtype() {}
    };
}
//  file:  filt.C
#include <iostream>
#include <locale>
#include "jctype"                 //  above
std::locale::id My::JCtype::id;   //  the static  JCtype  member
declared above.
int main()
{
    using namespace std;
    typedef ctype<wchar_t> wctype;
    locale loc(locale(""),              //  the user's preferred locale...
               new My::JCtype);         //  and a new feature ...
    wchar_t c = use_facet<wctype>(loc).widen('!');
    if (!use_facet<My::JCtype>(loc).is_kanji(c))
        cout << "no it isn't!" << endl;
    return 0;
}

220. ~ios_base() usage valid?

Section: 27.4.2.7 [lib.ios.base.cons]  Status: TC  Submitter: Jonathan Schilling, Howard Hinnant  Date: 13 Mar 2000

The pre-conditions for the ios_base destructor are described in 27.4.2.7 paragraph 2:

Effects: Destroys an object of class ios_base. Calls each registered callback pair (fn,index) (27.4.2.6) as (*fn)(erase_event,*this,index) at such time that any ios_base member function called from within fn has well defined results.

But what is not clear is: If no callback functions were ever registered, does it matter whether the ios_base members were ever initialized?

For instance, does this program have defined behavior:

#include <ios>
class D : public std::ios_base { };
int main() { D d; }

It seems that registration of a callback function would surely affect the state of an ios_base. That is, when you register a callback function with an ios_base, the ios_base must record that fact somehow.

But if after construction the ios_base is in an indeterminate state, and that state is not made determinate before the destructor is called, then how would the destructor know if any callbacks had indeed been registered? And if the number of callbacks that had been registered is indeterminate, then is not the behavior of the destructor undefined?

By comparison, the basic_ios class description in 27.4.4.1 paragraph 2 makes it explicit that destruction before initialization results in undefined behavior.

Proposed resolution:

Modify 27.4.2.7 paragraph 1 from

Effects: Each ios_base member has an indeterminate value after construction.

to

Effects: Each ios_base member has an indeterminate value after construction. These members must be initialized by calling basic_ios::init. If an ios_base object is destroyed before these initializations have taken place, the behavior is undefined.


221. num_get<>::do_get stage 2 processing broken

Section: 22.2.2.1.2 [lib.facet.num.get.virtuals]  Status: WP  Submitter: Matt Austern  Date: 14 Mar 2000

Stage 2 processing of numeric conversion is broken.

Table 55 in 22.2.2.1.2 says that when basefield is 0 the integral conversion specifier is %i. A %i specifier determines a number's base by its prefix (0 for octal, 0x for hex), so the intention is clearly that a 0x prefix is allowed. Paragraph 8 in the same section, however, describes very precisely how characters are processed. (It must be done "as if" by a specified code fragment.) That description does not allow a 0x prefix to be recognized.

Very roughly, stage 2 processing reads a char_type ct. It converts ct to a char, not by using narrow but by looking it up in a translation table that was created by widening the string literal "0123456789abcdefABCDEF+-". The character "x" is not found in that table, so it can't be recognized by stage 2 processing.

Proposed resolution:

In 22.2.2.1.2 paragraph 8, replace the line:

static const char src[] = "0123456789abcdefABCDEF+-";

with the line:

static const char src[] = "0123456789abcdefxABCDEFX+-";

Rationale:

If we're using the technique of widening a string literal, the string literal must contain every character we wish to recognize. This technique has the consequence that alternate representations of digits will not be recognized. This design decision was made deliberately, with full knowledge of that limitation.


222. Are throw clauses necessary if a throw is already implied by the effects clause?

Section: 17.3.1.3 [lib.structure.specifications]  Status: TC  Submitter: Judy Ward  Date: 17 Mar 2000

Section 21.3.6.8 describes the basic_string::compare function this way:

21.3.6.8 - basic_string::compare [lib.string::compare]

int compare(size_type pos1, size_type n1,
                const basic_string<charT,traits,Allocator>&  str ,
                size_type  pos2 , size_type  n2 ) const;

-4- Returns: 

    basic_string<charT,traits,Allocator>(*this,pos1,n1).compare(
                 basic_string<charT,traits,Allocator>(str,pos2,n2)) .

and the constructor that's implicitly called by the above is defined to throw an out-of-range exception if pos > str.size(). See section 21.3.1 [lib.string.cons] paragraph 4.

On the other hand, the compare function descriptions themselves don't have "Throws: " clauses and according to 17.3.1.3, paragraph 3, elements that do not apply to a function are omitted.

So it seems there is an inconsistency in the standard -- are the "Effects" clauses correct, or are the "Throws" clauses missing?

Proposed resolution:

In 17.3.1.3 [lib.structure.specifications] paragraph 3, the footnote 148 attached to the sentence "Descriptions of function semantics contain the following elements (as appropriate):", insert the word "further" so that the foot note reads:

To save space, items that do not apply to a function are omitted. For example, if a function does not specify any further preconditions, there will be no "Requires" paragraph.

Rationale:

The standard is somewhat inconsistent, but a failure to note a throw condition in a throws clause does not grant permission not to throw. The inconsistent wording is in a footnote, and thus non-normative. The proposed resolution from the LWG clarifies the footnote.


223. reverse algorithm should use iter_swap rather than swap

Section: 25.2.9 [lib.alg.reverse]  Status: TC  Submitter: Dave Abrahams  Date: 21 Mar 2000

Shouldn't the effects say "applies iter_swap to all pairs..."?

Proposed resolution:

In 25.2.9 [lib.alg.reverse], replace:

Effects: For each non-negative integer i <= (last - first)/2, applies swap to all pairs of iterators first + i, (last - i) - 1.

with:

Effects: For each non-negative integer i <= (last - first)/2, applies iter_swap to all pairs of iterators first + i, (last - i) - 1.

224. clear() complexity for associative containers refers to undefined N

Section: 23.1.2 [lib.associative.reqmts]  Status: TC  Submitter: Ed Brey  Date: 23 Mar 2000

In the associative container requirements table in 23.1.2 paragraph 7, a.clear() has complexity "log(size()) + N". However, the meaning of N is not defined.

Proposed resolution:

In the associative container requirements table in 23.1.2 paragraph 7, the complexity of a.clear(), change "log(size()) + N" to "linear in size()".

Rationale:

It's the "log(size())", not the "N", that is in error: there's no difference between O(N) and O(N + log(N)). The text in the standard is probably an incorrect cut-and-paste from the range version of erase.


225. std:: algorithms use of other unqualified algorithms

Section: 17.4.4.3 [lib.global.functions]  Status: WP  Submitter: Dave Abrahams  Date: 01 Apr 2000

Are algorithms in std:: allowed to use other algorithms without qualification, so functions in user namespaces might be found through Koenig lookup?

For example, a popular standard library implementation includes this implementation of std::unique:

namespace std {
    template <class _ForwardIter>
    _ForwardIter unique(_ForwardIter __first, _ForwardIter __last) {
      __first = adjacent_find(__first, __last);
      return unique_copy(__first, __last, __first);
    }
    }

Imagine two users on opposite sides of town, each using unique on his own sequences bounded by my_iterators . User1 looks at his standard library implementation and says, "I know how to implement a more efficient unique_copy for my_iterators", and writes:

namespace user1 {
    class my_iterator;
    // faster version for my_iterator
    my_iterator unique_copy(my_iterator, my_iterator, my_iterator);
    }

user1::unique_copy() is selected by Koenig lookup, as he intended.

User2 has other needs, and writes:

namespace user2 {
    class my_iterator;
    // Returns true iff *c is a unique copy of *a and *b.
    bool unique_copy(my_iterator a, my_iterator b, my_iterator c);
    }

User2 is shocked to find later that his fully-qualified use of std::unique(user2::my_iterator, user2::my_iterator, user2::my_iterator) fails to compile (if he's lucky). Looking in the standard, he sees the following Effects clause for unique():

Effects: Eliminates all but the first element from every consecutive group of equal elements referred to by the iterator i in the range [first, last) for which the following corresponding conditions hold: *i == *(i - 1) or pred(*i, *(i - 1)) != false

The standard gives user2 absolutely no reason to think he can interfere with std::unique by defining names in namespace user2. His standard library has been built with the template export feature, so he is unable to inspect the implementation. User1 eventually compiles his code with another compiler, and his version of unique_copy silently stops being called. Eventually, he realizes that he was depending on an implementation detail of his library and had no right to expect his unique_copy() to be called portably.

On the face of it, and given above scenario, it may seem obvious that the implementation of unique() shown is non-conforming because it uses unique_copy() rather than ::std::unique_copy(). Most standard library implementations, however, seem to disagree with this notion.

[Tokyo:  Steve Adamczyk from the core working group indicates that "std::" is sufficient;  leading "::" qualification is not required because any namespace qualification is sufficient to suppress Koenig lookup.]

Proposed resolution:

Add a paragraph and a note at the end of 17.4.4.3 [lib.global.functions]:

Unless otherwise specified, no global or non-member function in the standard library shall use a function from another namespace which is found through argument-dependent name lookup (3.4.2 [basic.lookup.koenig]).

[Note: the phrase "unless otherwise specified" is intended to allow Koenig lookup in cases like that of ostream_iterators:

Effects:

*out_stream << value;
if(delim != 0) *out_stream << delim;
return (*this);

--end note]

[Tokyo: The LWG agrees that this is a defect in the standard, but is as yet unsure if the proposed resolution is the best solution. Furthermore, the LWG believes that the same problem of unqualified library names applies to wording in the standard itself, and has opened issue 229 accordingly. Any resolution of issue 225 should be coordinated with the resolution of issue 229.]

[Toronto: The LWG is not sure if this is a defect in the standard. Most LWG members believe that an implementation of std::unique like the one quoted in this issue is already illegal, since, under certain circumstances, its semantics are not those specified in the standard. The standard's description of unique does not say that overloading adjacent_find should have any effect.]

[Curaçao: An LWG-subgroup spent an afternoon working on issues 225, 226, and 229. Their conclusion was that the issues should be separated into an LWG portion (Howard's paper, N1387=02-0045), and a EWG portion (Dave will write a proposal). The LWG and EWG had (separate) discussions of this plan the next day. The proposed resolution for this issue is in accordance with Howard's paper.]

Rationale:

It could be argued that this proposed isn't strictly necessary, that the Standard doesn't grant implementors license to write a standard function that behaves differently than specified in the Standard just because of an unrelated user-defined name in some other namespace. However, this is at worst a clarification. It is surely right that algorithsm shouldn't pick up random names, that user-defined names should have no effect unless otherwise specified. Issue 226 deals with the question of when it is appropriate for the standard to explicitly specify otherwise.


226. User supplied specializations or overloads of namespace std function templates

Section: 17.4.3.1 [lib.reserved.names]  Status: WP  Submitter: Dave Abrahams  Date: 01 Apr 2000

The issues are: 

1. How can a 3rd party library implementor (lib1) write a version of a standard algorithm which is specialized to work with his own class template? 

2. How can another library implementor (lib2) write a generic algorithm which will take advantage of the specialized algorithm in lib1?

This appears to be the only viable answer under current language rules:

namespace lib1
{
    // arbitrary-precision numbers using T as a basic unit
    template <class T>
    class big_num { //...
    };
    
    // defining this in namespace std is illegal (it would be an
    // overload), so we hope users will rely on Koenig lookup
    template <class T>
    void swap(big_int<T>&, big_int<T>&);
}
#include <algorithm>
namespace lib2
{
    template <class T>
    void generic_sort(T* start, T* end)
    {
            ...
        // using-declaration required so we can work on built-in types
        using std::swap;
        // use Koenig lookup to find specialized algorithm if available
        swap(*x, *y);
    }
}

This answer has some drawbacks. First of all, it makes writing lib2 difficult and somewhat slippery. The implementor needs to remember to write the using-declaration, or generic_sort will fail to compile when T is a built-in type. The second drawback is that the use of this style in lib2 effectively "reserves" names in any namespace which defines types which may eventually be used with lib2. This may seem innocuous at first when applied to names like swap, but consider more ambiguous names like unique_copy() instead. It is easy to imagine the user wanting to define these names differently in his own namespace. A definition with semantics incompatible with the standard library could cause serious problems (see issue 225).

Why, you may ask, can't we just partially specialize std::swap()? It's because the language doesn't allow for partial specialization of function templates. If you write:

namespace std
{
    template <class T>
    void swap(lib1::big_int<T>&, lib1::big_int<T>&);
}

You have just overloaded std::swap, which is illegal under the current language rules. On the other hand, the following full specialization is legal:

namespace std
{
    template <>
    void swap(lib1::other_type&, lib1::other_type&);
}

This issue reflects concerns raised by the "Namespace issue with specialized swap" thread on comp.lang.c++.moderated. A similar set of concerns was earlier raised on the boost.org mailing list and the ACCU-general mailing list. Also see library reflector message c++std-lib-7354.

J. C. van Winkel points out (in c++std-lib-9565) another unexpected fact: it's impossible to output a container of std::pair's using copy and an ostream_iterator, as long as both pair-members are built-in or std:: types. That's because a user-defined operator<< for (for example) std::pair<const std::string, int> will not be found: lookup for operator<< will be performed only in namespace std. Opinions differed on whether or not this was a defect, and, if so, whether the defect is that something is wrong with user-defined functionality and std, or whether it's that the standard library does not provide an operator<< for std::pair<>.

Proposed resolution:

Adopt the wording proposed in Howard Hinnant's paper N1523=03-0106, "Proposed Resolution To LWG issues 225, 226, 229".

[Tokyo: Summary, "There is no conforming way to extend std::swap for user defined templates."  The LWG agrees that there is a problem. Would like more information before proceeding. This may be a core issue. Core issue 229 has been opened to discuss the core aspects of this problem. It was also noted that submissions regarding this issue have been received from several sources, but too late to be integrated into the issues list. ]

[Post-Tokyo: A paper with several proposed resolutions, J16/00-0029==WG21/N1252, "Shades of namespace std functions " by Alan Griffiths, is in the Post-Tokyo mailing. It should be considered a part of this issue.]

[Toronto: Dave Abrahams and Peter Dimov have proposed a resolution that involves core changes: it would add partial specialization of function template. The Core Working Group is reluctant to add partial specialization of function templates. It is viewed as a large change, CWG believes that proposal presented leaves some syntactic issues unanswered; if the CWG does add partial specialization of function templates, it wishes to develop its own proposal. The LWG continues to believe that there is a serious problem: there is no good way for users to force the library to use user specializations of generic standard library functions, and in certain cases (e.g. transcendental functions called by valarray and complex) this is important. Koenig lookup isn't adequate, since names within the library must be qualified with std (see issue 225), specialization doesn't work (we don't have partial specialization of function templates), and users aren't permitted to add overloads within namespace std. ]

[Copenhagen: Discussed at length, with no consensus. Relevant papers in the pre-Copenhagen mailing: N1289, N1295, N1296. Discussion focused on four options. (1) Relax restrictions on overloads within namespace std. (2) Mandate that the standard library use unqualified calls for swap and possibly other functions. (3) Introduce helper class templates for swap and possibly other functions. (4) Introduce partial specialization of function templates. Every option had both support and opposition. Straw poll (first number is support, second is strongly opposed): (1) 6, 4; (2) 6, 7; (3) 3, 8; (4) 4, 4.]

[Redmond: Discussed, again no consensus. Herb presented an argument that a user who is defining a type T with an associated swap should not be expected to put that swap in namespace std, either by overloading or by partial specialization. The argument is that swap is part of T's interface, and thus should to in the same namespace as T and only in that namespace. If we accept this argument, the consequence is that standard library functions should use unqualified call of swap. (And which other functions? Any?) A small group (Nathan, Howard, Jeremy, Dave, Matt, Walter, Marc) will try to put together a proposal before the next meeting.]

[Curaçao: An LWG-subgroup spent an afternoon working on issues 225, 226, and 229. Their conclusion was that the issues should be separated into an LWG portion (Howard's paper, N1387=02-0045), and a EWG portion (Dave will write a proposal). The LWG and EWG had (separate) discussions of this plan the next day. The proposed resolution is the one proposed by Howard.]

[Santa Cruz: the LWG agreed with the general direction of Howard's paper, N1387. (Roughly: Koenig lookup is disabled unless we say otherwise; this issue is about when we do say otherwise.) However, there were concerns about wording. Howard will provide new wording. Bill and Jeremy will review it.]

[Kona: Howard proposed the new wording. The LWG accepted his proposed resolution.]

Rationale:

Informally: introduce a Swappable concept, and specify that the value types of the iterators passed to certain standard algorithms (such as iter_swap, swap_ranges, reverse, rotate, and sort) conform to that concept. The Swappable concept will make it clear that these algorithms use unqualified lookup for the calls to swap. Also, in 26.3.3.3 [lib.valarray.transcend] paragraph 1, state that the valarray transcendentals use unqualified lookup.


227. std::swap() should require CopyConstructible or DefaultConstructible arguments

Section: 25.2.2 [lib.alg.swap]  Status: TC  Submitter: Dave Abrahams  Date: 09 Apr 2000

25.2.2 reads:

template<class T> void swap(T& a, T& b);

Requires: Type T is Assignable (_lib.container.requirements_).
Effects: Exchanges values stored in two locations.

The only reasonable** generic implementation of swap requires construction of a new temporary copy of one of its arguments:

template<class T> void swap(T& a, T& b);
  {
      T tmp(a);
      a = b;
      b = tmp;
  }

But a type which is only Assignable cannot be swapped by this implementation.

**Yes, there's also an unreasonable implementation which would require T to be DefaultConstructible instead of CopyConstructible. I don't think this is worthy of consideration:

template<class T> void swap(T& a, T& b);
{
    T tmp;
    tmp = a;
    a = b;
    b = tmp;
}

Proposed resolution:

Change 25.2.2 paragraph 1 from:

Requires: Type T is Assignable (23.1).

to:

Requires: Type T is CopyConstructible (20.1.3) and Assignable (23.1)


228. Incorrect specification of "..._byname" facets

Section: 22.2 [lib.locale.categories]  Status: WP  Submitter: Dietmar Kühl  Date: 20 Apr 2000

The sections 22.2.1.2 [lib.locale.ctype.byname], 22.2.1.4 [lib.locale.ctype.byname.special], 22.2.1.6 [lib.locale.codecvt.byname], 22.2.3.2 [lib.locale.numpunct.byname], 22.2.4.2 [lib.locale.collate.byname], 22.2.5.4 [lib.locale.time.put.byname], 22.2.6.4 [lib.locale.moneypunct.byname], and 22.2.7.2 [lib.locale.messages.byname] overspecify the definitions of the "..._byname" classes by listing a bunch of virtual functions. At the same time, no semantics of these functions are defined. Real implementations do not define these functions because the functional part of the facets is actually implemented in the corresponding base classes and the constructor of the "..._byname" version just provides suitable date used by these implementations. For example, the 'numpunct' methods just return values from a struct. The base class uses a statically initialized struct while the derived version reads the contents of this struct from a table. However, no virtual function is defined in 'numpunct_byname'.

For most classes this does not impose a problem but specifically for 'ctype' it does: The specialization for 'ctype_byname<char>' is required because otherwise the semantics would change due to the virtual functions defined in the general version for 'ctype_byname': In 'ctype<char>' the method 'do_is()' is not virtual but it is made virtual in both 'ctype<cT>' and 'ctype_byname<cT>'. Thus, a class derived from 'ctype_byname<char>' can tell whether this class is specialized or not under the current specification: Without the specialization, 'do_is()' is virtual while with specialization it is not virtual.

Proposed resolution:

  Change section 22.2.1.2 (lib.locale.ctype.byname) to become:

     namespace std {
       template <class charT>
       class ctype_byname : public ctype<charT> {
       public:
         typedef ctype<charT>::mask mask;
         explicit ctype_byname(const char*, size_t refs = 0);
       protected:
        ~ctype_byname();             //  virtual
       };
     }

  Change section 22.2.1.6 (lib.locale.codecvt.byname) to become:

    namespace std {
      template <class internT, class externT, class stateT>
      class codecvt_byname : public codecvt<internT, externT, stateT> {
      public:
       explicit codecvt_byname(const char*, size_t refs = 0);
      protected:
      ~codecvt_byname();             //  virtual
       };
     }

  Change section 22.2.3.2 (lib.locale.numpunct.byname) to become:

     namespace std {
       template <class charT>
       class numpunct_byname : public numpunct<charT> {
     //  this class is specialized for  char  and  wchar_t.
       public:
         typedef charT                char_type;
         typedef basic_string<charT>  string_type;
         explicit numpunct_byname(const char*, size_t refs = 0);
       protected:
        ~numpunct_byname();          //  virtual
       };
     }

  Change section 22.2.4.2 (lib.locale.collate.byname) to become:

     namespace std {
       template <class charT>
       class collate_byname : public collate<charT> {
       public:
         typedef basic_string<charT> string_type;
         explicit collate_byname(const char*, size_t refs = 0);
       protected:
        ~collate_byname();           //  virtual
       };
     }

  Change section 22.2.5.2 (lib.locale.time.get.byname) to become:

     namespace std {
       template <class charT, class InputIterator = istreambuf_iterator<charT> >
       class time_get_byname : public time_get<charT, InputIterator> {
       public:
         typedef time_base::dateorder dateorder;
         typedef InputIterator        iter_type
         explicit time_get_byname(const char*, size_t refs = 0);
       protected:
        ~time_get_byname();          //  virtual
       };
     }

  Change section 22.2.5.4 (lib.locale.time.put.byname) to become:

     namespace std {
       template <class charT, class OutputIterator = ostreambuf_iterator<charT> >
       class time_put_byname : public time_put<charT, OutputIterator>
       {
       public:
         typedef charT          char_type;
         typedef OutputIterator iter_type;
         explicit time_put_byname(const char*, size_t refs = 0);
       protected:
        ~time_put_byname();          //  virtual
       };
     }"

  Change section 22.2.6.4 (lib.locale.moneypunct.byname) to become:

     namespace std {
       template <class charT, bool Intl = false>
       class moneypunct_byname : public moneypunct<charT, Intl> {
       public:
         typedef money_base::pattern pattern;
         typedef basic_string<charT> string_type;
         explicit moneypunct_byname(const char*, size_t refs = 0);
       protected:
        ~moneypunct_byname();        //  virtual
       };
     }

  Change section 22.2.7.2 (lib.locale.messages.byname) to become:

     namespace std {
       template <class charT>
       class messages_byname : public messages<charT> {
       public:
         typedef messages_base::catalog catalog;
         typedef basic_string<charT>    string_type;
         explicit messages_byname(const char*, size_t refs = 0);
       protected:
        ~messages_byname();          //  virtual
       };
     }

Remove section 22.2.1.4 [lib.locale.ctype.byname.special] completely (because in this case only those members are defined to be virtual which are defined to be virtual in 'ctype<cT>'.)

[Post-Tokyo: Dietmar Kühl submitted this issue at the request of the LWG to solve the underlying problems raised by issue 138.]

[Copenhagen: proposed resolution was revised slightly, to remove three last virtual functions from messages_byname.]


229. Unqualified references of other library entities

Section: 17.4.1.1 [lib.contents]  Status: WP  Submitter: Steve Clamage  Date: 19 Apr 2000

Throughout the library chapters, the descriptions of library entities refer to other library entities without necessarily qualifying the names.

For example, section 25.2.2 "Swap" describes the effect of swap_ranges in terms of the unqualified name "swap". This section could reasonably be interpreted to mean that the library must be implemented so as to do a lookup of the unqualified name "swap", allowing users to override any ::std::swap function when Koenig lookup applies.

Although it would have been best to use explicit qualification with "::std::" throughout, too many lines in the standard would have to be adjusted to make that change in a Technical Corrigendum.

Issue 182, which addresses qualification of size_t, is a special case of this.

Proposed resolution:

To section 17.4.1.1 "Library contents" Add the following paragraph:

Whenever a name x defined in the standard library is mentioned, the name x is assumed to be fully qualified as ::std::x, unless explicitly described otherwise. For example, if the Effects section for library function F is described as calling library function G, the function ::std::G is meant.

[Post-Tokyo: Steve Clamage submitted this issue at the request of the LWG to solve a problem in the standard itself similar to the problem within implementations of library identified by issue 225. Any resolution of issue 225 should be coordinated with the resolution of this issue.]

[post-Toronto: Howard is undecided about whether it is appropriate for all standard library function names referred to in other standard library functions to be explicitly qualified by std: it is common advice that users should define global functions that operate on their class in the same namespace as the class, and this requires argument-dependent lookup if those functions are intended to be called by library code. Several LWG members are concerned that valarray appears to require argument-dependent lookup, but that the wording may not be clear enough to fall under "unless explicitly described otherwise".]

[Curaçao: An LWG-subgroup spent an afternoon working on issues 225, 226, and 229. Their conclusion was that the issues should be separated into an LWG portion (Howard's paper, N1387=02-0045), and a EWG portion (Dave will write a proposal). The LWG and EWG had (separate) discussions of this plan the next day. This paper resolves issues 225 and 226. In light of that resolution, the proposed resolution for the current issue makes sense.]


230. Assignable specified without also specifying CopyConstructible

Section: 17 [lib.library]  Status: WP  Submitter: Beman Dawes  Date: 26 Apr 2000

Issue 227 identified an instance (std::swap) where Assignable was specified without also specifying CopyConstructible. The LWG asked that the standard be searched to determine if the same defect existed elsewhere.

There are a number of places (see proposed resolution below) where Assignable is specified without also specifying CopyConstructible. There are also several cases where both are specified. For example, 26.4.1 [lib.accumulate].

Proposed resolution:

In 23.1 [lib.container.requirements] table 65 for value_type: change "T is Assignable" to "T is CopyConstructible and Assignable"

In 23.1.2 [lib.associative.reqmts] table 69 X::key_type; change "Key is Assignable" to "Key is CopyConstructible and Assignable"

In 24.1.2 [lib.output.iterators] paragraph 1, change:

A class or a built-in type X satisfies the requirements of an output iterator if X is an Assignable type (23.1) and also the following expressions are valid, as shown in Table 73:

to:

A class or a built-in type X satisfies the requirements of an output iterator if X is a CopyConstructible (20.1.3) and Assignable type (23.1) and also the following expressions are valid, as shown in Table 73:

[Post-Tokyo: Beman Dawes submitted this issue at the request of the LWG. He asks that the 25.2.4 [lib.alg.replace] and 25.2.5 [lib.alg.fill] changes be studied carefully, as it is not clear that CopyConstructible is really a requirement and may be overspecification.]

[Portions of the resolution for issue 230 have been superceded by the resolution of issue 276.]

Rationale:

The original proposed resolution also included changes to input iterator, fill, and replace. The LWG believes that those changes are not necessary. The LWG considered some blanket statement, where an Assignable type was also required to be Copy Constructible, but decided against this because fill and replace really don't require the Copy Constructible property.


231. Precision in iostream?

Section: 22.2.2.2.2 [lib.facet.num.put.virtuals]  Status: WP  Submitter: James Kanze, Stephen Clamage  Date:  25 Apr 2000

What is the following program supposed to output?

#include <iostream>

    int
    main()
    {
        std::cout.setf( std::ios::scientific , std::ios::floatfield ) ;
        std::cout.precision( 0 ) ;
        std::cout << 1.00 << '\n' ;
        return 0 ;
    }

From my C experience, I would expect "1e+00"; this is what printf("%.0e" , 1.00 ); does. G++ outputs "1.000000e+00".

The only indication I can find in the standard is 22.2.2.2.2/11, where it says "For conversion from a floating-point type, if (flags & fixed) != 0 or if str.precision() > 0, then str.precision() is specified in the conversion specification." This is an obvious error, however, fixed is not a mask for a field, but a value that a multi-bit field may take -- the results of and'ing fmtflags with ios::fixed are not defined, at least not if ios::scientific has been set. G++'s behavior corresponds to what might happen if you do use (flags & fixed) != 0 with a typical implementation (floatfield == 3 << something, fixed == 1 << something, and scientific == 2 << something).

Presumably, the intent is either (flags & floatfield) != 0, or (flags & floatfield) == fixed; the first gives something more or less like the effect of precision in a printf floating point conversion. Only more or less, of course. In order to implement printf formatting correctly, you must know whether the precision was explicitly set or not. Say by initializing it to -1, instead of 6, and stating that for floating point conversions, if precision < -1, 6 will be used, for fixed point, if precision < -1, 1 will be used, etc. Plus, of course, if precision == 0 and flags & floatfield == 0, 1 should be = used. But it probably isn't necessary to emulate all of the anomalies of printf:-).

Proposed resolution:

Replace 22.2.2.2.2 [lib.facet.num.put.virtuals], paragraph 11, with the following sentence:

For conversion from a floating-point type, str.precision() is specified in the conversion specification.

Rationale:

The floatfield determines whether numbers are formatted as if with %f, %e, or %g. If the fixed bit is set, it's %f, if scientific it's %e, and if both bits are set, or neither, it's %g.

Turning to the C standard, a precision of 0 is meaningful for %f and %e. For %g, precision 0 is taken to be the same as precision 1.

The proposed resolution has the effect that if neither fixed nor scientific is set we'll be specifying a precision of 0, which will be internally turned into 1. There's no need to call it out as a special case.

The output of the above program will be "1e+00".

[Post-Curaçao: Howard provided improved wording covering the case where precision is 0 and mode is %g.]


232. "depends" poorly defined in 17.4.3.1

Section: 17.4.3.1 [lib.reserved.names]  Status: WP  Submitter: Peter Dimov  Date: 18 Apr 2000

17.4.3.1/1 uses the term "depends" to limit the set of allowed specializations of standard templates to those that "depend on a user-defined name of external linkage."

This term, however, is not adequately defined, making it possible to construct a specialization that is, I believe, technically legal according to 17.4.3.1/1, but that specializes a standard template for a built-in type such as 'int'.

The following code demonstrates the problem:

#include <algorithm>
template<class T> struct X
{
 typedef T type;
};
namespace std
{
 template<> void swap(::X<int>::type& i, ::X<int>::type& j);
}

Proposed resolution:

Change "user-defined name" to "user-defined type".

Rationale:

This terminology is used in section 2.5.2 and 4.1.1 of The C++ Programming Language. It disallows the example in the issue, since the underlying type itself is not user-defined. The only possible problem I can see is for non-type templates, but there's no possible way for a user to come up with a specialization for bitset, for example, that might not have already been specialized by the implementor?

[Toronto: this may be related to issue 120.]

[post-Toronto: Judy provided the above proposed resolution and rationale.]


234. Typos in allocator definition

Section: 20.4.1.1 [lib.allocator.members]  Status: WP  Submitter: Dietmar Kühl  Date: 24 Apr 2000

In paragraphs 12 and 13 the effects of construct() and destruct() are described as returns but the functions actually return void.

Proposed resolution:

Substitute "Returns" by "Effect".


235. No specification of default ctor for reverse_iterator

Section: 24.4.1.1 [lib.reverse.iterator]  Status: WP  Submitter: Dietmar Kühl  Date: 24 Apr 2000

The declaration of reverse_iterator lists a default constructor. However, no specification is given what this constructor should do.

Proposed resolution:

In section 24.4.1.3.1 [lib.reverse.iter.cons] add the following paragraph:

reverse_iterator()

Default initializes current. Iterator operations applied to the resulting iterator have defined behavior if and only if the corresponding operations are defined on a default constructed iterator of type Iterator.

[pre-Copenhagen: Dietmar provide wording for proposed resolution.]


237. Undefined expression in complexity specification

Section: 23.2.2.1 [lib.list.cons]  Status: WP  Submitter: Dietmar Kühl  Date: 24 Apr 2000

The complexity specification in paragraph 6 says that the complexity is linear in first - last. Even if operator-() is defined on iterators this term is in general undefined because it would have to be last - first.

Proposed resolution:

Change paragraph 6 from

Linear in first - last.

to become

Linear in distance(first, last).

238. Contradictory results of stringbuf initialization.

Section: 27.7.1.1 [lib.stringbuf.cons]  Status: WP  Submitter: Dietmar Kühl  Date: 11 May 2000

In 27.7.1.1 paragraph 4 the results of calling the constructor of 'basic_stringbuf' are said to be str() == str. This is fine that far but consider this code:

  std::basic_stringbuf<char> sbuf("hello, world", std::ios_base::openmode(0));
  std::cout << "'" << sbuf.str() << "'\n";

Paragraph 3 of 27.7.1.1 basically says that in this case neither the output sequence nor the input sequence is initialized and paragraph 2 of 27.7.1.2 basically says that str() either returns the input or the output sequence. None of them is initialized, ie. both are empty, in which case the return from str() is defined to be basic_string<cT>().

However, probably only test cases in some testsuites will detect this "problem"...

Proposed resolution:

Remove 27.7.1.1 paragraph 4.

Rationale:

We could fix 27.7.1.1 paragraph 4, but there would be no point. If we fixed it, it would say just the same thing as text that's already in the standard.


239. Complexity of unique() and/or unique_copy incorrect

Section: 25.2.8 [lib.alg.unique]  Status: WP  Submitter: Angelika Langer  Date: May 15 2000

The complexity of unique and unique_copy are inconsistent with each other and inconsistent with the implementations.  The standard specifies:

for unique():

-3- Complexity: If the range (last - first) is not empty, exactly (last - first) - 1 applications of the corresponding predicate, otherwise no applications of the predicate.

for unique_copy():

-7- Complexity: Exactly last - first applications of the corresponding predicate.

The implementations do it the other way round: unique() applies the predicate last-first times and unique_copy() applies it last-first-1 times.

As both algorithms use the predicate for pair-wise comparison of sequence elements I don't see a justification for unique_copy() applying the predicate last-first times, especially since it is not specified to which pair in the sequence the predicate is applied twice.

Proposed resolution:

Change both complexity sections in 25.2.8 [lib.alg.unique] to:

Complexity: For nonempty ranges, exactly last - first - 1 applications of the corresponding predicate.

240. Complexity of adjacent_find() is meaningless

Section: 25.1.5 [lib.alg.adjacent.find]  Status: WP  Submitter: Angelika Langer  Date: May 15 2000

The complexity section of adjacent_find is defective:

template <class ForwardIterator>
ForwardIterator adjacent_find(ForwardIterator first, ForwardIterator last
                              BinaryPredicate pred);

-1- Returns: The first iterator i such that both i and i + 1 are in the range [first, last) for which the following corresponding conditions hold: *i == *(i + 1), pred(*i, *(i + 1)) != false. Returns last if no such iterator is found.

-2- Complexity: Exactly find(first, last, value) - first applications of the corresponding predicate.

In the Complexity section, it is not defined what "value" is supposed to mean. My best guess is that "value" means an object for which one of the conditions pred(*i,value) or pred(value,*i) is true, where i is the iterator defined in the Returns section. However, the value type of the input sequence need not be equality-comparable and for this reason the term find(first, last, value) - first is meaningless.

A term such as find_if(first, last, bind2nd(pred,*i)) - first or find_if(first, last, bind1st(pred,*i)) - first might come closer to the intended specification. Binders can only be applied to function objects that have the function call operator declared const, which is not required of predicates because they can have non-const data members. For this reason, a specification using a binder could only be an "as-if" specification.

Proposed resolution:

Change the complexity section in 25.1.5 [lib.alg.adjacent.find] to:

For a nonempty range, exactly min((i - first) + 1, (last - first) - 1) applications of the corresponding predicate, where i is adjacent_find's return value.

[Copenhagen: the original resolution specified an upper bound. The LWG preferred an exact count.]


241. Does unique_copy() require CopyConstructible and Assignable?

Section: 25.2.8 [lib.alg.unique]  Status: WP  Submitter: Angelika Langer  Date: May 15 2000

Some popular implementations of unique_copy() create temporary copies of values in the input sequence, at least if the input iterator is a pointer. Such an implementation is built on the assumption that the value type is CopyConstructible and Assignable.

It is common practice in the standard that algorithms explicitly specify any additional requirements that they impose on any of the types used by the algorithm. An example of an algorithm that creates temporary copies and correctly specifies the additional requirements is accumulate(), 26.4.1 [lib.accumulate].

Since the specifications of unique() and unique_copy() do not require CopyConstructible and Assignable of the InputIterator's value type the above mentioned implementations are not standard-compliant. I cannot judge whether this is a defect in the standard or a defect in the implementations.

Proposed resolution:

In 25.2.8 change:

-4- Requires: The ranges [first, last) and [result, result+(last-first)) shall not overlap.

to:

-4- Requires: The ranges [first, last) and [result, result+(last-first)) shall not overlap. The expression *result = *first must be valid. If neither InputIterator nor OutputIterator meets the requirements of forward iterator then the value type of InputIterator must be copy constructible. Otherwise copy constructible is not required.

[Redmond: the original proposed resolution didn't impose an explicit requirement that the iterator's value type must be copy constructible, on the grounds that an input iterator's value type must always be copy constructible. Not everyone in the LWG thought that this requirement was clear from table 72. It has been suggested that it might be possible to implement unique_copy without requiring assignability, although current implementations do impose that requirement. Howard provided new wording.]

[ Curaçao: The LWG changed the PR editorially to specify "neither...nor...meet..." as clearer than "both...and...do not meet...". Change believed to be so minor as not to require re-review. ]


242. Side effects of function objects

Section: 25.2.3 [lib.alg.transform], 26.4 [lib.numeric.ops]  Status: WP  Submitter: Angelika Langer  Date: May 15 2000

The algorithms transform(), accumulate(), inner_product(), partial_sum(), and adjacent_difference() require that the function object supplied to them shall not have any side effects.

The standard defines a side effect in 1.9 [intro.execution] as:

-7- Accessing an object designated by a volatile lvalue (basic.lval), modifying an object, calling a library I/O function, or calling a function that does any of those operations are all side effects, which are changes in the state of the execution environment.

As a consequence, the function call operator of a function object supplied to any of the algorithms listed above cannot modify data members, cannot invoke any function that has a side effect, and cannot even create and modify temporary objects.  It is difficult to imagine a function object that is still useful under these severe limitations. For instance, any non-trivial transformator supplied to transform() might involve creation and modification of temporaries, which is prohibited according to the current wording of the standard.

On the other hand, popular implementations of these algorithms exhibit uniform and predictable behavior when invoked with a side-effect-producing function objects. It looks like the strong requirement is not needed for efficient implementation of these algorithms.

The requirement of  side-effect-free function objects could be replaced by a more relaxed basic requirement (which would hold for all function objects supplied to any algorithm in the standard library):

A function objects supplied to an algorithm shall not invalidate any iterator or sequence that is used by the algorithm. Invalidation of the sequence includes destruction of the sorting order if the algorithm relies on the sorting order (see section 25.3 - Sorting and related operations [lib.alg.sorting]).

I can't judge whether it is intended that the function objects supplied to transform(), accumulate(), inner_product(), partial_sum(), or adjacent_difference() shall not modify sequence elements through dereferenced iterators.

It is debatable whether this issue is a defect or a change request. Since the consequences for user-supplied function objects are drastic and limit the usefulness of the algorithms significantly I would consider it a defect.

Proposed resolution:

Things to notice about these changes:

  1. The fully-closed ("[]" as opposed to half-closed "[)" ranges are intentional. we want to prevent side-effects from invalidating the end iterators.
  2. That has the unintentional side-effect of prohibiting modification of the end element as a side-effect. This could conceivably be significant in some cases.
  3. The wording also prevents side-effects from modifying elements of the output sequence. I can't imagine why anyone would want to do this, but it is arguably a restriction that implementors don't need to place on users.
  4. Lifting the restrictions imposed in #2 and #3 above is possible and simple, but would require more verbiage.

Change 25.2.3/2 from:

-2- Requires: op and binary_op shall not have any side effects.

to:

-2- Requires: in the ranges [first1, last1], [first2, first2 + (last1 - first1)] and [result, result + (last1- first1)], op and binary_op shall neither modify elements nor invalidate iterators or subranges. [Footnote: The use of fully closed ranges is intentional --end footnote]

Change 25.2.3/2 from:

-2- Requires: op and binary_op shall not have any side effects.

to:

-2- Requires: op and binary_op shall not invalidate iterators or subranges, or modify elements in the ranges [first1, last1], [first2, first2 + (last1 - first1)], and [result, result + (last1 - first1)]. [Footnote: The use of fully closed ranges is intentional --end footnote]

Change 26.4.1/2 from:

-2- Requires: T must meet the requirements of CopyConstructible (lib.copyconstructible) and Assignable (lib.container.requirements) types. binary_op shall not cause side effects.

to:

-2- Requires: T must meet the requirements of CopyConstructible (lib.copyconstructible) and Assignable (lib.container.requirements) types. In the range [first, last], binary_op shall neither modify elements nor invalidate iterators or subranges. [Footnote: The use of a fully closed range is intentional --end footnote]

Change 26.4.2/2 from:

-2- Requires: T must meet the requirements of CopyConstructible (lib.copyconstructible) and Assignable (lib.container.requirements) types. binary_op1 and binary_op2 shall not cause side effects.

to:

-2- Requires: T must meet the requirements of CopyConstructible (lib.copyconstructible) and Assignable (lib.container.requirements) types. In the ranges [first, last] and [first2, first2 + (last - first)], binary_op1 and binary_op2 shall neither modify elements nor invalidate iterators or subranges. [Footnote: The use of fully closed ranges is intentional --end footnote]

Change 26.4.3/4 from:

-4- Requires: binary_op is expected not to have any side effects.

to:

-4- Requires: In the ranges [first, last] and [result, result + (last - first)], binary_op shall neither modify elements nor invalidate iterators or subranges. [Footnote: The use of fully closed ranges is intentional --end footnote]

Change 26.4.4/2 from:

-2- Requires: binary_op shall not have any side effects.

to:

-2- Requires: In the ranges [first, last] and [result, result + (last - first)], binary_op shall neither modify elements nor invalidate iterators or subranges. [Footnote: The use of fully closed ranges is intentional --end footnote]

[Toronto: Dave Abrahams supplied wording.]

[Copenhagen: Proposed resolution was modified slightly. Matt added footnotes pointing out that the use of closed ranges was intentional.]


243. get and getline when sentry reports failure

Section: 27.6.1.3 [lib.istream.unformatted]  Status: WP  Submitter: Martin Sebor  Date: May 15 2000

basic_istream<>::get(), and basic_istream<>::getline(), are unclear with respect to the behavior and side-effects of the named functions in case of an error.

27.6.1.3, p1 states that "... If the sentry object returns true, when converted to a value of type bool, the function endeavors to obtain the requested input..." It is not clear from this (or the rest of the paragraph) what precisely the behavior should be when the sentry ctor exits by throwing an exception or when the sentry object returns false. In particular, what is the number of characters extracted that gcount() returns supposed to be?

27.6.1.3 p8 and p19 say about the effects of get() and getline(): "... In any case, it then stores a null character (using charT()) into the next successive location of the array." Is not clear whether this sentence applies if either of the conditions above holds (i.e., when sentry fails).

Proposed resolution:

Add to 27.6.1.3, p1 after the sentence

"... If the sentry object returns true, when converted to a value of type bool, the function endeavors to obtain the requested input."

the following

"Otherwise, if the sentry constructor exits by throwing an exception or if the sentry object returns false, when converted to a value of type bool, the function returns without attempting to obtain any input. In either case the number of extracted characters is set to 0; unformatted input functions taking a character array of non-zero size as an argument shall also store a null character (using charT()) in the first location of the array."

Rationale:

Although the general philosophy of the input functions is that the argument should not be modified upon failure, getline historically added a terminating null unconditionally. Most implementations still do that. Earlier versions of the draft standard had language that made this an unambiguous requirement; those words were moved to a place where their context made them less clear. See Jerry Schwarz's message c++std-lib-7618.


248. time_get fails to set eofbit

Section: 22.2.5 [lib.category.time]  Status: WP  Submitter: Martin Sebor  Date: 22 June 2000

There is no requirement that any of time_get member functions set ios::eofbit when they reach the end iterator while parsing their input. Since members of both the num_get and money_get facets are required to do so (22.2.2.1.2, and 22.2.6.1.2, respectively), time_get members should follow the same requirement for consistency.

Proposed resolution:

Add paragraph 2 to section 22.2.5.1 with the following text:

If the end iterator is reached during parsing by any of the get() member functions, the member sets ios_base::eofbit in err.

Rationale:

Two alternative resolutions were proposed. The LWG chose this one because it was more consistent with the way eof is described for other input facets.


250. splicing invalidates iterators

Section: 23.2.2.4 [lib.list.ops]  Status: WP  Submitter: Brian Parker   Date: 14 Jul 2000

Section 23.2.2.4 [lib.list.ops] states that

  void splice(iterator position, list<T, Allocator>& x);

invalidates all iterators and references to list x.

This is unnecessary and defeats an important feature of splice. In fact, the SGI STL guarantees that iterators to x remain valid after splice.

Proposed resolution:

Add a footnote to 23.2.2.4 [lib.list.ops], paragraph 1:

[Footnote: As specified in 20.1.5 [lib.allocator.requirements], paragraphs 4-5, the semantics described in this clause applies only to the case where allocators compare equal. --end footnote]

In 23.2.2.4 [lib.list.ops], replace paragraph 4 with:

Effects: Inserts the contents of x before position and x becomes empty. Pointers and references to the moved elements of x now refer to those same elements but as members of *this. Iterators referring to the moved elements will continue to refer to their elements, but they now behave as iterators into *this, not into x.

In 23.2.2.4 [lib.list.ops], replace paragraph 7 with:

Effects: Inserts an element pointed to by i from list x before position and removes the element from x. The result is unchanged if position == i or position == ++i. Pointers and references to *i continue to refer to this same element but as a member of *this. Iterators to *i (including i itself) continue to refer to the same element, but now behave as iterators into *this, not into x.

In 23.2.2.4 [lib.list.ops], replace paragraph 12 with:

Requires: [first, last) is a valid range in x. The result is undefined if position is an iterator in the range [first, last). Pointers and references to the moved elements of x now refer to those same elements but as members of *this. Iterators referring to the moved elements will continue to refer to their elements, but they now behave as iterators into *this, not into x.

[pre-Copenhagen: Howard provided wording.]

Rationale:

The original proposed resolution said that iterators and references would remain "valid". The new proposed resolution clarifies what that means. Note that this only applies to the case of equal allocators. From 20.1.5 [lib.allocator.requirements] paragraph 4, the behavior of list when allocators compare nonequal is outside the scope of the standard.


251. basic_stringbuf missing allocator_type

Section: 27.7.1 [lib.stringbuf]  Status: WP  Submitter: Martin Sebor  Date: 28 Jul 2000

The synopsis for the template class basic_stringbuf doesn't list a typedef for the template parameter Allocator. This makes it impossible to determine the type of the allocator at compile time. It's also inconsistent with all other template classes in the library that do provide a typedef for the Allocator parameter.

Proposed resolution:

Add to the synopses of the class templates basic_stringbuf (27.7.1), basic_istringstream (27.7.2), basic_ostringstream (27.7.3), and basic_stringstream (27.7.4) the typedef:

  typedef Allocator allocator_type;

252. missing casts/C-style casts used in iostreams

Section: 27.7 [lib.string.streams]  Status: WP  Submitter: Martin Sebor  Date: 28 Jul 2000

27.7.2.2, p1 uses a C-style cast rather than the more appropriate const_cast<> in the Returns clause for basic_istringstream<>::rdbuf(). The same C-style cast is being used in 27.7.3.2, p1, D.7.2.2, p1, and D.7.3.2, p1, and perhaps elsewhere. 27.7.6, p1 and D.7.2.2, p1 are missing the cast altogether.

C-style casts have not been deprecated, so the first part of this issue is stylistic rather than a matter of correctness.

Proposed resolution:

In 27.7.2.2, p1 replace

  -1- Returns: (basic_stringbuf<charT,traits,Allocator>*)&sb.

with

  -1- Returns: const_cast<basic_stringbuf<charT,traits,Allocator>*>(&sb).

In 27.7.3.2, p1 replace

  -1- Returns: (basic_stringbuf<charT,traits,Allocator>*)&sb.

with

  -1- Returns: const_cast<basic_stringbuf<charT,traits,Allocator>*>(&sb).

In 27.7.6, p1, replace

  -1- Returns: &sb

with

  -1- Returns: const_cast<basic_stringbuf<charT,traits,Allocator>*>(&sb).

In D.7.2.2, p1 replace

  -2- Returns: &sb. 

with

  -2- Returns: const_cast<strstreambuf*>(&sb).

253. valarray helper functions are almost entirely useless

Section: 26.3.2.1 [lib.valarray.cons], 26.3.2.2 [lib.valarray.assign]  Status: WP  Submitter: Robert Klarer  Date: 31 Jul 2000

This discussion is adapted from message c++std-lib-7056 posted November 11, 1999. I don't think that anyone can reasonably claim that the problem described below is NAD.

These valarray constructors can never be called:

   template <class T>
         valarray<T>::valarray(const slice_array<T> &);
   template <class T>
         valarray<T>::valarray(const gslice_array<T> &);
   template <class T>
         valarray<T>::valarray(const mask_array<T> &);
   template <class T>
         valarray<T>::valarray(const indirect_array<T> &);

Similarly, these valarray assignment operators cannot be called:

     template <class T>
     valarray<T> valarray<T>::operator=(const slice_array<T> &);
     template <class T>
     valarray<T> valarray<T>::operator=(const gslice_array<T> &);
     template <class T>
     valarray<T> valarray<T>::operator=(const mask_array<T> &);
     template <class T>
     valarray<T> valarray<T>::operator=(const indirect_array<T> &);

Please consider the following example:

   #include <valarray>
   using namespace std;

   int main()
   {
       valarray<double> va1(12);
       valarray<double> va2(va1[slice(1,4,3)]); // line 1
   }

Since the valarray va1 is non-const, the result of the sub-expression va1[slice(1,4,3)] at line 1 is an rvalue of type const std::slice_array<double>. This slice_array rvalue is then used to construct va2. The constructor that is used to construct va2 is declared like this:

     template <class T>
     valarray<T>::valarray(const slice_array<T> &);

Notice the constructor's const reference parameter. When the constructor is called, a slice_array must be bound to this reference. The rules for binding an rvalue to a const reference are in 8.5.3, paragraph 5 (see also 13.3.3.1.4). Specifically, paragraph 5 indicates that a second slice_array rvalue is constructed (in this case copy-constructed) from the first one; it is this second rvalue that is bound to the reference parameter. Paragraph 5 also requires that the constructor that is used for this purpose be callable, regardless of whether the second rvalue is elided. The copy-constructor in this case is not callable, however, because it is private. Therefore, the compiler should report an error.

Since slice_arrays are always rvalues, the valarray constructor that has a parameter of type const slice_array<T> & can never be called. The same reasoning applies to the three other constructors and the four assignment operators that are listed at the beginning of this post. Furthermore, since these functions cannot be called, the valarray helper classes are almost entirely useless.

Proposed resolution:

slice_array:

gslice_array:

mask_array:

indirect_array:

[Proposed resolution was modified in Santa Cruz: explicitly make copy constructor and copy assignment operators public, instead of removing them.]

Rationale:

Keeping the valarray constructors private is untenable. Merely making valarray a friend of the helper classes isn't good enough, because access to the copy constructor is checked in the user's environment.

Making the assignment operator public is not strictly necessary to solve this problem. A majority of the LWG (straw poll: 13-4) believed we should make the assignment operators public, in addition to the copy constructors, for reasons of symmetry and user expectation.


256. typo in 27.4.4.2, p17: copy_event does not exist

Section: 27.4.4.2 [lib.basic.ios.members]  Status: WP  Submitter: Martin Sebor  Date: 21 Aug 2000

27.4.4.2, p17 says

-17- Before copying any parts of rhs, calls each registered callback pair (fn,index) as (*fn)(erase_event,*this,index). After all parts but exceptions() have been replaced, calls each callback pair that was copied from rhs as (*fn)(copy_event,*this,index).

The name copy_event isn't defined anywhere. The intended name was copyfmt_event.

Proposed resolution:

Replace copy_event with copyfmt_event in the named paragraph.


259. basic_string::operator[] and const correctness

Section: 21.3.4 [lib.string.access]  Status: WP  Submitter: Chris Newton   Date: 27 Aug 2000

Paraphrased from a message that Chris Newton posted to comp.std.c++:

The standard's description of basic_string<>::operator[] seems to violate const correctness.

The standard (21.3.4/1) says that "If pos < size(), returns data()[pos]." The types don't work. The return value of data() is const charT*, but operator[] has a non-const version whose return type is reference.

Proposed resolution:

In section 21.3.4, paragraph 1, change "data()[pos]" to "*(begin() + pos)".


260. Inconsistent return type of istream_iterator::operator++(int)

Section: 24.5.1.2 [lib.istream.iterator.ops]  Status: WP  Submitter: Martin Sebor  Date: 27 Aug 2000

The synopsis of istream_iterator::operator++(int) in 24.5.1 shows it as returning the iterator by value. 24.5.1.2, p5 shows the same operator as returning the iterator by reference. That's incorrect given the Effects clause below (since a temporary is returned). The `&' is probably just a typo.

Proposed resolution:

Change the declaration in 24.5.1.2, p5 from

 istream_iterator<T,charT,traits,Distance>& operator++(int);
 

to

 istream_iterator<T,charT,traits,Distance> operator++(int);
 

(that is, remove the `&').


261. Missing description of istream_iterator::operator!=

Section: 24.5.1.2 [lib.istream.iterator.ops]  Status: WP  Submitter: Martin Sebor  Date: 27 Aug 2000

24.5.1, p3 lists the synopsis for

   template <class T, class charT, class traits, class Distance>
        bool operator!=(const istream_iterator<T,charT,traits,Distance>& x,
                        const istream_iterator<T,charT,traits,Distance>& y);

but there is no description of what the operator does (i.e., no Effects or Returns clause) in 24.5.1.2.

Proposed resolution:

Add paragraph 7 to the end of section 24.5.1.2 with the following text:

   template <class T, class charT, class traits, class Distance>
        bool operator!=(const istream_iterator<T,charT,traits,Distance>& x,
                        const istream_iterator<T,charT,traits,Distance>& y);

-7- Returns: !(x == y).


262. Bitmask operator ~ specified incorrectly

Section: 17.3.2.1.2 [lib.bitmask.types]  Status: WP  Submitter: Beman Dawes  Date: 03 Sep 2000

The ~ operation should be applied after the cast to int_type.

Proposed resolution:

Change 17.3.2.1.2 [lib.bitmask.types] operator~ from:

   bitmask operator~ ( bitmask X )
     { return static_cast< bitmask>(static_cast<int_type>(~ X)); }

to:

   bitmask operator~ ( bitmask X )
     { return static_cast< bitmask>(~static_cast<int_type>(X)); }

263. Severe restriction on basic_string reference counting

Section: 21.3 [lib.basic.string]  Status: WP  Submitter: Kevlin Henney  Date: 04 Sep 2000

The note in paragraph 6 suggests that the invalidation rules for references, pointers, and iterators in paragraph 5 permit a reference- counted implementation (actually, according to paragraph 6, they permit a "reference counted implementation", but this is a minor editorial fix).

However, the last sub-bullet is so worded as to make a reference-counted implementation unviable. In the following example none of the conditions for iterator invalidation are satisfied:

    // first example: "*******************" should be printed twice
    string original = "some arbitrary text", copy = original;
    const string & alias = original;

    string::const_iterator i = alias.begin(), e = alias.end();
    for(string::iterator j = original.begin(); j != original.end(); ++j)
        *j = '*';
    while(i != e)
        cout << *i++;
    cout << endl;
    cout << original << endl;

Similarly, in the following example:

    // second example: "some arbitrary text" should be printed out
    string original = "some arbitrary text", copy = original;
    const string & alias = original;

    string::const_iterator i = alias.begin();
    original.begin();
    while(i != alias.end())
        cout << *i++;

I have tested this on three string implementations, two of which were reference counted. The reference-counted implementations gave "surprising behavior" because they invalidated iterators on the first call to non-const begin since construction. The current wording does not permit such invalidation because it does not take into account the first call since construction, only the first call since various member and non-member function calls.

Proposed resolution:

Change the following sentence in 21.3 paragraph 5 from

Subsequent to any of the above uses except the forms of insert() and erase() which return iterators, the first call to non-const member functions operator[](), at(), begin(), rbegin(), end(), or rend().

to

Following construction or any of the above uses, except the forms of insert() and erase() that return iterators, the first call to non- const member functions operator[](), at(), begin(), rbegin(), end(), or rend().

264. Associative container insert(i, j) complexity requirements are not feasible.

Section: 23.1.2 [lib.associative.reqmts]  Status: WP  Submitter: John Potter  Date: 07 Sep 2000

Table 69 requires linear time if [i, j) is sorted. Sorted is necessary but not sufficient. Consider inserting a sorted range of even integers into a set<int> containing the odd integers in the same range.

Related issue: 102

Proposed resolution:

In Table 69, in section 23.1.2, change the complexity clause for insertion of a range from "N log(size() + N) (N is the distance from i to j) in general; linear if [i, j) is sorted according to value_comp()" to "N log(size() + N), where N is the distance from i to j".

[Copenhagen: Minor fix in proposed resolution: fixed unbalanced parens in the revised wording.]

Rationale:

Testing for valid insertions could be less efficient than simply inserting the elements when the range is not both sorted and between two adjacent existing elements; this could be a QOI issue.

The LWG considered two other options: (a) specifying that the complexity was linear if [i, j) is sorted according to value_comp() and between two adjacent existing elements; or (b) changing to Klog(size() + N) + (N - K) (N is the distance from i to j and K is the number of elements which do not insert immediately after the previous element from [i, j) including the first). The LWG felt that, since we can't guarantee linear time complexity whenever the range to be inserted is sorted, it's more trouble than it's worth to say that it's linear in some special cases.


265. std::pair::pair() effects overly restrictive

Section: 20.2.2 [lib.pairs]  Status: WP  Submitter: Martin Sebor  Date: 11 Sep 2000

I don't see any requirements on the types of the elements of the std::pair container in 20.2.2. From the descriptions of the member functions it appears that they must at least satisfy the requirements of 20.1.3 [lib.copyconstructible] and 20.1.4 [lib.default.con.req], and in the case of the [in]equality operators also the requirements of 20.1.1 [lib.equalitycomparable] and 20.1.2 [lib.lessthancomparable].

I believe that the the CopyConstructible requirement is unnecessary in the case of 20.2.2, p2.

Proposed resolution:

Change the Effects clause in 20.2.2, p2 from

-2- Effects: Initializes its members as if implemented: pair() : first(T1()), second(T2()) {}

to

-2- Effects: Initializes its members as if implemented: pair() : first(), second() {}

Rationale:

The existing specification of pair's constructor appears to be a historical artifact: there was concern that pair's members be properly zero-initialized when they are built-in types. At one time there was uncertainty about whether they would be zero-initialized if the default constructor was written the obvious way. This has been clarified by core issue 178, and there is no longer any doubt that the straightforward implementation is correct.


266. bad_exception::~bad_exception() missing Effects clause

Section: 18.6.2.1 [lib.bad.exception]  Status: WP  Submitter: Martin Sebor  Date: 24 Sep 2000

The synopsis for std::bad_exception lists the function ~bad_exception() but there is no description of what the function does (the Effects clause is missing).

Proposed resolution:

Remove the destructor from the class synopses of bad_alloc (18.4.2.1 [lib.bad.alloc]), bad_cast (18.5.2 [lib.bad.cast]), bad_typeid (18.5.3 [lib.bad.typeid]), and bad_exception (18.6.2.1 [lib.bad.exception]).

Rationale:

This is a general problem with the exception classes in clause 18. The proposed resolution is to remove the destructors from the class synopses, rather than to document the destructors' behavior, because removing them is more consistent with how exception classes are described in clause 19.


268. Typo in locale synopsis

Section: 22.1.1 [lib.locale]  Status: WP  Submitter: Martin Sebor  Date: 5 Oct 2000

The synopsis of the class std::locale in 22.1.1 contains two typos: the semicolons after the declarations of the default ctor locale::locale() and the copy ctor locale::locale(const locale&) are missing.

Proposed resolution:

Add the missing semicolons, i.e., change

    //  construct/copy/destroy:
        locale() throw()
        locale(const locale& other) throw()

in the synopsis in 22.1.1 to

    //  construct/copy/destroy:
        locale() throw();
        locale(const locale& other) throw();

270. Binary search requirements overly strict

Section: 25.3.3 [lib.alg.binary.search]  Status: WP  Submitter: Matt Austern  Date: 18 Oct 2000

Each of the four binary search algorithms (lower_bound, upper_bound, equal_range, binary_search) has a form that allows the user to pass a comparison function object. According to 25.3, paragraph 2, that comparison function object has to be a strict weak ordering.

This requirement is slightly too strict. Suppose we are searching through a sequence containing objects of type X, where X is some large record with an integer key. We might reasonably want to look up a record by key, in which case we would want to write something like this:

    struct key_comp {
      bool operator()(const X& x, int n) const {
        return x.key() < n;
      }
    }

    std::lower_bound(first, last, 47, key_comp());

key_comp is not a strict weak ordering, but there is no reason to prohibit its use in lower_bound.

There's no difficulty in implementing lower_bound so that it allows the use of something like key_comp. (It will probably work unless an implementor takes special pains to forbid it.) What's difficult is formulating language in the standard to specify what kind of comparison function is acceptable. We need a notion that's slightly more general than that of a strict weak ordering, one that can encompass a comparison function that involves different types. Expressing that notion may be complicated.

Additional questions raised at the Toronto meeting:

Additional discussion from Copenhagen:

Proposed resolution:

Change 25.3 [lib.alg.sorting] paragraph 3 from:

3 For all algorithms that take Compare, there is a version that uses operator< instead. That is, comp(*i, *j) != false defaults to *i < *j != false. For the algorithms to work correctly, comp has to induce a strict weak ordering on the values.

to:

3 For all algorithms that take Compare, there is a version that uses operator< instead. That is, comp(*i, *j) != false defaults to *i < *j != false. For algorithms other than those described in lib.alg.binary.search (25.3.3) to work correctly, comp has to induce a strict weak ordering on the values.

Add the following paragraph after 25.3 [lib.alg.sorting] paragraph 5:

-6- A sequence [start, finish) is partitioned with respect to an expression f(e) if there exists an integer n such that for all 0 <= i < distance(start, finish), f(*(begin+i)) is true if and only if i < n.

Change 25.3.3 [lib.alg.binary.search] paragraph 1 from:

-1- All of the algorithms in this section are versions of binary search and assume that the sequence being searched is in order according to the implied or explicit comparison function. They work on non-random access iterators minimizing the number of comparisons, which will be logarithmic for all types of iterators. They are especially appropriate for random access iterators, because these algorithms do a logarithmic number of steps through the data structure. For non-random access iterators they execute a linear number of steps.

to:

-1- All of the algorithms in this section are versions of binary search and assume that the sequence being searched is partitioned with respect to an expression formed by binding the search key to an argument of the implied or explicit comparison function. They work on non-random access iterators minimizing the number of comparisons, which will be logarithmic for all types of iterators. They are especially appropriate for random access iterators, because these algorithms do a logarithmic number of steps through the data structure. For non-random access iterators they execute a linear number of steps.

Change 25.3.3.1 [lib.lower.bound] paragraph 1 from:

-1- Requires: Type T is LessThanComparable (lib.lessthancomparable).

to:

-1- Requires: The elements e of [first, last) are partitioned with respect to the expression e < value or comp(e, value)

Remove 25.3.3.1 [lib.lower.bound] paragraph 2:

-2- Effects: Finds the first position into which value can be inserted without violating the ordering.

Change 25.3.3.2 [lib.upper.bound] paragraph 1 from:

-1- Requires: Type T is LessThanComparable (lib.lessthancomparable).

to:

-1- Requires: The elements e of [first, last) are partitioned with respect to the expression !(value < e) or !comp(value, e)

Remove 25.3.3.2 [lib.upper.bound] paragraph 2:

-2- Effects: Finds the furthermost position into which value can be inserted without violating the ordering.

Change 25.3.3.3 [lib.equal.range] paragraph 1 from:

-1- Requires: Type T is LessThanComparable (lib.lessthancomparable).

to:

-1- Requires: The elements e of [first, last) are partitioned with respect to the expressions e < value and !(value < e) or comp(e, value) and !comp(value, e). Also, for all elements e of [first, last), e < value implies !(value < e) or comp(e, value) implies !comp(value, e)

Change 25.3.3.3 [lib.equal.range] paragraph 2 from:

-2- Effects: Finds the largest subrange [i, j) such that the value can be inserted at any iterator k in it without violating the ordering. k satisfies the corresponding conditions: !(*k < value) && !(value < *k) or comp(*k, value) == false && comp(value, *k) == false.

to:

   -2- Returns: 
         make_pair(lower_bound(first, last, value),
                   upper_bound(first, last, value))
       or
         make_pair(lower_bound(first, last, value, comp),
                   upper_bound(first, last, value, comp))

Change 25.3.3.3 [lib.binary.search] paragraph 1 from:

-1- Requires: Type T is LessThanComparable (lib.lessthancomparable).

to:

-1- Requires: The elements e of [first, last) are partitioned with respect to the expressions e < value and !(value < e) or comp(e, value) and !comp(value, e). Also, for all elements e of [first, last), e < value implies !(value < e) or comp(e, value) implies !comp(value, e)

[Copenhagen: Dave Abrahams provided this wording]

[Redmond: Minor changes in wording. (Removed "non-negative", and changed the "other than those described in" wording.) Also, the LWG decided to accept the "optional" part.]

Rationale:

The proposed resolution reinterprets binary search. Instead of thinking about searching for a value in a sorted range, we view that as an important special case of a more general algorithm: searching for the partition point in a partitioned range.

We also add a guarantee that the old wording did not: we ensure that the upper bound is no earlier than the lower bound, that the pair returned by equal_range is a valid range, and that the first part of that pair is the lower bound.


271. basic_iostream missing typedefs

Section: 27.6.1.5 [lib.iostreamclass]  Status: WP  Submitter: Martin Sebor  Date: 02 Nov 2000

Class template basic_iostream has no typedefs. The typedefs it inherits from its base classes can't be used, since (for example) basic_iostream<T>::traits_type is ambiguous.

Proposed resolution:

Add the following to basic_iostream's class synopsis in 27.6.1.5 [lib.iostreamclass], immediately after public:

  // types:
  typedef charT                     char_type;
  typedef typename traits::int_type int_type;
  typedef typename traits::pos_type pos_type;
  typedef typename traits::off_type off_type;
  typedef traits                    traits_type;

272. Missing parentheses around subexpression

Section: 27.4.4.3 [lib.iostate.flags]  Status: WP  Submitter: Martin Sebor  Date: 02 Nov 2000

27.4.4.3, p4 says about the postcondition of the function: If rdbuf()!=0 then state == rdstate(); otherwise rdstate()==state|ios_base::badbit.

The expression on the right-hand-side of the operator==() needs to be parenthesized in order for the whole expression to ever evaluate to anything but non-zero.

Proposed resolution:

Add parentheses like so: rdstate()==(state|ios_base::badbit).


273. Missing ios_base qualification on members of a dependent class

Section: 27 [lib.input.output]  Status: WP  Submitter: Martin Sebor  Date: 02 Nov 2000

27.5.2.4.2, p4, and 27.8.1.6, p2, 27.8.1.7, p3, 27.8.1.9, p2, 27.8.1.10, p3 refer to in and/or out w/o ios_base:: qualification. That's incorrect since the names are members of a dependent base class (14.6.2 [temp.dep]) and thus not visible.

Proposed resolution:

Qualify the names with the name of the class of which they are members, i.e., ios_base.


274. a missing/impossible allocator requirement

Section: 20.1.5 [lib.allocator.requirements]  Status: WP  Submitter: Martin Sebor  Date: 02 Nov 2000

I see that table 31 in 20.1.5, p3 allows T in std::allocator<T> to be of any type. But the synopsis in 20.4.1 calls for allocator<>::address() to be overloaded on reference and const_reference, which is ill-formed for all T = const U. In other words, this won't work:

template class std::allocator<const int>;

The obvious solution is to disallow specializations of allocators on const types. However, while containers' elements are required to be assignable (which rules out specializations on const T's), I think that allocators might perhaps be potentially useful for const values in other contexts. So if allocators are to allow const types a partial specialization of std::allocator<const T> would probably have to be provided.

Proposed resolution:

Change the text in row 1, column 2 of table 32 in 20.1.5, p3 from

any type

to

any non-const, non-reference type

[Redmond: previous proposed resolution was "any non-const, non-volatile, non-reference type". Got rid of the "non-volatile".]

Rationale:

Two resolutions were originally proposed: one that partially specialized std::allocator for const types, and one that said an allocator's value type may not be const. The LWG chose the second. The first wouldn't be appropriate, because allocators are intended for use by containers, and const value types don't work in containers. Encouraging the use of allocators with const value types would only lead to unsafe code.

The original text for proposed resolution 2 was modified so that it also forbids volatile types and reference types.

[Curaçao: LWG double checked and believes volatile is correctly excluded from the PR.]


275. Wrong type in num_get::get() overloads

Section: 22.2.2.1.1 [lib.facet.num.get.members]  Status: WP  Submitter: Matt Austern  Date: 02 Nov 2000

In 22.2.2.1.1, we have a list of overloads for num_get<>::get(). There are eight overloads, all of which are identical except for the last parameter. The overloads are:

There is a similar list, in 22.2.2.1.2, of overloads for num_get<>::do_get(). In this list, the last parameter has the types:

These two lists are not identical. They should be, since get is supposed to call do_get with exactly the arguments it was given.

Proposed resolution:

In 22.2.2.1.1 [lib.facet.num.get.members], change

  iter_type get(iter_type in, iter_type end, ios_base& str,
                ios_base::iostate& err, short& val) const;

to

  iter_type get(iter_type in, iter_type end, ios_base& str,
                ios_base::iostate& err, float& val) const;

276. Assignable requirement for container value type overly strict

Section: 23.1 [lib.container.requirements]  Status: WP  Submitter: Peter Dimov  Date: 07 Nov 2000

23.1/3 states that the objects stored in a container must be Assignable. 23.3.1 [lib.map], paragraph 2, states that map satisfies all requirements for a container, while in the same time defining value_type as pair<const Key, T> - a type that is not Assignable.

It should be noted that there exists a valid and non-contradictory interpretation of the current text. The wording in 23.1/3 avoids mentioning value_type, referring instead to "objects stored in a container." One might argue that map does not store objects of type map::value_type, but of map::mapped_type instead, and that the Assignable requirement applies to map::mapped_type, not map::value_type.

However, this makes map a special case (other containers store objects of type value_type) and the Assignable requirement is needlessly restrictive in general.

For example, the proposed resolution of active library issue 103 is to make set::iterator a constant iterator; this means that no set operations can exploit the fact that the stored objects are Assignable.

This is related to, but slightly broader than, closed issue 140.

Proposed resolution:

23.1/3: Strike the trailing part of the sentence:

, and the additional requirements of Assignable types from 23.1/3

so that it reads:

-3- The type of objects stored in these components must meet the requirements of CopyConstructible types (lib.copyconstructible).

23.1/4: Modify to make clear that this requirement is not for all containers. Change to:

-4- Table 64 defines the Assignable requirement. Some containers require this property of the types to be stored in the container. T is the type used to instantiate the container. t is a value of T, and u is a value of (possibly const) T.

23.1, Table 65: in the first row, change "T is Assignable" to "T is CopyConstructible".

23.2.1/2: Add sentence for Assignable requirement. Change to:

-2- A deque satisfies all of the requirements of a container and of a reversible container (given in tables in lib.container.requirements) and of a sequence, including the optional sequence requirements (lib.sequence.reqmts). In addition to the requirements on the stored object described in 23.1[lib.container.requirements], the stored object must also meet the requirements of Assignable. Descriptions are provided here only for operations on deque that are not described in one of these tables or for operations where there is additional semantic information.

23.2.2/2: Add Assignable requirement to specific methods of list. Change to:

-2- A list satisfies all of the requirements of a container and of a reversible container (given in two tables in lib.container.requirements) and of a sequence, including most of the the optional sequence requirements (lib.sequence.reqmts). The exceptions are the operator[] and at member functions, which are not provided. [Footnote: These member functions are only provided by containers whose iterators are random access iterators. --- end foonote]

list does not require the stored type T to be Assignable unless the following methods are instantiated: [Footnote: Implementors are permitted but not required to take advantage of T's Assignable properties for these methods. -- end foonote]

     list<T,Allocator>& operator=(const list<T,Allocator>&  x );
     template <class InputIterator>
       void assign(InputIterator first, InputIterator last);
     void assign(size_type n, const T& t);

Descriptions are provided here only for operations on list that are not described in one of these tables or for operations where there is additional semantic information.

23.2.4/2: Add sentence for Assignable requirement. Change to:

-2- A vector satisfies all of the requirements of a container and of a reversible container (given in two tables in lib.container.requirements) and of a sequence, including most of the optional sequence requirements (lib.sequence.reqmts). The exceptions are the push_front and pop_front member functions, which are not provided. In addition to the requirements on the stored object described in 23.1[lib.container.requirements], the stored object must also meet the requirements of Assignable. Descriptions are provided here only for operations on vector that are not described in one of these tables or for operations where there is additional semantic information.

Rationale:

list, set, multiset, map, multimap are able to store non-Assignables. However, there is some concern about list<T>: although in general there's no reason for T to be Assignable, some implementations of the member functions operator= and assign do rely on that requirement. The LWG does not want to forbid such implementations.

Note that the type stored in a standard container must still satisfy the requirements of the container's allocator; this rules out, for example, such types as "const int". See issue 274 for more details.

In principle we could also relax the "Assignable" requirement for individual vector member functions, such as push_back. However, the LWG did not see great value in such selective relaxation. Doing so would remove implementors' freedom to implement vector::push_back in terms of vector::insert.


278. What does iterator validity mean?

Section: 23.2.2.4 [lib.list.ops]  Status: WP  Submitter: P.J. Plauger  Date: 27 Nov 2000

Section 23.2.2.4 [lib.list.ops] states that

  void splice(iterator position, list<T, Allocator>& x);

invalidates all iterators and references to list x.

But what does the C++ Standard mean by "invalidate"? You can still dereference the iterator to a spliced list element, but you'd better not use it to delimit a range within the original list. For the latter operation, it has definitely lost some of its validity.

If we accept the proposed resolution to issue 250, then we'd better clarify that a "valid" iterator need no longer designate an element within the same container as it once did. We then have to clarify what we mean by invalidating a past-the-end iterator, as when a vector or string grows by reallocation. Clearly, such an iterator has a different kind of validity. Perhaps we should introduce separate terms for the two kinds of "validity."

Proposed resolution:

Add the following text to the end of section 24.1 [lib.iterator.requirements], after paragraph 5:

An invalid iterator is an iterator that may be singular. [Footnote: This definition applies to pointers, since pointers are iterators. The effect of dereferencing an iterator that has been invalidated is undefined.]

[post-Copenhagen: Matt provided wording.]

[Redmond: General agreement with the intent, some objections to the wording. Dave provided new wording.]

Rationale:

This resolution simply defines a term that the Standard uses but never defines, "invalid", in terms of a term that is defined, "singular".

Why do we say "may be singular", instead of "is singular"? That's becuase a valid iterator is one that is known to be nonsingular. Invalidating an iterator means changing it in such a way that it's no longer known to be nonsingular. An example: inserting an element into the middle of a vector is correctly said to invalidate all iterators pointing into the vector. That doesn't necessarily mean they all become singular.


281. std::min() and max() requirements overly restrictive

Section: 25.3.7 [lib.alg.min.max]  Status: WP  Submitter: Martin Sebor  Date: 02 Dec 2000

The requirements in 25.3.7, p1 and 4 call for T to satisfy the requirements of LessThanComparable (20.1.2 [lib.lessthancomparable]) and CopyConstructible (20.1.3 [lib.copyconstructible]). Since the functions take and return their arguments and result by const reference, I believe the CopyConstructible requirement is unnecessary.

Proposed resolution:

Remove the CopyConstructible requirement. Specifically, replace 25.3.7, p1 with

-1- Requires: Type T is LessThanComparable (20.1.2 [lib.lessthancomparable]).

and replace 25.3.7, p4 with

-4- Requires: Type T is LessThanComparable (20.1.2 [lib.lessthancomparable]).


282. What types does numpunct grouping refer to?

Section: 22.2.2.2.2 [lib.facet.num.put.virtuals]  Status: WP  Submitter: Howard Hinnant  Date: 5 Dec 2000

Paragraph 16 mistakenly singles out integral types for inserting thousands_sep() characters. This conflicts with the syntax for floating point numbers described under 22.2.3.1/2.

Proposed resolution:

Change paragraph 16 from:

For integral types, punct.thousands_sep() characters are inserted into the sequence as determined by the value returned by punct.do_grouping() using the method described in 22.2.3.1.2 [lib.facet.numpunct.virtuals].

To:

For arithmetic types, punct.thousands_sep() characters are inserted into the sequence as determined by the value returned by punct.do_grouping() using the method described in 22.2.3.1.2 [lib.facet.numpunct.virtuals].

[ Copenhagen: Opinions were divided about whether this is actually an inconsistency, but at best it seems to have been unintentional. This is only an issue for floating-point output: The standard is unambiguous that implementations must parse thousands_sep characters when performing floating-point. The standard is also unambiguous that this requirement does not apply to the "C" locale. ]

[ A survey of existing practice is needed; it is believed that some implementations do insert thousands_sep characters for floating-point output and others fail to insert thousands_sep characters for floating-point input even though this is unambiguously required by the standard. ]

[Post-Curaçao: the above proposed resolution is the consensus of Howard, Bill, Pete, Benjamin, Nathan, Dietmar, Boris, and Martin.]


283. std::replace() requirement incorrect/insufficient

Section: 25.2.4 [lib.alg.replace]  Status: WP  Submitter: Martin Sebor  Date: 15 Dec 2000

(revision of the further discussion) There are a number of problems with the requires clauses for the algorithms in 25.1 and 25.2. The requires clause of each algorithm should describe the necessary and sufficient requirements on the inputs to the algorithm such that the algorithm compiles and runs properly. Many of the requires clauses fail to do this. Here is a summary of the kinds of mistakes:

  1. Use of EqualityComparable, which only puts requirements on a single type, when in fact an equality operator is required between two different types, typically either T and the iterator's value type or between the value types of two different iterators.
  2. Use of Assignable for T when in fact what was needed is Assignable for the value_type of the iterator, and convertability from T to the value_type of the iterator. Or for output iterators, the requirement should be that T is writable to the iterator (output iterators do not have value types).

Here is the list of algorithms that contain mistakes:

Also, in the requirements for EqualityComparable, the requirement that the operator be defined for const objects is lacking.

Proposed resolution:

20.1.1 Change p1 from

In Table 28, T is a type to be supplied by a C++ program instantiating a template, a, b, and c are values of type T.

to

In Table 28, T is a type to be supplied by a C++ program instantiating a template, a, b, and c are values of type const T.

25 Between p8 and p9

Add the following sentence:

When the description of an algorithm gives an expression such as *first == value for a condition, it is required that the expression evaluate to either true or false in boolean contexts.

25.1.2 Change p1 by deleting the requires clause.

25.1.6 Change p1 by deleting the requires clause.

25.1.9

Change p4 from

-4- Requires: Type T is EqualityComparable (20.1.1), type Size is convertible to integral type (4.7.12.3).

to

-4- Requires: The type Size is convertible to integral type (4.7.12.3).

25.2.4 Change p1 from

-1- Requires: Type T is Assignable (23.1 ) (and, for replace(), EqualityComparable (20.1.1 )).

to

-1- Requires: The expression *first = new_value must be valid.

and change p4 from

-4- Requires: Type T is Assignable (23.1) (and, for replace_copy(), EqualityComparable (20.1.1)). The ranges [first, last) and [result, result + (last - first)) shall not overlap.

to

-4- Requires: The results of the expressions *first and new_value must be writable to the result output iterator. The ranges [first, last) and [result, result + (last - first)) shall not overlap.

25.2.5 Change p1 from

-1- Requires: Type T is Assignable (23.1). The type Size is convertible to an integral type (4.7.12.3).

to

-1- Requires: The expression value must be is writable to the output iterator. The type Size is convertible to an integral type (4.7.12.3).

25.2.7 Change p1 from

-1- Requires: Type T is EqualityComparable (20.1.1).

to

-1- Requires: The value type of the iterator must be Assignable (23.1).

Rationale:

The general idea of the proposed solution is to remove the faulty requires clauses and let the returns and effects clauses speak for themselves. That is, the returns clauses contain expressions that must be valid, and therefore already imply the correct requirements. In addition, a sentence is added at the beginning of chapter 25 saying that expressions given as conditions must evaluate to true or false in a boolean context. An alternative would be to say that the type of these condition expressions must be literally bool, but that would be imposing a greater restriction that what the standard currently says (which is convertible to bool).


284. unportable example in 20.3.7, p6

Section: 20.3.7 [lib.function.pointer.adaptors]  Status: WP  Submitter: Martin Sebor  Date: 26 Dec 2000

The example in 20.3.7 [lib.function.pointer.adaptors], p6 shows how to use the C library function strcmp() with the function pointer adapter ptr_fun(). But since it's unspecified whether the C library functions have extern "C" or extern "C++" linkage [17.4.2.2 [lib.using.linkage]], and since function pointers with different the language linkage specifications (7.5 [dcl.link]) are incompatible, whether this example is well-formed is unspecified.

Proposed resolution:

Change 20.3.7 [lib.function.pointer.adaptors] paragraph 6 from:

[Example:

    replace_if(v.begin(), v.end(), not1(bind2nd(ptr_fun(strcmp), "C")), "C++");
  

replaces each C with C++ in sequence v.

to:

[Example:

    int compare(const char*, const char*);
    replace_if(v.begin(), v.end(),
               not1(bind2nd(ptr_fun(compare), "abc")), "def");
  

replaces each abc with def in sequence v.

Also, remove footnote 215 in that same paragraph.

[Copenhagen: Minor change in the proposed resolution. Since this issue deals in part with C and C++ linkage, it was believed to be too confusing for the strings in the example to be "C" and "C++". ]

[Redmond: More minor changes. Got rid of the footnote (which seems to make a sweeping normative requirement, even though footnotes aren't normative), and changed the sentence after the footnote so that it corresponds to the new code fragment.]


285. minor editorial errors in fstream ctors

Section: 27.8.1.6 [lib.ifstream.cons]  Status: WP  Submitter: Martin Sebor  Date: 31 Dec 2000

27.8.1.6 [lib.ifstream.cons], p2, 27.8.1.9 [lib.ofstream.cons], p2, and 27.8.1.12 [lib.fstream.cons], p2 say about the effects of each constructor:

... If that function returns a null pointer, calls setstate(failbit) (which may throw ios_base::failure).

The parenthetical note doesn't apply since the ctors cannot throw an exception due to the requirement in 27.4.4.1 [lib.basic.ios.cons], p3 that exceptions() be initialized to ios_base::goodbit.

Proposed resolution:

Strike the parenthetical note from the Effects clause in each of the paragraphs mentioned above.


286. <cstdlib> requirements missing size_t typedef

Section: 25.4 [lib.alg.c.library]  Status: WP  Submitter: Judy Ward  Date: 30 Dec 2000

The <cstdlib> header file contains prototypes for bsearch and qsort (C++ Standard section 25.4 paragraphs 3 and 4) and other prototypes (C++ Standard section 21.4 paragraph 1 table 49) that require the typedef size_t. Yet size_t is not listed in the <cstdlib> synopsis table 78 in section 25.4.

Proposed resolution:

Add the type size_t to Table 78 (section 25.4) and add the type size_t <cstdlib> to Table 97 (section C.2).

Rationale:

Since size_t is in <stdlib.h>, it must also be in <cstdlib>.


288. <cerrno> requirements missing macro EILSEQ

Section: 19.3 [lib.errno]  Status: WP  Submitter: Judy Ward  Date: 30 Dec 2000

ISO/IEC 9899:1990/Amendment1:1994 Section 4.3 States: "The list of macros defined in <errno.h> is adjusted to include a new macro, EILSEQ"

ISO/IEC 14882:1998(E) section 19.3 does not refer to the above amendment.

Proposed resolution:

Update Table 26 (section 19.3) "Header <cerrno> synopsis" and Table 95 (section C.2) "Standard Macros" to include EILSEQ.


291. Underspecification of set algorithms

Section: 25.3.5 [lib.alg.set.operations]  Status: WP  Submitter: Matt Austern  Date: 03 Jan 2001

The standard library contains four algorithms that compute set operations on sorted ranges: set_union, set_intersection, set_difference, and set_symmetric_difference. Each of these algorithms takes two sorted ranges as inputs, and writes the output of the appropriate set operation to an output range. The elements in the output range are sorted.

The ordinary mathematical definitions are generalized so that they apply to ranges containing multiple copies of a given element. Two elements are considered to be "the same" if, according to an ordering relation provided by the user, neither one is less than the other. So, for example, if one input range contains five copies of an element and another contains three, the output range of set_union will contain five copies, the output range of set_intersection will contain three, the output range of set_difference will contain two, and the output range of set_symmetric_difference will contain two.

Because two elements can be "the same" for the purposes of these set algorithms, without being identical in other respects (consider, for example, strings under case-insensitive comparison), this raises a number of unanswered questions:

The standard should either answer these questions, or explicitly say that the answers are unspecified. I prefer the former option, since, as far as I know, all existing implementations behave the same way.

Proposed resolution:

Add the following to the end of 25.3.5.2 [lib.set.union] paragraph 5:

If [first1, last1) contains m elements that are equivalent to each other and [first2, last2) contains n elements that are equivalent to them, then max(m, n) of these elements will be copied to the output range: all m of these elements from [first1, last1), and the last max(n-m, 0) of them from [first2, last2), in that order.

Add the following to the end of 25.3.5.3 [lib.set.intersection] paragraph 5:

If [first1, last1) contains m elements that are equivalent to each other and [first2, last2) contains n elements that are equivalent to them, the first min(m, n) of those elements from [first1, last1) are copied to the output range.

Add a new paragraph, Notes, after 25.3.5.4 [lib.set.difference] paragraph 4:

If [first1, last1) contains m elements that are equivalent to each other and [first2, last2) contains n elements that are equivalent to them, the last max(m-n, 0) elements from [first1, last1) are copied to the output range.

Add a new paragraph, Notes, after 25.3.5.5 [lib.set.symmetric.difference] paragraph 4:

If [first1, last1) contains m elements that are equivalent to each other and [first2, last2) contains n elements that are equivalent to them, then |m - n| of those elements will be copied to the output range: the last m - n of these elements from [first1, last1) if m > n, and the last n - m of these elements from [first2, last2) if m < n.

[Santa Cruz: it's believed that this language is clearer than what's in the Standard. However, it's also believed that the Standard may already make these guarantees (although not quite in these words). Bill and Howard will check and see whether they think that some or all of these changes may be redundant. If so, we may close this issue as NAD.]

Rationale:

For simple cases, these descriptions are equivalent to what's already in the Standard. For more complicated cases, they describe the behavior of existing implementations.


292. effects of a.copyfmt (a)

Section: 27.4.4.2 [lib.basic.ios.members]  Status: WP  Submitter: Martin Sebor  Date: 05 Jan 2001

The Effects clause of the member function copyfmt() in 27.4.4.2, p15 doesn't consider the case where the left-hand side argument is identical to the argument on the right-hand side, that is (this == &rhs). If the two arguments are identical there is no need to copy any of the data members or call any callbacks registered with register_callback(). Also, as Howard Hinnant points out in message c++std-lib-8149 it appears to be incorrect to allow the object to fire erase_event followed by copyfmt_event since the callback handling the latter event may inadvertently attempt to access memory freed by the former.

Proposed resolution:

Change the Effects clause in 27.4.4.2, p15 from

-15- Effects:Assigns to the member objects of *this the corresponding member objects of rhs, except that...

to

-15- Effects:If (this == &rhs) does nothing. Otherwise assigns to the member objects of *this the corresponding member objects of rhs, except that...

295. Is abs defined in <cmath>?

Section: 26.5 [lib.c.math]  Status: WP  Submitter: Jens Maurer  Date: 12 Jan 2001

Table 80 lists the contents of the <cmath> header. It does not list abs(). However, 26.5, paragraph 6, which lists added signatures present in <cmath>, does say that several overloads of abs() should be defined in <cmath>.

Proposed resolution:

Add abs to Table 80. Also, remove the parenthetical list of functions "(abs(), div(), rand(), srand())" from 26.5 [lib.c.math], paragraph 1.

[Copenhagen: Modified proposed resolution so that it also gets rid of that vestigial list of functions in paragraph 1.]

Rationale:

All this DR does is fix a typo; it's uncontroversial. A separate question is whether we're doing the right thing in putting some overloads in <cmath> that we aren't also putting in <cstdlib>. That's issue 323.


297. const_mem_fun_t<>::argument_type should be const T*

Section: 20.3.8 [lib.member.pointer.adaptors]  Status: WP  Submitter: Martin Sebor  Date: 6 Jan 2001

The class templates const_mem_fun_t in 20.3.8, p8 and const_mem_fun1_t in 20.3.8, p9 derive from unary_function<T*, S>, and binary_function<T*, A, S>, respectively. Consequently, their argument_type, and first_argument_type members, respectively, are both defined to be T* (non-const). However, their function call member operator takes a const T* argument. It is my opinion that argument_type should be const T* instead, so that one can easily refer to it in generic code. The example below derived from existing code fails to compile due to the discrepancy:

template <class T>
void foo (typename T::argument_type arg)   // #1
{
    typename T::result_type (T::*pf) (typename T::argument_type) const =   // #2
        &T::operator();
}

struct X { /* ... */ };

int main ()
{
    const X x;
    foo<std::const_mem_fun_t<void, X> >(&x);   // #3
}

#1 foo() takes a plain unqualified X* as an argument
#2 the type of the pointer is incompatible with the type of the member function
#3 the address of a constant being passed to a function taking a non-const X*

Proposed resolution:

Replace the top portion of the definition of the class template const_mem_fun_t in 20.3.8, p8

template <class S, class T> class const_mem_fun_t
          : public unary_function<T*, S> {

with

template <class S, class T> class const_mem_fun_t
          : public unary_function<const T*, S> {

Also replace the top portion of the definition of the class template const_mem_fun1_t in 20.3.8, p9

template <class S, class T, class A> class const_mem_fun1_t
          : public binary_function<T*, A, S> {

with

template <class S, class T, class A> class const_mem_fun1_t
          : public binary_function<const T*, A, S> {

Rationale:

This is simply a contradiction: the argument_type typedef, and the argument type itself, are not the same.


298. ::operator delete[] requirement incorrect/insufficient

Section: 18.4.1.2 [lib.new.delete.array]  Status: WP  Submitter: John A. Pedretti  Date: 10 Jan 2001

The default behavior of operator delete[] described in 18.4.1.2, p12 - namely that for non-null value of ptr, the operator reclaims storage allocated by the earlier call to the default operator new[] - is not correct in all cases. Since the specified operator new[] default behavior is to call operator new (18.4.1.2, p4, p8), which can be replaced, along with operator delete, by the user, to implement their own memory management, the specified default behavior of operator delete[] must be to call operator delete.

Proposed resolution:

Change 18.4.1.2, p12 from

-12- Default behavior:

to

-12- Default behavior: Calls operator delete(ptr) or operator delete(ptr, std::nothrow) respectively.

and expunge paragraph 13.


300. list::merge() specification incomplete

Section: 23.2.2.4 [lib.list.ops]  Status: WP  Submitter: John Pedretti  Date: 23 Jan 2001

The "Effects" clause for list::merge() (23.2.2.4, p23) appears to be incomplete: it doesn't cover the case where the argument list is identical to *this (i.e., this == &x). The requirement in the note in p24 (below) is that x be empty after the merge which is surely unintended in this case.

Proposed resolution:

In 23.2.2.4 [lib.list.ops], replace paragraps 23-25 with:

23 Effects: if (&x == this) does nothing; otherwise, merges the two sorted ranges [begin(), end()) and [x.begin(), x.end()). The result is a range in which the elements will be sorted in non-decreasing order according to the ordering defined by comp; that is, for every iterator i in the range other than the first, the condition comp(*i, *(i - 1)) will be false.

24 Notes: Stable: if (&x != this), then for equivalent elements in the two original ranges, the elements from the original range [begin(), end()) always precede the elements from the original range [x.begin(), x.end()). If (&x != this) the range [x.begin(), x.end()) is empty after the merge.

25 Complexity: At most size() + x.size() - 1 applications of comp if (&x ! = this); otherwise, no applications of comp are performed. If an exception is thrown other than by a comparison there are no effects.

[Copenhagen: The original proposed resolution did not fix all of the problems in 23.2.2.4 [lib.list.ops], p22-25. Three different paragraphs (23, 24, 25) describe the effects of merge. Changing p23, without changing the other two, appears to introduce contradictions. Additionally, "merges the argument list into the list" is excessively vague.]

[Post-Curaçao: Robert Klarer provided new wording.]


301. basic_string template ctor effects clause omits allocator argument

Section: 21.3.1 [lib.string.cons]  Status: WP  Submitter: Martin Sebor  Date: 27 Jan 2001

The effects clause for the basic_string template ctor in 21.3.1, p15 leaves out the third argument of type Allocator. I believe this to be a mistake.

Proposed resolution:

Replace

-15- Effects: If InputIterator is an integral type, equivalent to

basic_string(static_cast<size_type>(begin), static_cast<value_type>(end))

with

-15- Effects: If InputIterator is an integral type, equivalent to

basic_string(static_cast<size_type>(begin), static_cast<value_type>(end), a)

303. Bitset input operator underspecified

Section: 23.3.5.3 [lib.bitset.operators]  Status: WP  Submitter: Matt Austern  Date: 5 Feb 2001

In 23.3.5.3, we are told that bitset's input operator "Extracts up to N (single-byte) characters from is.", where is is a stream of type basic_istream<charT, traits>.

The standard does not say what it means to extract single byte characters from a stream whose character type, charT, is in general not a single-byte character type. Existing implementations differ.

A reasonable solution will probably involve widen() and/or narrow(), since they are the supplied mechanism for converting a single character between char and arbitrary charT.

Narrowing the input characters is not the same as widening the literals '0' and '1', because there may be some locales in which more than one wide character maps to the narrow character '0'. Narrowing means that alternate representations may be used for bitset input, widening means that they may not be.

Note that for numeric input, num_get<> (22.2.2.1.2/8) compares input characters to widened version of narrow character literals.

From Pete Becker, in c++std-lib-8224:

Different writing systems can have different representations for the digits that represent 0 and 1. For example, in the Unicode representation of the Devanagari script (used in many of the Indic languages) the digit 0 is 0x0966, and the digit 1 is 0x0967. Calling narrow would translate those into '0' and '1'. But Unicode also provides the ASCII values 0x0030 and 0x0031 for for the Latin representations of '0' and '1', as well as code points for the same numeric values in several other scripts (Tamil has no character for 0, but does have the digits 1-9), and any of these values would also be narrowed to '0' and '1'.

...

It's fairly common to intermix both native and Latin representations of numbers in a document. So I think the rule has to be that if a wide character represents a digit whose value is 0 then the bit should be cleared; if it represents a digit whose value is 1 then the bit should be set; otherwise throw an exception. So in a Devanagari locale, both 0x0966 and 0x0030 would clear the bit, and both 0x0967 and 0x0031 would set it. Widen can't do that. It would pick one of those two values, and exclude the other one.

From Jens Maurer, in c++std-lib-8233:

Whatever we decide, I would find it most surprising if bitset conversion worked differently from int conversion with regard to alternate local representations of numbers.

Thus, I think the options are:

Proposed resolution:

Replace the first two sentences of paragraph 5 with:

Extracts up to N characters from is. Stores these characters in a temporary object str of type basic_string<charT, traits>, then evaluates the expression x = bitset<N>(str).

Replace the third bullet item in paragraph 5 with:

Rationale:

Input for bitset should work the same way as numeric input. Using widen does mean that alternative digit representations will not be recognized, but this was a known consequence of the design choice.


305. Default behavior of codecvt<wchar_t, char, mbstate_t>::length()

Section: 22.2.1.5.2 [lib.locale.codecvt.virtuals]  Status: WP  Submitter: Howard Hinnant  Date: 24 Jan 2001

22.2.1.5/3 introduces codecvt in part with:

codecvt<wchar_t,char,mbstate_t> converts between the native character sets for tiny and wide characters. Instantiations on mbstate_t perform conversion between encodings known to the library implementor.

But 22.2.1.5.2/10 describes do_length in part with:

... codecvt<wchar_t, char, mbstate_t> ... return(s) the lesser of max and (from_end-from).

The semantics of do_in and do_length are linked. What one does must be consistent with what the other does. 22.2.1.5/3 leads me to believe that the vendor is allowed to choose the algorithm that codecvt<wchar_t,char,mbstate_t>::do_in performs so that it makes his customers happy on a given platform. But 22.2.1.5.2/10 explicitly says what codecvt<wchar_t,char,mbstate_t>::do_length must return. And thus indirectly specifies the algorithm that codecvt<wchar_t,char,mbstate_t>::do_in must perform. I believe that this is not what was intended and is a defect.

Discussion from the -lib reflector:
This proposal would have the effect of making the semantics of all of the virtual functions in codecvt<wchar_t, char, mbstate_t> implementation specified. Is that what we want, or do we want to mandate specific behavior for the base class virtuals and leave the implementation specified behavior for the codecvt_byname derived class? The tradeoff is that former allows implementors to write a base class that actually does something useful, while the latter gives users a way to get known and specified---albeit useless---behavior, and is consistent with the way the standard handles other facets. It is not clear what the original intention was.

Nathan has suggest a compromise: a character that is a widened version of the characters in the basic execution character set must be converted to a one-byte sequence, but there is no such requirement for characters that are not part of the basic execution character set.

Proposed resolution:

Change 22.2.1.5.2/5 from:

The instantiations required in Table 51 (lib.locale.category), namely codecvt<wchar_t,char,mbstate_t> and codecvt<char,char,mbstate_t>, store no characters. Stores no more than (to_limit-to) destination elements. It always leaves the to_next pointer pointing one beyond the last element successfully stored.

to:

Stores no more than (to_limit-to) destination elements, and leaves the to_next pointer pointing one beyond the last element successfully stored. codecvt<char,char,mbstate_t> stores no characters.

Change 22.2.1.5.2/10 from:

-10- Returns: (from_next-from) where from_next is the largest value in the range [from,from_end] such that the sequence of values in the range [from,from_next) represents max or fewer valid complete characters of type internT. The instantiations required in Table 51 (21.1.1.1.1), namely codecvt<wchar_t, char, mbstate_t> and codecvt<char, char, mbstate_t>, return the lesser of max and (from_end-from).

to:

-10- Returns: (from_next-from) where from_next is the largest value in the range [from,from_end] such that the sequence of values in the range [from,from_next) represents max or fewer valid complete characters of type internT. The instantiation codecvt<char, char, mbstate_t> returns the lesser of max and (from_end-from).

[Redmond: Nathan suggested an alternative resolution: same as above, but require that, in the default encoding, a character from the basic execution character set would map to a single external character. The straw poll was 8-1 in favor of the proposed resolution.]

Rationale:

The default encoding should be whatever users of a given platform would expect to be the most natural. This varies from platform to platform. In many cases there is a preexisting C library, and users would expect the default encoding to be whatever C uses in the default "C" locale. We could impose a guarantee like the one Nathan suggested (a character from the basic execution character set must map to a single external character), but this would rule out important encodings that are in common use: it would rule out JIS, for example, and it would rule out a fixed-width encoding of UCS-4.

[Curaçao: fixed rationale typo at the request of Ichiro Koshida; "shift-JIS" changed to "JIS".]


306. offsetof macro and non-POD types

Section: 18.1 [lib.support.types]  Status: WP  Submitter: Steve Clamage  Date: 21 Feb 2001

Spliced together from reflector messages c++std-lib-8294 and -8295:

18.1, paragraph 5, reads: "The macro offsetof accepts a restricted set of type arguments in this International Standard. type shall be a POD structure or a POD union (clause 9). The result of applying the offsetof macro to a field that is a static data member or a function member is undefined."

For the POD requirement, it doesn't say "no diagnostic required" or "undefined behavior". I read 1.4 [intro.compliance], paragraph 1, to mean that a diagnostic is required. It's not clear whether this requirement was intended. While it's possible to provide such a diagnostic, the extra complication doesn't seem to add any value.

Proposed resolution:

Change 18.1, paragraph 5, to "If type is not a POD structure or a POD union the results are undefined."

[Copenhagen: straw poll was 7-4 in favor. It was generally agreed that requiring a diagnostic was inadvertent, but some LWG members thought that diagnostics should be required whenever possible.]


307. Lack of reference typedefs in container adaptors

Section: 23.2.3 [lib.container.adaptors]  Status: WP  Submitter: Howard Hinnant  Date: 13 Mar 2001

From reflector message c++std-lib-8330. See also lib-8317.

The standard is currently inconsistent in 23.2.3.2 [lib.priority.queue] paragraph 1 and 23.2.3.3 [lib.stack] paragraph 1. 23.2.3.3/1, for example, says:

-1- Any sequence supporting operations back(), push_back() and pop_back() can be used to instantiate stack. In particular, vector (lib.vector), list (lib.list) and deque (lib.deque) can be used.

But this is false: vector<bool> can not be used, because the container adaptors return a T& rather than using the underlying container's reference type.

This is a contradiction that can be fixed by:

  1. Modifying these paragraphs to say that vector<bool> is an exception.
  2. Removing the vector<bool> specialization.
  3. Changing the return types of stack and priority_queue to use reference typedef's.

I propose 3. This does not preclude option 2 if we choose to do it later (see issue 96); the issues are independent. Option 3 offers a small step towards support for proxied containers. This small step fixes a current contradiction, is easy for vendors to implement, is already implemented in at least one popular lib, and does not break any code.

Proposed resolution:

Summary: Add reference and const_reference typedefs to queue, priority_queue and stack. Change return types of "value_type&" to "reference". Change return types of "const value_type&" to "const_reference". Details:

Change 23.2.3.1/1 from:

  namespace std {
    template <class T, class Container = deque<T> >
    class queue {
    public:
      typedef typename Container::value_type            value_type;
      typedef typename Container::size_type             size_type;
      typedef          Container                        container_type;
    protected:
      Container c;

    public:
      explicit queue(const Container& = Container());

      bool      empty() const             { return c.empty(); }
      size_type size()  const             { return c.size(); }
      value_type&       front()           { return c.front(); }
      const value_type& front() const     { return c.front(); }
      value_type&       back()            { return c.back(); }
      const value_type& back() const      { return c.back(); }
      void push(const value_type& x)      { c.push_back(x); }
      void pop()                          { c.pop_front(); }
    };

to:

  namespace std {
    template <class T, class Container = deque<T> >
    class queue {
    public:
      typedef typename Container::value_type            value_type;
      typedef typename Container::reference             reference;
      typedef typename Container::const_reference       const_reference;
      typedef typename Container::value_type            value_type;
      typedef typename Container::size_type             size_type;
      typedef          Container                        container_type;
    protected:
      Container c;

    public:
      explicit queue(const Container& = Container());

      bool      empty() const             { return c.empty(); }
      size_type size()  const             { return c.size(); }
      reference         front()           { return c.front(); }
      const_reference   front() const     { return c.front(); }
      reference         back()            { return c.back(); }
      const_reference   back() const      { return c.back(); }
      void push(const value_type& x)      { c.push_back(x); }
      void pop()                          { c.pop_front(); }
    };

Change 23.2.3.2/1 from:

  namespace std {
    template <class T, class Container = vector<T>,
              class Compare = less<typename Container::value_type> >
    class priority_queue {
    public:
      typedef typename Container::value_type            value_type;
      typedef typename Container::size_type             size_type;
      typedef          Container                        container_type;
    protected:
      Container c;
      Compare comp;

    public:
      explicit priority_queue(const Compare& x = Compare(),
                              const Container& = Container());
      template <class InputIterator>
        priority_queue(InputIterator first, InputIterator last,
                       const Compare& x = Compare(),
                       const Container& = Container());

      bool      empty() const       { return c.empty(); }
      size_type size()  const       { return c.size(); }
      const value_type& top() const { return c.front(); }
      void push(const value_type& x);
      void pop();
    };
                                  //  no equality is provided
  }

to:

  namespace std {
    template <class T, class Container = vector<T>,
              class Compare = less<typename Container::value_type> >
    class priority_queue {
    public:
      typedef typename Container::value_type            value_type;
      typedef typename Container::reference             reference;
      typedef typename Container::const_reference       const_reference;
      typedef typename Container::size_type             size_type;
      typedef          Container                        container_type;
    protected:
      Container c;
      Compare comp;

    public:
      explicit priority_queue(const Compare& x = Compare(),
                              const Container& = Container());
      template <class InputIterator>
        priority_queue(InputIterator first, InputIterator last,
                       const Compare& x = Compare(),
                       const Container& = Container());

      bool      empty() const       { return c.empty(); }
      size_type size()  const       { return c.size(); }
      const_reference   top() const { return c.front(); }
      void push(const value_type& x);
      void pop();
    };
                                  //  no equality is provided
  }

And change 23.2.3.3/1 from:

  namespace std {
    template <class T, class Container = deque<T> >
    class stack {
    public:
      typedef typename Container::value_type            value_type;
      typedef typename Container::size_type             size_type;
      typedef          Container                        container_type;
    protected:
      Container c;

    public:
      explicit stack(const Container& = Container());

      bool      empty() const             { return c.empty(); }
      size_type size()  const             { return c.size(); }
      value_type&       top()             { return c.back(); }
      const value_type& top() const       { return c.back(); }
      void push(const value_type& x)      { c.push_back(x); }
      void pop()                          { c.pop_back(); }
    };

    template <class T, class Container>
      bool operator==(const stack<T, Container>& x,
                      const stack<T, Container>& y);
    template <class T, class Container>
      bool operator< (const stack<T, Container>& x,
                      const stack<T, Container>& y);
    template <class T, class Container>
      bool operator!=(const stack<T, Container>& x,
                      const stack<T, Container>& y);
    template <class T, class Container>
      bool operator> (const stack<T, Container>& x,
                      const stack<T, Container>& y);
    template <class T, class Container>
      bool operator>=(const stack<T, Container>& x,
                      const stack<T, Container>& y);
    template <class T, class Container>
      bool operator<=(const stack<T, Container>& x,
                      const stack<T, Container>& y);
  }

to:

  namespace std {
    template <class T, class Container = deque<T> >
    class stack {
    public:
      typedef typename Container::value_type            value_type;
      typedef typename Container::reference             reference;
      typedef typename Container::const_reference       const_reference;
      typedef typename Container::size_type             size_type;
      typedef          Container                        container_type;
    protected:
      Container c;

    public:
      explicit stack(const Container& = Container());

      bool      empty() const             { return c.empty(); }
      size_type size()  const             { return c.size(); }
      reference         top()             { return c.back(); }
      const_reference   top() const       { return c.back(); }
      void push(const value_type& x)      { c.push_back(x); }
      void pop()                          { c.pop_back(); }
    };

    template <class T, class Container>
      bool operator==(const stack<T, Container>& x,
                      const stack<T, Container>& y);
    template <class T, class Container>
      bool operator< (const stack<T, Container>& x,
                      const stack<T, Container>& y);
    template <class T, class Container>
      bool operator!=(const stack<T, Container>& x,
                      const stack<T, Container>& y);
    template <class T, class Container>
      bool operator> (const stack<T, Container>& x,
                      const stack<T, Container>& y);
    template <class T, class Container>
      bool operator>=(const stack<T, Container>& x,
                      const stack<T, Container>& y);
    template <class T, class Container>
      bool operator<=(const stack<T, Container>& x,
                      const stack<T, Container>& y);
  }

[Copenhagen: This change was discussed before the IS was released and it was deliberately not adopted. Nevertheless, the LWG believes (straw poll: 10-2) that it is a genuine defect.]


308. Table 82 mentions unrelated headers

Section: 27 [lib.input.output]  Status: WP  Submitter: Martin Sebor  Date: 15 Mar 2001

Table 82 in section 27 mentions the header <cstdlib> for String streams (27.7 [lib.string.streams]) and the headers <cstdio> and <cwchar> for File streams (27.8 [lib.file.streams]). It's not clear why these headers are mentioned in this context since they do not define any of the library entities described by the subclauses. According to 17.4.1.1 [lib.contents], only such headers are to be listed in the summary.

Proposed resolution:

Remove <cstdlib> and <cwchar> from Table 82.

[Copenhagen: changed the proposed resolution slightly. The original proposed resolution also said to remove <cstdio> from Table 82. However, <cstdio> is mentioned several times within section 27.8 [lib.file.streams], including 27.8.2 [lib.c.files].]


310. Is errno a macro?

Section: 17.4.1.2 [lib.headers], 19.3 [lib.errno]  Status: WP  Submitter: Steve Clamage  Date: 21 Mar 2001

Exactly how should errno be declared in a conforming C++ header?

The C standard says in 7.1.4 that it is unspecified whether errno is a macro or an identifier with external linkage. In some implementations it can be either, depending on compile-time options. (E.g., on Solaris in multi-threading mode, errno is a macro that expands to a function call, but is an extern int otherwise. "Unspecified" allows such variability.)

The C++ standard:

I find no other references to errno.

We should either explicitly say that errno must be a macro, even though it need not be a macro in C, or else explicitly leave it unspecified. We also need to say something about namespace std. A user who includes <cerrno> needs to know whether to write errno, or ::errno, or std::errno, or else <cerrno> is useless.

Two acceptable fixes:

[ This issue was first raised in 1999, but it slipped through the cracks. ]

Proposed resolution:

Change the Note in section 17.4.1.2p5 from

Note: the names defined as macros in C include the following: assert, errno, offsetof, setjmp, va_arg, va_end, and va_start.

to

Note: the names defined as macros in C include the following: assert, offsetof, setjmp, va_arg, va_end, and va_start.

In section 19.3, change paragraph 2 from

The contents are the same as the Standard C library header <errno.h>.

to

The contents are the same as the Standard C library header <errno.h>, except that errno shall be defined as a macro.

Rationale:

C++ must not leave it up to the implementation to decide whether or not a name is a macro; it must explicitly specify exactly which names are required to be macros. The only one that really works is for it to be a macro.

[Curaçao: additional rationale added.]


311. Incorrect wording in basic_ostream class synopsis

Section: 27.6.2.1 [lib.ostream]  Status: WP  Submitter: Andy Sawyer  Date: 21 Mar 2001

In 27.6.2.1 [lib.ostream], the synopsis of class basic_ostream says:

  // partial specializationss
  template<class traits>
    basic_ostream<char,traits>& operator<<( basic_ostream<char,traits>&,
                                            const char * );

Problems:

Proposed resolution:

In the synopsis in 27.6.2.1 [lib.ostream], remove the // partial specializationss comment. Also remove the same comment (correctly spelled, but still incorrect) from the synopsis in 27.6.2.5.4 [lib.ostream.inserters.character].

[ Pre-Redmond: added 27.6.2.5.4 [lib.ostream.inserters.character] because of Martin's comment in c++std-lib-8939. ]


312. Table 27 is missing headers

Section: 20 [lib.utilities]  Status: WP  Submitter: Martin Sebor  Date: 29 Mar 2001

Table 27 in section 20 lists the header <memory> (only) for Memory (lib.memory) but neglects to mention the headers <cstdlib> and <cstring> that are discussed in 20.4.6 [lib.c.malloc].

Proposed resolution:

Add <cstdlib> and <cstring> to Table 27, in the same row as <memory>.


315. Bad "range" in list::unique complexity

Section: 23.2.2.4 [lib.list.ops]  Status: WP  Submitter: Andy Sawyer  Date: 1 May 2001

23.2.2.4 [lib.list.ops], Para 21 describes the complexity of list::unique as: "If the range (last - first) is not empty, exactly (last - first) -1 applications of the corresponding predicate, otherwise no applications of the predicate)".

"(last - first)" is not a range.

Proposed resolution:

Change the "range" from (last - first) to [first, last).


316. Vague text in Table 69

Section: 23.1.2 [lib.associative.reqmts]  Status: WP  Submitter: Martin Sebor  Date: 4 May 2001

Table 69 says this about a_uniq.insert(t):

inserts t if and only if there is no element in the container with key equivalent to the key of t. The bool component of the returned pair indicates whether the insertion takes place and the iterator component of the pair points to the element with key equivalent to the key of t.

The description should be more specific about exactly how the bool component indicates whether the insertion takes place.

Proposed resolution:

Change the text in question to

...The bool component of the returned pair is true if and only if the insertion takes place...

317. Instantiation vs. specialization of facets

Section: 22 [lib.localization]  Status: WP  Submitter: Martin Sebor  Date: 4 May 2001

The localization section of the standard refers to specializations of the facet templates as instantiations even though the required facets are typically specialized rather than explicitly (or implicitly) instantiated. In the case of ctype<char> and ctype_byname<char> (and the wchar_t versions), these facets are actually required to be specialized. The terminology should be corrected to make it clear that the standard doesn't mandate explicit instantiation (the term specialization encompasses both explicit instantiations and specializations).

Proposed resolution:

In the following paragraphs, replace all occurrences of the word instantiation or instantiations with specialization or specializations, respectively:

22.1.1.1.1, p4, Table 52, 22.2.1.1, p2, 22.2.1.5, p3, 22.2.1.5.1, p5, 22.2.1.5.2, p10, 22.2.2, p2, 22.2.3.1, p1, 22.2.3.1.2, p1, p2 and p3, 22.2.4.1, p1, 22.2.4.1.2, p1, 22,2,5, p1, 22,2,6, p2, 22.2.6.3.2, p7, and Footnote 242.

And change the text in 22.1.1.1.1, p4 from

An implementation is required to provide those instantiations for facet templates identified as members of a category, and for those shown in Table 52:

to

An implementation is required to provide those specializations...

[Nathan will review these changes, and will look for places where explicit specialization is necessary.]

Rationale:

This is a simple matter of outdated language. The language to describe templates was clarified during the standardization process, but the wording in clause 22 was never updated to reflect that change.


318. Misleading comment in definition of numpunct_byname

Section: 22.2.3.2 [lib.locale.numpunct.byname]  Status: WP  Submitter: Martin Sebor  Date: 12 May 2001

The definition of the numpunct_byname template contains the following comment:

    namespace std {
        template <class charT>
        class numpunct_byname : public numpunct<charT> {
    // this class is specialized for char and wchar_t.
        ...

There is no documentation of the specializations and it seems conceivable that an implementation will not explicitly specialize the template at all, but simply provide the primary template.

Proposed resolution:

Remove the comment from the text in 22.2.3.2 and from the proposed resolution of library issue 228.


319. Storage allocation wording confuses "Required behavior", "Requires"

Section: 18.4.1.1 [lib.new.delete.single], 18.4.1.2 [lib.new.delete.array]  Status: WP  Submitter: Beman Dawes  Date: 15 May 2001

The standard specifies 17.3.1.3 [lib.structure.specifications] that "Required behavior" elements describe "the semantics of a function definition provided by either the implementation or a C++ program."

The standard specifies 17.3.1.3 [lib.structure.specifications] that "Requires" elements describe "the preconditions for calling the function."

In the sections noted below, the current wording specifies "Required Behavior" for what are actually preconditions, and thus should be specified as "Requires".

Proposed resolution:

In 18.4.1.1 [lib.new.delete.single] Para 12 Change:

Required behavior: accept a value of ptr that is null or that was returned by an earlier call ...

to:

Requires: the value of ptr is null or the value returned by an earlier call ...

In 18.4.1.2 [lib.new.delete.array] Para 11 Change:

Required behavior: accept a value of ptr that is null or that was returned by an earlier call ...

to:

Requires: the value of ptr is null or the value returned by an earlier call ...


320. list::assign overspecified

Section: 23.2.2.1 [lib.list.cons]  Status: WP  Submitter: Howard Hinnant  Date: 17 May 2001

Section 23.2.2.1, paragraphs 6-8 specify that list assign (both forms) have the "effects" of a call to erase followed by a call to insert.

I would like to document that implementers have the freedom to implement assign by other methods, as long as the end result is the same and the exception guarantee is as good or better than the basic guarantee.

The motivation for this is to use T's assignment operator to recycle existing nodes in the list instead of erasing them and reallocating them with new values. It is also worth noting that, with careful coding, most common cases of assign (everything but assignment with true input iterators) can elevate the exception safety to strong if T's assignment has a nothrow guarantee (with no extra memory cost). Metrowerks does this. However I do not propose that this subtlety be standardized. It is a QoI issue.

Existing practise: Metrowerks and SGI recycle nodes, Dinkumware and Rogue Wave don't.

Proposed resolution:

Change 23.2.2.1/7 from:

Effects:

   erase(begin(), end());
   insert(begin(), first, last);

to:

Effects: Replaces the contents of the list with the range [first, last).

In 23.1.1 [lib.sequence.reqmts], in Table 67 (sequence requirements), add two new rows:

      a.assign(i,j)     void      pre: i,j are not iterators into a.
                                  Replaces elements in a with a copy
                                  of [i, j).

      a.assign(n,t)     void      pre: t is not a reference into a.
                                  Replaces elements in a with n copies
                                  of t.

Change 23.2.2.1/8 from:

Effects:

   erase(begin(), end());
   insert(begin(), n, t);

to:

Effects: Replaces the contents of the list with n copies of t.

[Redmond: Proposed resolution was changed slightly. Previous version made explicit statement about exception safety, which wasn't consistent with the way exception safety is expressed elsewhere. Also, the change in the sequence requirements is new. Without that change, the proposed resolution would have required that assignment of a subrange would have to work. That too would have been overspecification; it would effectively mandate that assignment use a temporary. Howard provided wording. ]

[Curaçao: Made editorial improvement in wording; changed "Replaces elements in a with copies of elements in [i, j)." with "Replaces the elements of a with a copy of [i, j)." Changes not deemed serious enough to requre rereview.]


321. Typo in num_get

Section: 22.2.2.1.2 [lib.facet.num.get.virtuals]  Status: WP  Submitter: Kevin Djang  Date: 17 May 2001

Section 22.2.2.1.2 at p7 states that "A length specifier is added to the conversion function, if needed, as indicated in Table 56." However, Table 56 uses the term "length modifier", not "length specifier".

Proposed resolution:

In 22.2.2.1.2 at p7, change the text "A length specifier is added ..." to be "A length modifier is added ..."

Rationale:

C uses the term "length modifier". We should be consistent.


322. iterator and const_iterator should have the same value type

Section: 23.1 [lib.container.requirements]  Status: WP  Submitter: Matt Austern  Date: 17 May 2001

It's widely assumed that, if X is a container, iterator_traits<X::iterator>::value_type and iterator_traits<X::const_iterator>::value_type should both be X::value_type. However, this is nowhere stated. The language in Table 65 is not precise about the iterators' value types (it predates iterator_traits), and could even be interpreted as saying that iterator_traits<X::const_iterator>::value_type should be "const X::value_type".

Related issue: 279.

Proposed resolution:

In Table 65 ("Container Requirements"), change the return type for X::iterator to "iterator type whose value type is T". Change the return type for X::const_iterator to "constant iterator type whose value type is T".

Rationale:

This belongs as a container requirement, rather than an iterator requirement, because the whole notion of iterator/const_iterator pairs is specific to containers' iterator.

It is existing practice that (for example) iterator_traits<list<int>::const_iterator>::value_type is "int", rather than "const int". This is consistent with the way that const pointers are handled: the standard already requires that iterator_traits<const int*>::value_type is int.


324. Do output iterators have value types?

Section: 24.1.2 [lib.output.iterators]  Status: WP  Submitter: Dave Abrahams  Date: 7 June 2001

Table 73 suggests that output iterators have value types. It requires the expression "*a = t". Additionally, although Table 73 never lists "a = t" or "X(a) = t" in the "expressions" column, it contains a note saying that "a = t" and "X(a) = t" have equivalent (but nowhere specified!) semantics.

According to 24.1/9, t is supposed to be "a value of value type T":

In the following sections, a and b denote values of X, n denotes a value of the difference type Distance, u, tmp, and m denote identifiers, r denotes a value of X&, t denotes a value of value type T.

Two other parts of the standard that are relevant to whether output iterators have value types:

The first of these passages suggests that "*i" is supposed to return a useful value, which contradicts the note in 24.1.2/2 saying that the only valid use of "*i" for output iterators is in an expression of the form "*i = t". The second of these passages appears to contradict Table 73, because it suggests that "*i"'s return value should be void. The second passage is also broken in the case of a an iterator type, like non-const pointers, that satisfies both the output iterator requirements and the forward iterator requirements.

What should the standard say about *i's return value when i is an output iterator, and what should it say about that t is in the expression "*i = t"? Finally, should the standard say anything about output iterators' pointer and reference types?

Proposed resolution:

24.1 p1, change

All iterators i support the expression *i, resulting in a value of some class, enumeration, or built-in type T, called the value type of the iterator.

to

All input iterators i support the expression *i, resulting in a value of some class, enumeration, or built-in type T, called the value type of the iterator. All output iterators support the expression *i = o where o is a value of some type that is in the set of types that are writable to the particular iterator type of i.

24.1 p9, add

o denotes a value of some type that is writable to the output iterator.

Table 73, change

*a = t

to

*r = o

and change

*r++ = t

to

*r++ = o

[post-Redmond: Jeremy provided wording]

Rationale:

The LWG considered two options: change all of the language that seems to imply that output iterators have value types, thus making it clear that output iterators have no value types, or else define value types for output iterator consistently. The LWG chose the former option, because it seems clear that output iterators were never intended to have value types. This was a deliberate design decision, and any language suggesting otherwise is simply a mistake.

A future revision of the standard may wish to revisit this design decision.


325. Misleading text in moneypunct<>::do_grouping

Section: 22.2.6.3.2 [lib.locale.moneypunct.virtuals]  Status: WP  Submitter: Martin Sebor  Date: 02 Jul 2001

The Returns clause in 22.2.6.3.2, p3 says about moneypunct<charT>::do_grouping()

Returns: A pattern defined identically as the result of numpunct<charT>::do_grouping().241)

Footnote 241 then reads

This is most commonly the value "\003" (not "3").

The returns clause seems to imply that the two member functions must return an identical value which in reality may or may not be true, since the facets are usually implemented in terms of struct std::lconv and return the value of the grouping and mon_grouping, respectively. The footnote also implies that the member function of the moneypunct facet (rather than the overridden virtual functions in moneypunct_byname) most commonly return "\003", which contradicts the C standard which specifies the value of "" for the (most common) C locale.

Proposed resolution:

Replace the text in Returns clause in 22.2.6.3.2, p3 with the following:

Returns: A pattern defined identically as, but not necessarily equal to, the result of numpunct<charT>::do_grouping().241)

and replace the text in Footnote 241 with the following:

To specify grouping by 3s the value is "\003", not "3".

Rationale:

The fundamental problem is that the description of the locale facet virtuals serves two purposes: describing the behavior of the base class, and describing the meaning of and constraints on the behavior in arbitrary derived classes. The new wording makes that separation a little bit clearer. The footnote (which is nonnormative) is not supposed to say what the grouping is in the "C" locale or in any other locale. It is just a reminder that the values are interpreted as small integers, not ASCII characters.


327. Typo in time_get facet in table 52

Section: 22.1.1.1.1 [lib.locale.category]  Status: WP  Submitter: Tiki Wan  Date: 06 Jul 2001

The wchar_t versions of time_get and time_get_byname are listed incorrectly in table 52, required instantiations. In both cases the second template parameter is given as OutputIterator. It should instead be InputIterator, since these are input facets.

Proposed resolution:

In table 52, required instantiations, in 22.1.1.1.1 [lib.locale.category], change

    time_get<wchar_t, OutputIterator>
    time_get_byname<wchar_t, OutputIterator>

to

    time_get<wchar_t, InputIterator>
    time_get_byname<wchar_t, InputIterator>

[Redmond: Very minor change in proposed resolution. Original had a typo, wchart instead of wchar_t.]


328. Bad sprintf format modifier in money_put<>::do_put()

Section: 22.2.6.2.2 [lib.locale.money.put.virtuals]  Status: WP  Submitter: Martin Sebor  Date: 07 Jul 2001

The sprintf format string , "%.01f" (that's the digit one), in the description of the do_put() member functions of the money_put facet in 22.2.6.2.2, p1 is incorrect. First, the f format specifier is wrong for values of type long double, and second, the precision of 01 doesn't seem to make sense. What was most likely intended was "%.0Lf"., that is a precision of zero followed by the L length modifier.

Proposed resolution:

Change the format string to "%.0Lf".

Rationale:

Fixes an obvious typo


329. vector capacity, reserve and reallocation

Section: 23.2.4.2 [lib.vector.capacity], 23.2.4.3 [lib.vector.modifiers]  Status: WP  Submitter: Anthony Williams  Date: 13 Jul 2001

There is an apparent contradiction about which circumstances can cause a reallocation of a vector in Section 23.2.4.2 [lib.vector.capacity] and section 23.2.4.3 [lib.vector.modifiers].

23.2.4.2p5 says:

Notes: Reallocation invalidates all the references, pointers, and iterators referring to the elements in the sequence. It is guaranteed that no reallocation takes place during insertions that happen after a call to reserve() until the time when an insertion would make the size of the vector greater than the size specified in the most recent call to reserve().

Which implies if I do

  std::vector<int> vec;
  vec.reserve(23);
  vec.reserve(0);
  vec.insert(vec.end(),1);

then the implementation may reallocate the vector for the insert, as the size specified in the previous call to reserve was zero.

However, the previous paragraphs (23.2.4.2, p1-2) state:

(capacity) Returns: The total number of elements the vector can hold without requiring reallocation

...After reserve(), capacity() is greater or equal to the argument of reserve if reallocation happens; and equal to the previous value of capacity() otherwise...

This implies that vec.capacity() is still 23, and so the insert() should not require a reallocation, as vec.size() is 0. This is backed up by 23.2.4.3p1:

(insert) Notes: Causes reallocation if the new size is greater than the old capacity.

Though this doesn't rule out reallocation if the new size is less than the old capacity, I think the intent is clear.

Proposed resolution:

Change the wording of 23.2.4.2 [lib.vector.capacity] paragraph 5 to:

Notes: Reallocation invalidates all the references, pointers, and iterators referring to the elements in the sequence. It is guaranteed that no reallocation takes place during insertions that happen after a call to reserve() until the time when an insertion would make the size of the vector greater than the value of capacity().

[Redmond: original proposed resolution was modified slightly. In the original, the guarantee was that there would be no reallocation until the size would be greater than the value of capacity() after the most recent call to reserve(). The LWG did not believe that the "after the most recent call to reserve()" added any useful information.]

Rationale:

There was general agreement that, when reserve() is called twice in succession and the argument to the second invocation is smaller than the argument to the first, the intent was for the second invocation to have no effect. Wording implying that such cases have an effect on reallocation guarantees was inadvertant.


331. bad declaration of destructor for ios_base::failure

Section: 27.4.2.1.1 [lib.ios::failure]  Status: WP  Submitter: PremAnand M. Rao  Date: 23 Aug 2001

With the change in 17.4.4.8 [lib.res.on.exception.handling] to state "An implementation may strengthen the exception-specification for a non-virtual function by removing listed exceptions." (issue 119) and the following declaration of ~failure() in ios_base::failure

    namespace std {
       class ios_base::failure : public exception {
       public:
           ...
           virtual ~failure();
           ...
       };
     }

the class failure cannot be implemented since in 18.6.1 [lib.exception] the destructor of class exception has an empty exception specification:

    namespace std {
       class exception {
       public:
         ...
         virtual ~exception() throw();
         ...
       };
     }

Proposed resolution:

Remove the declaration of ~failure().

Rationale:

The proposed resolution is consistent with the way that destructors of other classes derived from exception are handled.


333. does endl imply synchronization with the device?

Section: 27.6.2.7 [lib.ostream.manip]  Status: WP  Submitter: PremAnand M. Rao  Date: 27 Aug 2001

A footnote in 27.6.2.7 [lib.ostream.manip] states:

[Footnote: The effect of executing cout << endl is to insert a newline character in the output sequence controlled by cout, then synchronize it with any external file with which it might be associated. --- end foonote]

Does the term "file" here refer to the external device? This leads to some implementation ambiguity on systems with fully buffered files where a newline does not cause a flush to the device.

Choosing to sync with the device leads to significant performance penalties for each call to endl, while not sync-ing leads to errors under special circumstances.

I could not find any other statement that explicitly defined the behavior one way or the other.

Proposed resolution:

Remove footnote 300 from section 27.6.2.7 [lib.ostream.manip].

Rationale:

We already have normative text saying what endl does: it inserts a newline character and calls flush. This footnote is at best redundant, at worst (as this issue says) misleading, because it appears to make promises about what flush does.


334. map::operator[] specification forces inefficient implementation

Section: 23.3.1.2 [lib.map.access]  Status: WP  Submitter: Andrea Griffini  Date: 02 Sep 2001

The current standard describes map::operator[] using a code example. That code example is however quite inefficient because it requires several useless copies of both the passed key_type value and of default constructed mapped_type instances. My opinion is that was not meant by the comitee to require all those temporary copies.

Currently map::operator[] behaviour is specified as:

  Returns:
    (*((insert(make_pair(x, T()))).first)).second.

This specification however uses make_pair that is a template function of which parameters in this case will be deduced being of type const key_type& and const T&. This will create a pair<key_type,T> that isn't the correct type expected by map::insert so another copy will be required using the template conversion constructor available in pair to build the required pair<const key_type,T> instance.

If we consider calling of key_type copy constructor and mapped_type default constructor and copy constructor as observable behaviour (as I think we should) then the standard is in this place requiring two copies of a key_type element plus a default construction and two copy construction of a mapped_type (supposing the addressed element is already present in the map; otherwise at least another copy construction for each type).

A simple (half) solution would be replacing the description with:

  Returns:
    (*((insert(value_type(x, T()))).first)).second.

This will remove the wrong typed pair construction that requires one extra copy of both key and value.

However still the using of map::insert requires temporary objects while the operation, from a logical point of view, doesn't require any.

I think that a better solution would be leaving free an implementer to use a different approach than map::insert that, because of its interface, forces default constructed temporaries and copies in this case. The best solution in my opinion would be just requiring map::operator[] to return a reference to the mapped_type part of the contained element creating a default element with the specified key if no such an element is already present in the container. Also a logarithmic complexity requirement should be specified for the operation.

This would allow library implementers to write alternative implementations not using map::insert and reaching optimal performance in both cases of the addressed element being present or absent from the map (no temporaries at all and just the creation of a new pair inside the container if the element isn't present). Some implementer has already taken this option but I think that the current wording of the standard rules that as non-conforming.

Proposed resolution:

Replace 23.3.1.2 [lib.map.access] paragraph 1 with

-1- Effects: If there is no key equivalent to x in the map, inserts value_type(x, T()) into the map.

-2- Returns: A reference to the mapped_type corresponding to x in *this.

-3- Complexity: logarithmic.

[This is the second option mentioned above. Howard provided wording. We may also wish to have a blanket statement somewhere in clause 17 saying that we do not intend the semantics of sample code fragments to be interpreted as specifing exactly how many copies are made. See issue 98 for a similar problem.]

Rationale:

This is the second solution described above; as noted, it is consistent with existing practice.

Note that we now need to specify the complexity explicitly, because we are no longer defining operator[] in terms of insert.


335. minor issue with char_traits, table 37

Section: 21.1.1 [lib.char.traits.require]  Status: WP  Submitter: Andy Sawyer  Date: 06 Sep 2001

Table 37, in 21.1.1 [lib.char.traits.require], descibes char_traits::assign as:

  X::assign(c,d)   assigns c = d.

And para 1 says:

[...] c and d denote values of type CharT [...]

Naturally, if c and d are values, then the assignment is (effectively) meaningless. It's clearly intended that (in the case of assign, at least), 'c' is intended to be a reference type.

I did a quick survey of the four implementations I happened to have lying around, and sure enough they all have signatures:

    assign( charT&, const charT& );

(or the equivalent). It's also described this way in Nico's book. (Not to mention the synopses of char_traits<char> in 21.1.3.1 and char_traits<wchar_t> in 21.1.3.2...)

Proposed resolution:

Add the following to 21.1.1 para 1:

r denotes an lvalue of CharT

and change the description of assign in the table to:

  X::assign(r,d)   assigns r = d

336. Clause 17 lack of references to deprecated headers

Section: 17 [lib.library]  Status: WP  Submitter: Detlef Vollmann  Date: 05 Sep 2001

From c++std-edit-873:

17.4.1.2 [lib.headers], Table 11. In this table, the header <strstream> is missing.

This shows a general problem: The whole clause 17 refers quite often to clauses 18 through 27, but D.7 is also a part of the standard library (though a deprecated one).

Proposed resolution:

To 17.4.1.2 [lib.headers] Table 11, C++ Library Headers, add "<strstream>".

In the following places, change "clauses 17 through 27" to "clauses 17 through 27 and Annex D":


337. replace_copy_if's template parameter should be InputIterator

Section: 25.2.4 [lib.alg.replace]  Status: WP  Submitter: Detlef Vollmann  Date: 07 Sep 2001

From c++std-edit-876:

In section 25.2.4 [lib.alg.replace] before p4: The name of the first parameter of template replace_copy_if should be "InputIterator" instead of "Iterator". According to 17.3.2.1 [lib.type.descriptions] p1 the parameter name conveys real normative meaning.

Proposed resolution:

Change Iterator to InputIterator.


338.  is whitespace allowed between `-' and a digit?

Section: 22.2 [lib.locale.categories]  Status: WP  Submitter: Martin Sebor  Date: 17 Sep 2001

From Stage 2 processing in 22.2.2.1.2 [lib.facet.num.get.virtuals], p8 and 9 (the original text or the text corrected by the proposed resolution of issue 221) it seems clear that no whitespace is allowed within a number, but 22.2.3.1 [lib.locale.numpunct], p2, which gives the format for integer and floating point values, says that whitespace is optional between a plusminus and a sign.

The text needs to be clarified to either consistently allow or disallow whitespace between a plusminus and a sign. It might be worthwhile to consider the fact that the C library stdio facility does not permit whitespace embedded in numbers and neither does the C or C++ core language (the syntax of integer-literals is given in 2.13.1 [lex.icon], that of floating-point-literals in 2.13.3 [lex.fcon] of the C++ standard).

Proposed resolution:

Change the first part of 22.2.3.1 [lib.locale.numpunct] paragraph 2 from:

The syntax for number formats is as follows, where digit represents the radix set specified by the fmtflags argument value, whitespace is as determined by the facet ctype<charT> (22.2.1.1), and thousands-sep and decimal-point are the results of corresponding numpunct<charT> members. Integer values have the format:

  integer   ::= [sign] units
  sign      ::= plusminus [whitespace]
  plusminus ::= '+' | '-'
  units     ::= digits [thousands-sep units]
  digits    ::= digit [digits]

to:

The syntax for number formats is as follows, where digit represents the radix set specified by the fmtflags argument value, and thousands-sep and decimal-point are the results of corresponding numpunct<charT> members. Integer values have the format:

  integer   ::= [sign] units
  sign      ::= plusminus
  plusminus ::= '+' | '-'
  units     ::= digits [thousands-sep units]
  digits    ::= digit [digits]

Rationale:

It's not clear whether the format described in 22.2.3.1 [lib.locale.numpunct] paragraph 2 has any normative weight: nothing in the standard says how, or whether, it's used. However, there's no reason for it to differ gratuitously from the very specific description of numeric processing in 22.2.2.1.2 [lib.facet.num.get.virtuals]. The proposed resolution removes all mention of "whitespace" from that format.


339. definition of bitmask type restricted to clause 27

Section: 22.2.1 [lib.category.ctype], 17.3.2.1.2 [lib.bitmask.types]  Status: WP  Submitter: Martin Sebor  Date: 17 September 2001

The ctype_category::mask type is declared to be an enum in 22.2.1 [lib.category.ctype] with p1 then stating that it is a bitmask type, most likely referring to the definition of bitmask type in 17.3.2.1.2 [lib.bitmask.types], p1. However, the said definition only applies to clause 27, making the reference in 22.2.1 somewhat dubious.

Proposed resolution:

Clarify 17.3.2.1.2, p1 by changing the current text from

Several types defined in clause 27 are bitmask types. Each bitmask type can be implemented as an enumerated type that overloads certain operators, as an integer type, or as a bitset (23.3.5 [lib.template.bitset]).

to read

Several types defined in clauses lib.language.support through lib.input.output and Annex D are bitmask types. Each bitmask type can be implemented as an enumerated type that overloads certain operators, as an integer type, or as a bitset (lib.template.bitset).

Additionally, change the definition in 22.2.1 to adopt the same convention as in clause 27 by replacing the existing text with the following (note, in particluar, the cross-reference to 17.3.2.1.2 in 22.2.1, p1):

22.2.1 The ctype category [lib.category.ctype]

namespace std {
    class ctype_base {
    public:
        typedef T mask;

        // numeric values are for exposition only.
        static const mask space = 1 << 0;
        static const mask print = 1 << 1;
        static const mask cntrl = 1 << 2;
        static const mask upper = 1 << 3;
        static const mask lower = 1 << 4;
        static const mask alpha = 1 << 5;
        static const mask digit = 1 << 6;
        static const mask punct = 1 << 7;
        static const mask xdigit = 1 << 8;
        static const mask alnum = alpha | digit;
        static const mask graph = alnum | punct;
    };
}

The type mask is a bitmask type (17.3.2.1.2 [lib.bitmask.types]).

[Curaçao: The LWG notes that T above should be bold-italics to be consistent with the rest of the standard.]


340. interpretation of has_facet<Facet>(loc)

Section: 22.1.1.1.1 [lib.locale.category]  Status: WP  Submitter: Martin Sebor  Date: 18 Sep 2001

It's unclear whether 22.1.1.1.1, p3 says that has_facet<Facet>(loc) returns true for any Facet from Table 51 or whether it includes Table 52 as well:

For any locale loc either constructed, or returned by locale::classic(), and any facet Facet that is a member of a standard category, has_facet<Facet>(loc) is true. Each locale member function which takes a locale::category argument operates on the corresponding set of facets.

It seems that it comes down to which facets are considered to be members of a standard category. Intuitively, I would classify all the facets in Table 52 as members of their respective standard categories, but there are an unbounded set of them...

The paragraph implies that, for instance, has_facet<num_put<C, OutputIterator> >(loc) must always return true. I don't think that's possible. If it were, then use_facet<num_put<C, OutputIterator> >(loc) would have to return a reference to a distinct object for each valid specialization of num_put<C, OutputIteratory>, which is clearly impossible.

On the other hand, if none of the facets in Table 52 is a member of a standard category then none of the locale member functions that operate on entire categories of facets will work properly.

It seems that what p3 should mention that it's required (permitted?) to hold only for specializations of Facet from Table 52 on C from the set { char, wchar_t }, and InputIterator and OutputIterator from the set of { {i,o}streambuf_iterator<{char,wchar_t}> }.

Proposed resolution:

In 22.1.1.1.1 [lib.locale.category], paragraph 3, change "that is a member of a standard category" to "shown in Table 51".

Rationale:

The facets in Table 52 are an unbounded set. Locales should not be required to contain an infinite number of facets.

It's not necessary to talk about which values of InputIterator and OutputIterator must be supported. Table 51 already contains a complete list of the ones we need.


341. Vector reallocation and swap

Section: 23.2.4.2 [lib.vector.capacity]  Status: WP  Submitter: Anthony Williams  Date: 27 Sep 2001

It is a common idiom to reduce the capacity of a vector by swapping it with an empty one:

  std::vector<SomeType> vec;
  // fill vec with data
  std::vector<SomeType>().swap(vec);
  // vec is now empty, with minimal capacity

However, the wording of 23.2.4.2 [lib.vector.capacity]paragraph 5 prevents the capacity of a vector being reduced, following a call to reserve(). This invalidates the idiom, as swap() is thus prevented from reducing the capacity. The proposed wording for issue 329 does not affect this. Consequently, the example above requires the temporary to be expanded to cater for the contents of vec, and the contents be copied across. This is a linear-time operation.

However, the container requirements state that swap must have constant complexity (23.1 [lib.container.requirements] note to table 65).

This is an important issue, as reallocation affects the validity of references and iterators.

If the wording of 23.2.4.2p5 is taken to be the desired intent, then references and iterators remain valid after a call to swap, if they refer to an element before the new end() of the vector into which they originally pointed, in which case they refer to the element at the same index position. Iterators and references that referred to an element whose index position was beyond the new end of the vector are invalidated.

If the note to table 65 is taken as the desired intent, then there are two possibilities with regard to iterators and references:

  1. All Iterators and references into both vectors are invalidated.
  2. Iterators and references into either vector remain valid, and remain pointing to the same element. Consequently iterators and references that referred to one vector now refer to the other, and vice-versa.

Proposed resolution:

Add a new paragraph after 23.2.4.2 [lib.vector.capacity] paragraph 5:

  void swap(vector<T,Allocator>& x);

Effects: Exchanges the contents and capacity() of *this with that of x.

Complexity: Constant time.

[This solves the problem reported for this issue. We may also have a problem with a circular definition of swap() for other containers.]

Rationale:

swap should be constant time. The clear intent is that it should just do pointer twiddling, and that it should exchange all properties of the two vectors, including their reallocation guarantees.


345. type tm in <cwchar>

Section: 21.4 [lib.c.strings]  Status: WP  Submitter: Clark Nelson  Date: 19 Oct 2001

C99, and presumably amendment 1 to C90, specify that <wchar.h> declares struct tm as an incomplete type. However, table 48 in 21.4 [lib.c.strings] does not mention the type tm as being declared in <cwchar>. Is this omission intentional or accidental?

Proposed resolution:

In section 21.4 [lib.c.strings], add "tm" to table 48.


346. Some iterator member functions should be const

Section: 24.1 [lib.iterator.requirements]  Status: WP  Submitter: Jeremy Siek  Date: 20 Oct 2001

Iterator member functions and operators that do not change the state of the iterator should be defined as const member functions or as functions that take iterators either by const reference or by value. The standard does not explicitly state which functions should be const. Since this a fairly common mistake, the following changes are suggested to make this explicit.

The tables almost indicate constness properly through naming: r for non-const and a,b for const iterators. The following changes make this more explicit and also fix a couple problems.

Proposed resolution:

In 24.1 [lib.iterator.requirements] Change the first section of p9 from "In the following sections, a and b denote values of X..." to "In the following sections, a and b denote values of type const X...".

In Table 73, change

    a->m   U&         ...

to

    a->m   const U&   ...
    r->m   U&         ...

In Table 73 expression column, change

    *a = t

to

    *r = t

[Redmond: The container requirements should be reviewed to see if the same problem appears there.]


347. locale::category and bitmask requirements

Section: 22.1.1.1.1 [lib.locale.category]  Status: WP  Submitter: P.J. Plauger, Nathan Myers  Date: 23 Oct 2001

In 22.1.1.1.1 [lib.locale.category] paragraph 1, the category members are described as bitmask elements. In fact, the bitmask requirements in 17.3.2.1.2 [lib.bitmask.types] don't seem quite right: none and all are bitmask constants, not bitmask elements.

In particular, the requirements for none interact poorly with the requirement that the LC_* constants from the C library must be recognizable as C++ locale category constants. LC_* values should not be mixed with these values to make category values.

We have two options for the proposed resolution. Informally: option 1 removes the requirement that LC_* values be recognized as category arguments. Option 2 changes the category type so that this requirement is implementable, by allowing none to be some value such as 0x1000 instead of 0.

Nathan writes: "I believe my proposed resolution [Option 2] merely re-expresses the status quo more clearly, without introducing any changes beyond resolving the DR.

Proposed resolution:

Replace the first two paragraphs of 22.1.1.1 [lib.locale.types] with:

    typedef int category;

Valid category values include the locale member bitmask elements collate, ctype, monetary, numeric, time, and messages, each of which represents a single locale category. In addition, locale member bitmask constant none is defined as zero and represents no category. And locale member bitmask constant all is defined such that the expression

    (collate | ctype | monetary | numeric | time | messages | all) == all

is true, and represents the union of all categories. Further the expression (X | Y), where X and Y each represent a single category, represents the union of the two categories.

locale member functions expecting a category argument require one of the category values defined above, or the union of two or more such values. Such a category argument identifies a set of locale categories. Each locale category, in turn, identifies a set of locale facets, including at least those shown in Table 51:

[Curaçao: need input from locale experts.]

Rationale:

The LWG considered, and rejected, an alternate proposal (described as "Option 2" in the discussion). The main reason for rejecting it was that library implementors were concerened about implementation difficult, given that getting a C++ library to work smoothly with a separately written C library is already a delicate business. Some library implementers were also concerned about the issue of adding extra locale categories.

Option 2:
Replace the first paragraph of 22.1.1.1 [lib.locale.types] with:

Valid category values include the enumerated values. In addition, the result of applying commutative operators | and & to any two valid values is valid, and results in the setwise union and intersection, respectively, of the argument categories. The values all and none are defined such that for any valid value cat, the expressions (cat | all == all), (cat & all == cat), (cat | none == cat) and (cat & none == none) are true. For non-equal values cat1 and cat2 of the remaining enumerated values, (cat1 & cat2 == none) is true. For any valid categories cat1 and cat2, the result of (cat1 & ~cat2) is valid, and equals the setwise union of those categories found in cat1 but not found in cat2. [Footnote: it is not required that all equal the setwise union of the other enumerated values; implementations may add extra categories.]


349. Minor typographical error in ostream_iterator

Section: 24.5.2 [lib.ostream.iterator]  Status: WP  Submitter: Andy Sawyer  Date: 24 Oct 2001

24.5.2 [lib.ostream.iterator] states:

    [...]

    private:
    // basic_ostream<charT,traits>* out_stream; exposition only
    // const char* delim; exposition only

Whilst it's clearly marked "exposition only", I suspect 'delim' should be of type 'const charT*'.

Proposed resolution:

In 24.5.2 [lib.ostream.iterator], replace const char* delim with const charT* delim.


352. missing fpos requirements

Section: 21.1.2 [lib.char.traits.typedefs]  Status: WP  Submitter: Martin Sebor  Date: 2 Dec 2001

(1) There are no requirements on the stateT template parameter of fpos listed in 27.4.3. The interface appears to require that the type be at least Assignable and CopyConstructible (27.4.3.1, p1), and I think also DefaultConstructible (to implement the operations in Table 88).

21.1.2, p3, however, only requires that char_traits<charT>::state_type meet the requirements of CopyConstructible types.

(2) Additionally, the stateT template argument has no corresponding typedef in fpos which might make it difficult to use in generic code.

Proposed resolution:

Modify 21.1.2, p4 from

Requires: state_type shall meet the requirements of CopyConstructible types (20.1.3).

Requires: state_type shall meet the requirements of Assignable (23.1, p4), CopyConstructible (20.1.3), and DefaultConstructible (20.1.4) types.

Rationale:

The LWG feels this is two issues, as indicated above. The first is a defect---std::basic_fstream is unimplementable without these additional requirements---and the proposed resolution fixes it. The second is questionable; who would use that typedef? The class template fpos is used only in a very few places, all of which know the state type already. Unless motivation is provided, the second should be considered NAD.


354. Associative container lower/upper bound requirements

Section: 23.1.2 [lib.associative.reqmts]  Status: WP  Submitter: Hans Aberg  Date: 17 Dec 2001

Discussions in the thread "Associative container lower/upper bound requirements" on comp.std.c++ suggests that there is a defect in the C++ standard, Table 69 of section 23.1.2, "Associative containers", [lib.associative.reqmts]. It currently says:

a.find(k): returns an iterator pointing to an element with the key equivalent to k, or a.end() if such an element is not found.

a.lower_bound(k): returns an iterator pointing to the first element with key not less than k.

a.upper_bound(k): returns an iterator pointing to the first element with key greater than k.

We have "or a.end() if such an element is not found" for find, but not for upper_bound or lower_bound. As the text stands, one would be forced to insert a new element into the container and return an iterator to that in case the sought iterator does not exist, which does not seem to be the intention (and not possible with the "const" versions).

Proposed resolution:

Change Table 69 of section 23.1.2 [lib.associative.reqmts] indicated entries to:

a.lower_bound(k): returns an iterator pointing to the first element with key not less than k, or a.end() if such an element is not found.

a.upper_bound(k): returns an iterator pointing to the first element with key greater than k, or a.end() if such an element is not found.

[Curaçao: LWG reviewed PR.]


355. Operational semantics for a.back()

Section: 23.1.1 [lib.sequence.reqmts]  Status: WP  Submitter: Yaroslav Mironov  Date: 23 Jan 2002

Table 68 "Optional Sequence Operations" in 23.1.1/12 specifies operational semantics for "a.back()" as "*--a.end()", which may be ill-formed [because calling operator-- on a temporary (the return) of a built-in type is ill-formed], provided a.end() returns a simple pointer rvalue (this is almost always the case for std::vector::end(), for example). Thus, the specification is not only incorrect, it demonstrates a dangerous construct: "--a.end()" may successfully compile and run as intended, but after changing the type of the container or the mode of compilation it may produce compile-time error.

Proposed resolution:

Change the specification in table 68 "Optional Sequence Operations" in 23.1.1/12 for "a.back()" from

*--a.end()

to

{ iterator tmp = a.end(); --tmp; return *tmp; }

and the specification for "a.pop_back()" from

a.erase(--a.end())

to

{ iterator tmp = a.end(); --tmp; a.erase(tmp); }

[Curaçao: LWG changed PR from "{ X::iterator tmp = a.end(); return *--tmp; }" to "*a.rbegin()", and from "{ X::iterator tmp = a.end(); a.erase(--tmp); }" to "a.erase(rbegin())".]

[There is a second possible defect; table 68 "Optional Sequence Operations" in the "Operational Semantics" column uses operations present only in the "Reversible Container" requirements, yet there is no stated dependency between these separate requirements tables. Ask in Santa Cruz if the LWG would like a new issue opened.]

[Santa Cruz: the proposed resolution is even worse than what's in the current standard: erase is undefined for reverse iterator. If we're going to make the change, we need to define a temporary and use operator--. Additionally, we don't know how prevalent this is: do we need to make this change in more than one place? Martin has volunteered to review the standard and see if this problem occurs elsewhere.]

[Oxford: Matt provided new wording to address the concerns raised in Santa Cruz. It does not appear that this problem appears anywhere else in clauses 23 or 24.]

[Kona: In definition of operational semantics of back(), change "*tmp" to "return *tmp;"]


358. interpreting thousands_sep after a decimal_point

Section: 22.2.2.1.2 [lib.facet.num.get.virtuals]  Status: WP  Submitter: Martin Sebor  Date: 12 Mar 2002

I don't think thousands_sep is being treated correctly after decimal_point has been seen. Since grouping applies only to the integral part of the number, the first such occurrence should, IMO, terminate Stage 2. (If it does not terminate it, then 22.2.2.1.2, p12 and 22.2.3.1.2, p3 need to explain how thousands_sep is to be interpreted in the fractional part of a number.)

The easiest change I can think of that resolves this issue would be something like below.

Proposed resolution:

Change the first sentence of 22.2.2.1.2, p9 from

If discard is true then the position of the character is remembered, but the character is otherwise ignored. If it is not discarded, then a check is made to determine if c is allowed as the next character of an input field of the conversion specifier returned by stage 1. If so it is accumulated.

to

If discard is true, then if '.' has not yet been accumulated, then the position of the character is remembered, but the character is otherwise ignored. Otherwise, if '.' has already been accumulated, the character is discarded and Stage 2 terminates. ...

Rationale:

We believe this reflects the intent of the Standard. Thousands sep characters after the decimal point are not useful in any locale. Some formatting conventions do group digits that follow the decimal point, but they usually introduce a different grouping character instead of reusing the thousand sep character. If we want to add support for such conventions, we need to do so explicitly.


359. num_put<>::do_put (..., bool) undocumented

Section: 22.2.2.2.1 [lib.facet.num.put.members]  Status: WP  Submitter: Martin Sebor  Date: 12 Mar 2002

22.2.2.2.1, p1:

    iter_type put (iter_type out, ios_base& str, char_type fill,
                   bool val) const;
    ...

    1   Returns: do_put (out, str, fill, val).
    

AFAICS, the behavior of do_put (..., bool) is not documented anywhere, however, 22.2.2.2.2, p23:

iter_type put (iter_type out, ios_base& str, char_type fill,
               bool val) const;
Effects: If (str.flags() & ios_base::boolalpha) == 0 then do out = do_put(out, str, fill, (int)val) Otherwise do
             string_type s =
                 val ? use_facet<ctype<charT> >(loc).truename()
                     : use_facet<ctype<charT> >(loc).falsename();
and then insert the characters of s into out. out.

This means that the bool overload of do_put() will never be called, which contradicts the first paragraph. Perhaps the declaration should read do_put(), and not put()?

Note also that there is no Returns clause for this function, which should probably be corrected, just as should the second occurrence of "out." in the text.

I think the least invasive change to fix it would be something like the following:

Proposed resolution:

In 22.2.2.2.2 [lib.facet.num.put.virtuals], just above paragraph 1, remove the bool overload.

In 22.2.2.2.2 [lib.facet.num.put.virtuals], p23, make the following changes

Replace put() with do_put() in the declaration of the member function.
Change the Effects clause to a Returns clause (to avoid the requirement to call do_put(..., int) from do_put (..., bool)) like so:
23 Returns: If (str.flags() & ios_base::boolalpha) == 0 then do_put (out, str, fill, (long)val) Otherwise the function obtains a string s as if by
             string_type s =
                val ? use_facet<ctype<charT> >(loc).truename()
                    : use_facet<ctype<charT> >(loc).falsename();
and then inserts each character c of s into out via *out++ = c and returns out.

Rationale:

This fixes a couple of obvious typos, and also fixes what appears to be a requirement of gratuitous inefficiency.


360. locale mandates inefficient implementation

Section: 22.1.1 [lib.locale]  Status: WP  Submitter: Martin Sebor  Date: 12 Mar 2002

22.1.1, p7 (copied below) allows iostream formatters and extractors to make assumptions about the values returned from facet members. However, such assumptions are apparently not guaranteed to hold in other cases (e.g., when the facet members are being called directly rather than as a result of iostream calls, or between successive calls to the same iostream functions with no interevening calls to imbue(), or even when the facet member functions are called from other member functions of other facets). This restriction prevents locale from being implemented efficiently.

Proposed resolution:

Change the first sentence in 22.1.1, p7 from

In successive calls to a locale facet member function during a call to an iostream inserter or extractor or a streambuf member function, the returned result shall be identical. [Note: This implies that such results may safely be reused without calling the locale facet member function again, and that member functions of iostream classes cannot safely call imbue() themselves, except as specified elsewhere. --end note]

to

In successive calls to a locale facet member function on a facet object installed in the same locale, the returned result shall be identical. ...

Rationale:

This change is reasonable becuase it clarifies the intent of this part of the standard.


363. Missing exception specification in 27.4.2.1.1

Section: 27.4.2.1.1 [lib.ios::failure]  Status: WP  Submitter: Walter Brown and Marc Paterno  Date: 20 May 2002

The destructor of ios_base::failure should have an empty throw specification, because the destructor of its base class, exception, is declared in this way.

Proposed resolution:

Change the destructor to

  virtual ~failure() throw();

Rationale:

Fixes an obvious glitch. This is almost editorial.


364. Inconsistent wording in 27.5.2.4.2

Section: 27.5.2.4.2 [lib.streambuf.virt.buffer]  Status: WP  Submitter: Walter Brown, Marc Paterno  Date: 10 May 2002

27.5.2.4.2 [lib.streambuf.virt.buffer] paragraph 1 is inconsistent with the Effects clause for seekoff.

Proposed resolution:

Make this paragraph, the Effects clause for setbuf, consistent in wording with the Effects clause for seekoff in paragraph 3 by amending paragraph 1 to indicate the purpose of setbuf:

Original text:

1 Effects: Performs an operation that is defined separately for each class derived from basic_streambuf in this clause (27.7.1.3, 27.8.1.4).

Proposed text:

1 Effects: Influences stream buffering in a way that is defined separately for each class derived from basic_streambuf in this clause (27.7.1.3, 27.8.1.4).

Rationale:

The LWG doesn't believe there is any normative difference between the existing wording and what's in the proposed resolution, but the change may make the intent clearer.


365. Lack of const-qualification in clause 27

Section: 27 [lib.input.output]  Status: WP  Submitter: Walter Brown, Marc Paterno  Date: 10 May 2002

Some stream and streambuf member functions are declared non-const, even thought they appear only to report information rather than to change an object's logical state. They should be declared const. See document N1360 for details and rationale.

The list of member functions under discussion: in_avail, showmanyc, tellg, tellp, is_open.

Related issue: 73

Proposed resolution:

In 27.8.1.5, 27.8.1.7, 27.8.1.8, 27.8.1.10, 27.8.1.11, and 27.8.1.13

Replace

  bool is_open();

with

  bool is_open() const;

Rationale:

Of the changes proposed in N1360, the only one that is safe is changing the filestreams' is_open to const. The LWG believed that this was NAD the first time it considered this issue (issue 73), but now thinks otherwise. The corresponding streambuf member function, after all,is already const.

The other proposed changes are less safe, because some streambuf functions that appear merely to report a value do actually perform mutating operations. It's not even clear that they should be considered "logically const", because streambuf has two interfaces, a public one and a protected one. These functions may, and often do, change the state as exposed by the protected interface, even if the state exposed by the public interface is unchanged.

Note that implementers can make this change in a binary compatible way by providing both overloads; this would be a conforming extension.


370. Minor error in basic_istream::get

Section: 27.6.1.3 [lib.istream.unformatted]  Status: WP  Submitter: Ray Lischner  Date: 15 Jul 2002

Defect report for description of basic_istream::get (section 27.6.1.3 [lib.istream.unformatted]), paragraph 15. The description for the get function with the following signature:

  basic_istream<charT,traits>& get(basic_streambuf<char_type,traits>&
  sb);

is incorrect. It reads

Effects: Calls get(s,n,widen('\n'))

which I believe should be:

Effects: Calls get(sb,widen('\n'))

Proposed resolution:

Change the Effects paragraph to:

Effects: Calls get(sb,this->widen('\n'))

[Pre-Oxford: Minor correction from Howard: replaced 'widen' with 'this->widen'.]

Rationale:

Fixes an obvious typo.


373. Are basic_istream and basic_ostream to use (exceptions()&badbit) != 0 ?

Section: 27.6.1.2.1 [lib.istream.formatted.reqmts], 27.6.2.5.1 [lib.ostream.formatted.reqmts]  Status: WP  Submitter: Keith Baker  Date: 23 Jul 2002

In 27.6.1.2.1 [lib.istream.formatted.reqmts] and 27.6.2.5.1 [lib.ostream.formatted.reqmts] (exception()&badbit) != 0 is used in testing for rethrow, yet exception() is the constructor to class std::exception in 18.6.1 [lib.exception] that has no return type. Should member function exceptions() found in 27.4.4 [lib.ios] be used instead?

Proposed resolution:

In 27.6.1.2.1 [lib.istream.formatted.reqmts] and 27.6.2.5.1 [lib.ostream.formatted.reqmts], change "(exception()&badbit) != 0" to "(exceptions()&badbit) != 0".

Rationale:

Fixes an obvious typo.


375. basic_ios should be ios_base in 27.7.1.3

Section: 27.7.1.3 [lib.stringbuf.virtuals]  Status: WP  Submitter: Ray Lischner  Date: 14 Aug 2002

In Section 27.7.1.3 [lib.stringbuf.virtuals]: Table 90, Table 91, and paragraph 14 all contain references to "basic_ios::" which should be "ios_base::".

Proposed resolution:

Change all references to "basic_ios" in Table 90, Table 91, and paragraph 14 to "ios_base".

Rationale:

Fixes an obvious typo.


379. nonsensical ctype::do_widen() requirement

Section: 22.2.1.1.2 [lib.locale.ctype.virtuals]  Status: WP  Submitter: Martin Sebor  Date: 6 Sep 2002

The last sentence in 22.2.1.1.2, p11 below doesn't seem to make sense.

  charT do_widen (char c) const;

  -11- Effects: Applies the simplest reasonable transformation from
       a char value or sequence of char values to the corresponding
       charT value or values. The only characters for which unique
       transformations are required are those in the basic source
       character set (2.2). For any named ctype category with a
       ctype<charT> facet ctw and valid ctype_base::mask value
       M (is(M, c) || !ctw.is(M, do_widen(c))) is true.

Shouldn't the last sentence instead read

       For any named ctype category with a ctype<char> facet ctc
       and valid ctype_base::mask value M
       (ctc.is(M, c) || !is(M, do_widen(c))) is true.

I.e., if the narrow character c is not a member of a class of characters then neither is the widened form of c. (To paraphrase footnote 224.)

Proposed resolution:

Replace the last sentence of 22.2.1.1.2 [lib.locale.ctype.virtuals], p11 with the following text:

       For any named ctype category with a ctype<char> facet ctc
       and valid ctype_base::mask value M,
       (ctc.is(M, c) || !is(M, do_widen(c))) is true.

[Kona: Minor edit. Added a comma after the M for clarity.]

Rationale:

The LWG believes this is just a typo, and that this is the correct fix.


380. typos in codecvt tables 53 and 54

Section: 22.2.1.5.2 [lib.locale.codecvt.virtuals]  Status: WP  Submitter: Martin Sebor  Date: 6 Sep 2002

Tables 53 and 54 in 22.2.1.5.2 [lib.locale.codecvt.virtuals] are both titled "convert result values," when surely "do_in/do_out result values" must have been intended for Table 53 and "do_unshift result values" for Table 54.

Table 54, row 3 says that the meaning of partial is "more characters needed to be supplied to complete termination." The function is not supplied any characters, it is given a buffer which it fills with characters or, more precisely, destination elements (i.e., an escape sequence). So partial means that space for more than (to_limit - to) destination elements was needed to terminate a sequence given the value of state.

Proposed resolution:

Change the title of Table 53 to "do_in/do_out result values" and the title of Table 54 to "do_unshift result values."

Change the text in Table 54, row 3 (the partial row), under the heading Meaning, to "space for more than (to_limit - to) destination elements was needed to terminate a sequence given the value of state."


381. detection of invalid mbstate_t in codecvt

Section: 22.2.1.5.2 [lib.locale.codecvt.virtuals]  Status: WP  Submitter: Martin Sebor  Date: 6 Sep 2002

All but one codecvt member functions that take a state_type argument list as one of their preconditions that the state_type argument have a valid value. However, according to 22.2.1.5.2, p6, codecvt::do_unshift() is the only codecvt member that is supposed to return error if the state_type object is invalid.

It seems to me that the treatment of state_type by all codecvt member functions should be the same and the current requirements should be changed. Since the detection of invalid state_type values may be difficult in general or computationally expensive in some specific cases, I propose the following:

Proposed resolution:

Add a new paragraph before 22.2.1.5.2, p5, and after the function declaration below

    result do_unshift(stateT& state,
    externT* to, externT* to_limit, externT*& to_next) const;

as follows:

    Requires: (to <= to_end) well defined and true; state initialized,
    if at the beginning of a sequence, or else equal to the result of
    converting the preceding characters in the sequence.

and change the text in Table 54, row 4, the error row, under the heading Meaning, from

    state has invalid value

to

    an unspecified error has occurred

Rationale:

The intent is that implementations should not be required to detect invalid state values; such a requirement appears nowhere else. An invalid state value is a precondition violation, i.e. undefined behavior. Implementations that do choose to detect invalid state values, or that choose to detect any other kind of error, may return error as an indication.


383. Bidirectional iterator assertion typo

Section: 24.1.4 [lib.bidirectional.iterators]  Status: WP  Submitter: ysapir (submitted via comp.std.c++)  Date: 17 Oct 2002

Following a discussion on the boost list regarding end iterators and the possibility of performing operator--() on them, it seems to me that there is a typo in the standard. This typo has nothing to do with that discussion.

I have checked this newsgroup, as well as attempted a search of the Active/Defect/Closed Issues List on the site for the words "s is derefer" so I believe this has not been proposed before. Furthermore, the "Lists by Index" mentions only DR 299 on section 24.1.4, and DR 299 is not related to this issue.

The standard makes the following assertion on bidirectional iterators, in section 24.1.4 [lib.bidirectional.iterators], Table 75:

                         operational  assertion/note
expression  return type   semantics    pre/post-condition

--r          X&                        pre: there exists s such
                                       that r == ++s.
                                       post: s is dereferenceable.
                                       --(++r) == r.
                                       --r == --s implies r == s.
                                       &r == &--r.

(See http://aspn.activestate.com/ASPN/Mail/Message/boost/1395763.)

In particular, "s is dereferenceable" seems to be in error. It seems that the intention was to say "r is dereferenceable".

If it were to say "r is dereferenceable" it would make perfect sense. Since s must be dereferenceable prior to operator++, then the natural result of operator-- (to undo operator++) would be to make r dereferenceable. Furthermore, without other assertions, and basing only on precondition and postconditions, we could not otherwise know this. So it is also interesting information.

Proposed resolution:

Change the guarantee to "postcondition: r is dereferenceable."

Rationale:

Fixes an obvious typo


386. Reverse iterator's operator[] has impossible return type

Section: 24.4.1.3.11 [lib.reverse.iter.opindex]  Status: DR  Submitter: Matt Austern  Date: 23 Oct 2002

In 24.4.1.3.11 [lib.reverse.iter.opindex], reverse_iterator<>::operator[] is specified as having a return type of reverse_iterator::reference, which is the same as iterator_traits<Iterator>::reference. (Where Iterator is the underlying iterator type.)

The trouble is that Iterator's own operator[] doesn't necessarily have a return type of iterator_traits<Iterator>::reference. Its return type is merely required to be convertible to Iterator's value type. The return type specified for reverse_iterator's operator[] would thus appear to be impossible.

With the resolution of issue 299, the type of a[n] will continue to be required (for random access iterators) to be convertible to the value type, and also a[n] = t will be a valid expression. Implementations of reverse_iterator will likely need to return a proxy from operator[] to meet these requirements. As mentioned in the comment from Dave Abrahams, the simplest way to specify that reverse_iterator meet this requirement to just mandate it and leave the return type of operator[] unspecified.

Proposed resolution:

In 24.4.1.2 [lib.reverse.iter.requirements] change:

reference operator[](difference_type n) const;

to:

unspecified operator[](difference_type n) const; // see lib.random.access.iterators

[ Comments from Dave Abrahams: IMO we should resolve 386 by just saying that the return type of reverse_iterator's operator[] is unspecified, allowing the random access iterator requirements to impose an appropriate return type. If we accept 299's proposed resolution (and I think we should), the return type will be readable and writable, which is about as good as we can do. ]


389. Const overload of valarray::operator[] returns by value

Section: 26.3.2 [lib.template.valarray]  Status: WP  Submitter: Gabriel Dos Reis  Date: 8 Nov 2002

Consider the following program:

    #include <iostream>
    #include <ostream>
    #include <vector>
    #include <valarray>
    #include <algorithm>
    #include <iterator>
    template<typename Array>
    void print(const Array& a)
    {
    using namespace std;
    typedef typename Array::value_type T;
    copy(&a[0], &a[0] + a.size(),
    ostream_iterator<T>(std::cout, " "));
    }
    template<typename T, unsigned N>
    unsigned size(T(&)[N]) { return N; }
    int main()
    {
    double array[] = { 0.89, 9.3, 7, 6.23 };
    std::vector<double> v(array, array + size(array));
    std::valarray<double> w(array, size(array));
    print(v); // #1
    std::cout << std::endl;
    print(w); // #2
    std::cout << std::endl;
    }

While the call numbered #1 succeeds, the call numbered #2 fails because the const version of the member function valarray<T>::operator[](size_t) returns a value instead of a const-reference. That seems to be so for no apparent reason, no benefit. Not only does that defeats users' expectation but it also does hinder existing software (written either in C or Fortran) integration within programs written in C++. There is no reason why subscripting an expression of type valarray<T> that is const-qualified should not return a const T&.

Proposed resolution:

In the class synopsis in 26.3.2 [lib.template.valarray], and in 26.3.2.3 [lib.valarray.access] just above paragraph 1, change

  T operator[](size_t const);

to

  const T& operator[](size_t const);

[Kona: fixed a minor typo: put semicolon at the end of the line wehre it belongs.]

Rationale:

Return by value seems to serve no purpose. Valaray was explicitly designed to have a specified layout so that it could easily be integrated with libraries in other languages, and return by value defeats that purpose. It is believed that this change will have no impact on allowable optimizations.


391. non-member functions specified as const

Section: 22.1.3.2 [lib.conversions]  Status: WP  Submitter: James Kanze  Date: 10 Dec 2002

The specifications of toupper and tolower both specify the functions as const, althought they are not member functions, and are not specified as const in the header file synopsis in section 22.1 [lib.locales].

Proposed resolution:

In 22.1.3.2 [lib.conversions], remove const from the function declarations of std::toupper and std::tolower

Rationale:

Fixes an obvious typo


395. inconsistencies in the definitions of rand() and random_shuffle()

Section: 26.5 [lib.c.math]  Status: WP  Submitter: James Kanze  Date: 3 Jan 2003

In 26.5 [lib.c.math], the C++ standard refers to the C standard for the definition of rand(); in the C standard, it is written that "The implementation shall behave as if no library function calls the rand function."

In 25.2.11 [lib.alg.random.shuffle], there is no specification as to how the two parameter version of the function generates its random value. I believe that all current implementations in fact call rand() (in contradiction with the requirement avove); if an implementation does not call rand(), there is the question of how whatever random generator it does use is seeded. Something is missing.

Proposed resolution:

In [lib.c.math], add a paragraph specifying that the C definition of rand shal be modified to say that "Unless otherwise specified, the implementation shall behave as if no library function calls the rand function."

In [lib.alg.random.shuffle], add a sentence to the effect that "In the two argument form of the function, the underlying source of random numbers is implementation defined. [Note: in particular, an implementation is permitted to use rand.]

Rationale:

The original proposed resolution proposed requiring the two-argument from of random_shuffle to use rand. We don't want to do that, because some existing implementations already use something else: gcc uses lrand48, for example. Using rand presents a problem if the number of elements in the sequence is greater than RAND_MAX.


400. redundant type cast in lib.allocator.members

Section: 20.4.1.1 [lib.allocator.members]  Status: WP  Submitter: Markus Mauhart  Date: 27 Feb 2003

20.4.1.1 [lib.allocator.members] allocator members, contains the following 3 lines:

  12 Returns: new((void *) p) T( val)
     void destroy(pointer p);
  13 Returns: ((T*) p)->~T()

The type cast "(T*) p" in the last line is redundant cause we know that std::allocator<T>::pointer is a typedef for T*.

Proposed resolution:

Replace "((T*) p)" with "p".

Rationale:

Just a typo, this is really editorial.


402. wrong new expression in [some_]allocator::construct

Section: 20.1.5 [lib.allocator.requirements], 20.4.1.1 [lib.allocator.members],   Status: WP  Submitter: Markus Mauhart  Date: 27 Feb 2003

This applies to the new expression that is contained in both par12 of 20.4.1.1 [lib.allocator.members] and in par2 (table 32) of 20.1.5 [lib.allocator.requirements]. I think this new expression is wrong, involving unintended side effects.

20.4.1.1 [lib.allocator.members] contains the following 3 lines:

  11 Returns: the largest value N for which the call allocate(N,0) might succeed.
     void construct(pointer p, const_reference val);
  12 Returns: new((void *) p) T( val)

20.1.5 [lib.allocator.requirements] in table 32 has the following line:

  a.construct(p,t)   Effect: new((void*)p) T(t)

.... with the prerequisits coming from the preceding two paragraphs, especially from table 31:

  alloc<T>             a     ;// an allocator for T
  alloc<T>::pointer    p     ;// random access iterator
                              // (may be different from T*)
  alloc<T>::reference  r = *p;// T&
  T const&             t     ;

Cause of using "new" but not "::new", any existing "T::operator new" function will hide the global placement new function. When there is no "T::operator new" with adequate signature, every_alloc<T>::construct(..) is ill-formed, and most std::container<T,every_alloc<T>> use it; a workaround would be adding placement new and delete functions with adequate signature and semantic to class T, but class T might come from another party. Maybe even worse is the case when T has placement new and delete functions with adequate signature but with "unknown" semantic: I dont like to speculate about it, but whoever implements any_container<T,any_alloc> and wants to use construct(..) probably must think about it.

Proposed resolution:

Replace "new" with "::new" in both cases.


403. basic_string::swap should not throw exceptions

Section: 21.3.5.8 [lib.string::swap]  Status: WP  Submitter: Beman Dawes  Date: 25 Mar 2003

std::basic_string, 21.3 [lib.basic.string] paragraph 2 says that basic_string "conforms to the requirements of a Sequence, as specified in (23.1.1)." The sequence requirements specified in (23.1.1) to not include any prohibition on swap members throwing exceptions.

Section 23.1 [lib.container.requirements] paragraph 10 does limit conditions under which exceptions may be thrown, but applies only to "all container types defined in this clause" and so excludes basic_string::swap because it is defined elsewhere.

Eric Niebler points out that 21.3 [lib.basic.string] paragraph 5 explicitly permits basic_string::swap to invalidates iterators, which is disallowed by 23.1 [lib.container.requirements] paragraph 10. Thus the standard would be contradictory if it were read or extended to read as having basic_string meet 23.1 [lib.container.requirements] paragraph 10 requirements.

Yet several LWG members have expressed the belief that the original intent was that basic_string::swap should not throw exceptions as specified by 23.1 [lib.container.requirements] paragraph 10, and that the standard is unclear on this issue. The complexity of basic_string::swap is specified as "constant time", indicating the intent was to avoid copying (which could cause a bad_alloc or other exception). An important use of swap is to ensure that exceptions are not thrown in exception-safe code.

Note: There remains long standing concern over whether or not it is possible to reasonably meet the 23.1 [lib.container.requirements] paragraph 10 swap requirements when allocators are unequal. The specification of basic_string::swap exception requirements is in no way intended to address, prejudice, or otherwise impact that concern.

Proposed resolution:

In 21.3.5.8 [lib.string::swap], add a throws clause:

Throws: Shall not throw exceptions.


404. May a replacement allocation function be declared inline?

Section: 17.4.3.4 [lib.replacement.functions], 18.4.1 [lib.new.delete]  Status: WP  Submitter: Matt Austern  Date: 24 Apr 2003

The eight basic dynamic memory allocation functions (single-object and array versions of ::operator new and ::operator delete, in the ordinary and nothrow forms) are replaceable. A C++ program may provide an alternative definition for any of them, which will be used in preference to the implementation's definition.

Three different parts of the standard mention requirements on replacement functions: 17.4.3.4 [lib.replacement.functions], 18.4.1.1 [lib.new.delete.single] and 18.4.1.2 [lib.new.delete.array], and 3.7.3 [basic.stc.dynamic].

None of these three places say whether a replacement function may be declared inline. 18.4.1.1 [lib.new.delete.single] paragraph 2 specifies a signature for the replacement function, but that's not enough: the inline specifier is not part of a function's signature. One might also reason from 7.1.2 [dcl.fct.spec] paragraph 2, which requires that "an inline function shall be defined in every translation unit in which it is used," but this may not be quite specific enough either. We should either explicitly allow or explicitly forbid inline replacement memory allocation functions.

Proposed resolution:

Add a new sentence to the end of 17.4.3.4 [lib.replacement.functions] paragraph 3: "The program's definitions shall not be specified as inline. No diagnostic is required."

[Kona: added "no diagnostic is required"]

Rationale:

The fact that inline isn't mentioned appears to have been nothing more than an oversight. Existing implementations do not permit inline functions as replacement memory allocation functions. Providing this functionality would be difficult in some cases, and is believed to be of limited value.


405. qsort and POD

Section: 25.4 [lib.alg.c.library]  Status: WP  Submitter: Ray Lischner  Date: 08 Apr 2003

Section 25.4 [lib.alg.c.library] describes bsearch and qsort, from the C standard library. Paragraph 4 does not list any restrictions on qsort, but it should limit the base parameter to point to POD. Presumably, qsort sorts the array by copying bytes, which requires POD.

Proposed resolution:

In 25.4 [lib.alg.c.library] paragraph 4, just after the declarations and before the nonnormative note, add these words: "both of which have the same behavior as the original declaration. The behavior is undefined unless the objects in the array pointed to by base are of POD type."

[Something along these lines is clearly necessary. Matt provided wording.]


406. vector::insert(s) exception safety

Section: 23.2.4.3 [lib.vector.modifiers]  Status: DR  Submitter: Dave Abrahams  Date: 27 Apr 2003

There is a possible defect in the standard: the standard text was never intended to prevent arbitrary ForwardIterators, whose operations may throw exceptions, from being passed, and it also wasn't intended to require a temporary buffer in the case where ForwardIterators were passed (and I think most implementations don't use one). As is, the standard appears to impose requirements that aren't met by any existing implementation.

Proposed resolution:

Replace 23.2.4.3 [lib.vector.modifiers] paragraph 1 with:

1- Notes: Causes reallocation if the new size is greater than the old capacity. If no reallocation happens, all the iterators and references before the insertion point remain valid. If an exception is thrown other than by the copy constructor or assignment operator of T or by any InputIterator operation there are no effects.

[We probably need to say something similar for deque.]


407. Can singular iterators be destroyed?

Section: 24.1 [lib.iterator.requirements]  Status: WP  Submitter: Nathan Myers  Date: 3 June 2003

Clause 24.1 [lib.iterator.requirements], paragraph 5, says that the only expression that is defined for a singular iterator is "an assignment of a non-singular value to an iterator that holds a singular value". This means that destroying a singular iterator (e.g. letting an automatic variable go out of scope) is technically undefined behavior. This seems overly strict, and probably unintentional.

Proposed resolution:

Change the sentence in question to "... the only exceptions are destroying an iterator that holds a singular value, or the assignment of a non-singular value to an iterator that holds a singular value."


409. Closing an fstream should clear error state

Section: 27.8.1.7 [lib.ifstream.members], 27.8.1.10 [lib.ofstream.members]  Status: DR  Submitter: Nathan Myers  Date: 3 June 2003

A strict reading of 27.8.1 [lib.fstreams] shows that opening or closing a basic_[io]fstream does not affect the error bits. This means, for example, that if you read through a file up to EOF, and then close the stream and reopen it at the beginning of the file, the EOF bit in the stream's error state is still set. This is counterintuitive.

The LWG considered this issue once before, as issue 22, and put in a footnote to clarify that the strict reading was indeed correct. We did that because we believed the standard was unambiguous and consistent, and that we should not make architectural changes in a TC. Now that we're working on a new revision of the language, those considerations no longer apply.

Proposed resolution:

Change 27.8.1.7 [lib.ifstream.members], para. 3 from:

Calls rdbuf()->open(s,mode|in). If that function returns a null pointer, calls setstate(failbit) (which may throw ios_base::failure [Footnote: (lib.iostate.flags)].

to:

Calls rdbuf()->open(s,mode|in). If that function returns a null pointer, calls setstate(failbit) (which may throw ios_base::failure [Footnote: (lib.iostate.flags)), else calls clear().

Change 27.8.1.10 [lib.ofstream.members], para. 3 from:

Calls rdbuf()->open(s,mode|out). If that function returns a null pointer, calls setstate(failbit) (which may throw ios_base::failure [Footnote: (lib.iostate.flags)).

to:

Calls rdbuf()->open(s,mode|out). If that function returns a null pointer, calls setstate(failbit) (which may throw ios_base::failure [Footnote: (lib.iostate.flags)), else calls clear().

Change 27.8.1.13 [lib.fstream.members], para. 3 from:

Calls rdbuf()->open(s,mode), If that function returns a null pointer, calls setstate(failbit), (which may throw ios_base::failure). (lib.iostate.flags) )

to:

Calls rdbuf()->open(s,mode), If that function returns a null pointer, calls setstate(failbit), (which may throw ios_base::failure). (lib.iostate.flags) ), else calls clear().

[Kona: the LWG agrees this is a good idea. Post-Kona: Bill provided wording. He suggests having open, not close, clear the error flags.]

[Post-Sydney: Howard provided a new proposed resolution. The old one didn't make sense because it proposed to fix this at the level of basic_filebuf, which doesn't have access to the stream's error state. Howard's proposed resolution fixes this at the level of the three fstream class template instead.]


410. Missing semantics for stack and queue comparison operators

Section: 23.2.3.1 [lib.queue], 23.2.3.3 [lib.stack]  Status: WP  Submitter: Hans Bos  Date: 7 Jun 2003

Sections 23.2.3.1 [lib.queue] and 23.2.3.3 [lib.stack] list comparison operators (==, !=, <, <=, >, =>) for queue and stack. Only the semantics for queue::operator== (23.2.3.1 [lib.queue] par2) and queue::operator< (23.2.3.1 [lib.queue] par3) are defined.

Proposed resolution:

Add the following new paragraphs after 23.2.3.1 [lib.queue] paragraph 3:

  operator!=

Returns: x.c != y.c

  operator>

Returns: x.c > y.c

  operator<=

Returns: x.c <= y.c

  operator>=

Returns: x.c >= y.c

Add the following paragraphs at the end of 23.2.3.3 [lib.stack]:

  operator==

Returns: x.c == y.c

  operator<

Returns: x.c < y.c

  operator!=

Returns: x.c != y.c

  operator>

Returns: x.c > y.c

  operator<=

Returns: x.c <= y.c

  operator>=

Returns: x.c >= y.c

[Kona: Matt provided wording.]

Rationale:

There isn't any real doubt about what these operators are supposed to do, but we ought to spell it out.

411. Wrong names of set member functions

Section: 25.3.5 [lib.alg.set.operations]  Status: WP  Submitter: Daniel Frey  Date: 9 Jul 2003

25.3.5 [lib.alg.set.operations] paragraph 1 reads: "The semantics of the set operations are generalized to multisets in a standard way by defining union() to contain the maximum number of occurrences of every element, intersection() to contain the minimum, and so on."

This is wrong. The name of the functions are set_union() and set_intersection(), not union() and intersection().

Proposed resolution:

Change that sentence to use the correct names.


412. Typo in 27.4.4.3

Section: 27.4.4.3 [lib.iostate.flags]  Status: WP  Submitter: Martin Sebor  Date: 10 Jul 2003

The Effects clause in 27.4.4.3 [lib.iostate.flags] paragraph 5 says that the function only throws if the respective bits are already set prior to the function call. That's obviously not the intent. The typo ought to be corrected and the text reworded as: "If (state & exceptions()) == 0, returns. ..."

Proposed resolution:

In 27.4.4.3 [lib.iostate.flags] paragraph 5, replace "If (rdstate() & exceptions()) == 0" with "If ((state | (rdbuf() ? goodbit : badbit)) & exceptions()) == 0".

[Kona: the original proposed resolution wasn't quite right. We really do mean rdstate(); the ambiguity is that the wording in the standard doesn't make it clear whether we mean rdstate() before setting the new state, or rdsate() after setting it. We intend the latter, of course. Post-Kona: Martin provided wording.]


413. Proposed resolution to LDR#64 still wrong

Section: 27.6.1.2.3 [lib.istream::extractors]  Status: DR  Submitter: Bo Persson  Date: 13 Jul 2003

The second sentence of the proposed resolution says:

"If it inserted no characters because it caught an exception thrown while extracting characters from sb and ..."

However, we are not extracting from sb, but extracting from the basic_istream (*this) and inserting into sb. I can't really tell if "extracting" or "sb" is a typo.

[ Sydney: Definitely a real issue. We are, indeed, extracting characters from an istream and not from sb. The problem was there in the FDIS and wasn't fixed by issue 64. Probably what was intended was to have *this instead of sb. We're talking about the exception flag state of a basic_istream object, and there's only one basic_istream object in this discussion, so that would be a consistent interpretation. (But we need to be careful: the exception policy of this member function must be consistent with that of other extractors.) PJP will provide wording. ]

Proposed resolution:

Change the sentence from:

If it inserted no characters because it caught an exception thrown while extracting characters from sb and failbit is on in exceptions(), then the caught exception is rethrown.

to:

If it inserted no characters because it caught an exception thrown while extracting characters from *this and failbit is on in exceptions(), then the caught exception is rethrown.

414. Which iterators are invalidated by v.erase()?

Section: 23.2.4.3 [lib.vector.modifiers]  Status: WP  Submitter: Matt Austern  Date: 19 Aug 2003

Consider the following code fragment:

int A[8] = { 1,3,5,7,9,8,4,2 };
std::vector<int> v(A, A+8);

std::vector<int>::iterator i1 = v.begin() + 3;
std::vector<int>::iterator i2 = v.begin() + 4;
v.erase(i1);

Which iterators are invalidated by v.erase(i1): i1, i2, both, or neither?

On all existing implementations that I know of, the status of i1 and i2 is the same: both of them will be iterators that point to some elements of the vector (albeit not the same elements they did before). You won't get a crash if you use them. Depending on exactly what you mean by "invalidate", you might say that neither one has been invalidated because they still point to something, or you might say that both have been invalidated because in both cases the elements they point to have been changed out from under the iterator.

The standard doesn't say either of those things. It says that erase invalidates all iterators and references "after the point of the erase". This doesn't include i1, since it's at the point of the erase instead of after it. I can't think of any sensible definition of invalidation by which one can say that i2 is invalidated but i1 isn't.

(This issue is important if you try to reason about iterator validity based only on the guarantees in the standard, rather than reasoning from typical implementation techniques. Strict debugging modes, which some programmers find useful, do not use typical implementation techniques.)

Proposed resolution:

In 23.2.4.3 [lib.vector.modifiers] paragraph 3, change "Invalidates all the iterators and references after the point of the erase" to "Invalidates iterators and references at or after the point of the erase".

Rationale:

I believe this was essentially a typographical error, and that it was taken for granted that erasing an element invalidates iterators that point to it. The effects clause in question treats iterators and references in parallel, and it would seem counterintuitive to say that a reference to an erased value remains valid.


415. behavior of std::ws

Section: 27.6.1.4 [lib.istream.manip]  Status: WP  Submitter: Martin Sebor  Date: 18 Sep 2003

According to 27.6.1.4, the ws() manipulator is not required to construct the sentry object. The manipulator is also not a member function so the text in 27.6.1, p1 through 4 that describes the exception policy for istream member functions does not apply. That seems inconsistent with the rest of extractors and all the other input functions (i.e., ws will not cause a tied stream to be flushed before extraction, it doesn't check the stream's exceptions or catch exceptions thrown during input, and it doesn't affect the stream's gcount).

Proposed resolution:

Add to 27.6.1.4 [lib.istream.manip], immediately before the first sentence of paragraph 1, the following text:

Behaves as an unformatted input function (as described in 27.6.1.3, paragraph 1), except that it does not count the number of characters extracted and does not affect the value returned by subsequent calls to is.gcount(). After constructing a sentry object...

[Post-Kona: Martin provided wording]


420. is std::FILE a complete type?

Section: 27.8.1 [lib.fstreams]  Status: WP  Submitter: Martin Sebor  Date: 18 Sep 2003

7.19.1, p2, of C99 requires that the FILE type only be declared in <stdio.h>. None of the (implementation-defined) members of the struct is mentioned anywhere for obvious reasons.

C++ says in 27.8.1, p2 that FILE is a type that's defined in <cstdio>. Is it really the intent that FILE be a complete type or is an implementation allowed to just declare it without providing a full definition?

Proposed resolution:

In the first sentence of 27.8.1 [lib.fstreams] paragraph 2, change "defined" to "declared".

Rationale:

We don't want to impose any restrictions beyond what the C standard already says. We don't want to make anything implementation defined, because that imposes new requirements in implementations.


425. return value of std::get_temporary_buffer

Section: 20.4.3 [lib.temporary.buffer]  Status: WP  Submitter: Martin Sebor  Date: 18 Sep 2003

The standard is not clear about the requirements on the value returned from a call to get_temporary_buffer(0). In particular, it fails to specify whether the call should return a distinct pointer each time it is called (like operator new), or whether the value is unspecified (as if returned by malloc). The standard also fails to mention what the required behavior is when the argument is less than 0.

Proposed resolution:

Change 20.4.3 [lib.temporary.buffer] paragraph 2 from "...or a pair of 0 values if no storage can be obtained" to "...or a pair of 0 values if no storage can be obtained or if n <= 0."

[Kona: Matt provided wording]


426. search_n(), fill_n(), and generate_n() with negative n

Section: 25.1.9 [lib.alg.search], 25.2.5 [lib.alg.fill], 25.2.6 [lib.alg.generate]  Status: WP  Submitter: Martin Sebor  Date: 18 Sep 2003

The complexity requirements for these function templates are incorrect (or don't even make sense) for negative n:

25.1.9, p7 (search_n):
Complexity: At most (last1 - first1) * count applications of the corresponding predicate.

25.2.5, p3 (fill_n):
Complexity: Exactly last - first (or n) assignments.


25.2.6, p3 (generate_n):
Complexity: Exactly last - first (or n) assignments.

In addition, the Requirements or the Effects clauses for the latter two templates don't say anything about the behavior when n is negative.

Proposed resolution:

Change 25.1.9, p7 to

Complexity: At most (last1 - first1) * count applications of the corresponding predicate if count is positive, or 0 otherwise.

Change 25.2.5, p2 to

Effects: Assigns value through all the iterators in the range [first, last), or [first, first + n) if n is positive, none otherwise.

Change 25.2.5, p3 to:

Complexity: Exactly last - first (or n if n is positive, or 0 otherwise) assignments.

Change 25.2.6, p1 to (notice the correction for the misspelled "through"):

Effects: Invokes the function object genand assigns the return value of gen through all the iterators in the range [first, last), or [first, first + n) if n is positive, or [first, first) otherwise.

Change 25.2.6, p3 to:

Complexity: Exactly last - first (or n if n is positive, or 0 otherwise) assignments.

Rationale:

Informally, we want to say that whenever we see a negative number we treat it the same as if it were zero. We believe the above changes do that (although they may not be the minimal way of saying so). The LWG considered and rejected the alternative of saying that negative numbers are undefined behavior.


428. string::erase(iterator) validity

Section: 21.3.5.5 [lib.string::erase]  Status: WP  Submitter: Martin Sebor  Date: 18 Sep 2003

23.1.1, p3 along with Table 67 specify as a prerequisite for a.erase(q) that q must be a valid dereferenceable iterator into the sequence a.

However, 21.3.5.5, p5 describing string::erase(p) only requires that p be a valid iterator.

This may be interepreted as a relaxation of the general requirement, which is most likely not the intent.

Proposed resolution:

Remove 21.3.5.5 [lib.string::erase] paragraph 5.

Rationale:

The LWG considered two options: changing the string requirements to match the general container requirements, or just removing the erroneous string requirements altogether. The LWG chose the latter option, on the grounds that duplicating text always risks the possibility that it might be duplicated incorrectly.


432. stringbuf::overflow() makes only one write position available

Section: 27.7.1.3 [lib.stringbuf.virtuals]  Status: WP  Submitter: Christian W Brock  Date: 24 Sep 2003

27.7.1.3 par 8 says:

Notes: The function can make a write position available only if ( mode & ios_base::out) != 0. To make a write position available, the function reallocates (or initially allocates) an array object with a sufficient number of elements to hold the current array object (if any), plus one additional write position. If ( mode & ios_base::in) != 0, the function alters the read end pointer egptr() to point just past the new write position (as does the write end pointer epptr()).

The sentences "plus one additional write position." and especially "(as does the write end pointer epptr())" COULD by interpreted (and is interpreted by at least my library vendor) as:

post-condition: epptr() == pptr()+1

This WOULD force sputc() to call the virtual overflow() each time.

The proposed change also affects Defect Report 169.

Proposed resolution:

27.7.1.1/2 Change:

2- Notes: The function allocates no array object.

to:

2- Postcondition: str() == "".

27.7.1.1/3 Change:

-3- Effects: Constructs an object of class basic_stringbuf, initializing the base class with basic_streambuf() (lib.streambuf.cons), and initializing mode with which . Then copies the content of str into the basic_stringbuf underlying character sequence and initializes the input and output sequences according to which. If which & ios_base::out is true, initializes the output sequence with the underlying sequence. If which & ios_base::in is true, initializes the input sequence with the underlying sequence.

to:

-3- Effects: Constructs an object of class basic_stringbuf, initializing the base class with basic_streambuf() (lib.streambuf.cons), and initializing mode with which. Then copies the content of str into the basic_stringbuf underlying character sequence. If which & ios_base::out is true, initializes the output sequence such that pbase() points to the first underlying character, epptr() points one past the last underlying character, and if (which & ios_base::ate) is true, pptr() is set equal to epptr() else pptr() is set equal to pbase(). If which & ios_base::in is true, initializes the input sequence such that eback() and gptr() point to the first underlying character and egptr() points one past the last underlying character.

27.7.1.2/1 Change:

-1- Returns: A basic_string object whose content is equal to the basic_stringbuf underlying character sequence. If the buffer is only created in input mode, the underlying character sequence is equal to the input sequence; otherwise, it is equal to the output sequence. In case of an empty underlying character sequence, the function returns basic_string<charT,traits,Allocator>().

to:

-1- Returns: A basic_string object whose content is equal to the basic_stringbuf underlying character sequence. If the basic_stringbuf was created only in input mode, the resultant basic_string contains the character sequence in the range [eback(), egptr()). If the basic_stringbuf was created with (which & ios_base::out) being true then the resultant basic_string contains the character sequence in the range [pbase(), high_mark) where high_mark represents the position one past the highest initialized character in the buffer. Characters can be initialized either through writing to the stream, or by constructing the basic_stringbuf with a basic_string, or by calling the str(basic_string) member function. In the case of calling the str(basic_string) member function, all characters initialized prior to the call are now considered uninitialized (except for those characters re-initialized by the new basic_string). Otherwise the basic_stringbuf has been created in neither input nor output mode and a zero length basic_string is returned.

27.7.1.2/2 Change:

-2- Effects: If the basic_stringbuf's underlying character sequence is not empty, deallocates it. Then copies the content of s into the basic_stringbuf underlying character sequence and initializes the input and output sequences according to the mode stored when creating the basic_stringbuf object. If (mode&ios_base::out) is true, then initializes the output sequence with the underlying sequence. If (mode&ios_base::in) is true, then initializes the input sequence with the underlying sequence.

to:

-2- Effects: Copies the content of s into the basic_stringbuf underlying character sequence. If mode & ios_base::out is true, initializes the output sequence such that pbase() points to the first underlying character, epptr() points one past the last underlying character, and if (mode & ios_base::ate) is true, pptr() is set equal to epptr() else pptr() is set equal to pbase(). If mode & ios_base::in is true, initializes the input sequence such that eback() and gptr() point to the first underlying character and egptr() points one past the last underlying character.

Remove 27.2.1.2/3. (Same rationale as issue 238: incorrect and unnecessary.)

27.7.1.3/1 Change:

1- Returns: If the input sequence has a read position available, returns traits::to_int_type(*gptr()). Otherwise, returns traits::eof().

to:

1- Returns: If the input sequence has a read position available, returns traits::to_int_type(*gptr()). Otherwise, returns traits::eof(). Any character in the underlying buffer which has been initialized is considered to be part of the input sequence.

27.7.1.3/9 Change:

-9- Notes: The function can make a write position available only if ( mode & ios_base::out) != 0. To make a write position available, the function reallocates (or initially allocates) an array object with a sufficient number of elements to hold the current array object (if any), plus one additional write position. If ( mode & ios_base::in) != 0, the function alters the read end pointer egptr() to point just past the new write position (as does the write end pointer epptr()).

to:

-9- The function can make a write position available only if ( mode & ios_base::out) != 0. To make a write position available, the function reallocates (or initially allocates) an array object with a sufficient number of elements to hold the current array object (if any), plus one additional write position. If ( mode & ios_base::in) != 0, the function alters the read end pointer egptr() to point just past the new write position.

27.7.1.3/12 Change:

-12- _ If (newoff + off) < 0, or (xend - xbeg) < (newoff + off), the positioning operation fails. Otherwise, the function assigns xbeg + newoff + off to the next pointer xnext .

to:

-12- _ If (newoff + off) < 0, or if (newoff + off) refers to an uninitialized character (as defined in 27.7.1.2 [lib.stringbuf.members] paragraph 1), the positioning operation fails. Otherwise, the function assigns xbeg + newoff + off to the next pointer xnext .

[post-Kona: Howard provided wording. At Kona the LWG agreed that something along these lines was a good idea, but the original proposed resolution didn't say enough about the effect of various member functions on the underlying character sequences.]

Rationale:

The current basic_stringbuf description is over-constrained in such a way as to prohibit vendors from making this the high-performance in-memory stream it was meant to be. The fundamental problem is that the pointers: eback(), gptr(), egptr(), pbase(), pptr(), epptr() are observable from a derived client, and the current description restricts the range [pbase(), epptr()) from being grown geometrically. This change allows, but does not require, geometric growth of this range.

Backwards compatibility issues: These changes will break code that derives from basic_stringbuf, observes epptr(), and depends upon [pbase(), epptr()) growing by one character on each call to overflow() (i.e. test suites). Otherwise there are no backwards compatibility issues.

27.7.1.1/2: The non-normative note is non-binding, and if it were binding, would be over specification. The recommended change focuses on the important observable fact.

27.7.1.1/3: This change does two things: 1. It describes exactly what must happen in terms of the sequences. The terms "input sequence" and "output sequence" are not well defined. 2. It introduces a common extension: open with app or ate mode. I concur with issue 238 that paragraph 4 is both wrong and unnecessary.

27.7.1.2/1: This change is the crux of the efficiency issue. The resultant basic_string is not dependent upon epptr(), and thus implementors are free to grow the underlying buffer geometrically during overflow() *and* place epptr() at the end of that buffer.

27.7.1.2/2: Made consistent with the proposed 27.7.1.1/3.

27.7.1.3/1: Clarifies that characters written to the stream beyond the initially specified string are available for reading in an i/o basic_streambuf.

27.7.1.3/9: Made normative by removing "Notes:", and removed the trailing parenthetical comment concerning epptr().

27.7.1.3/12: Restricting the positioning to [xbeg, xend) is no longer allowable since [pbase(), epptr()) may now contain uninitialized characters. Positioning is only allowable over the initialized range.


434. bitset::to_string() hard to use

Section: 23.3.5.2 [lib.bitset.members]  Status: DR  Submitter: Martin Sebor  Date: 15 Oct 2003

It has been pointed out a number of times that the bitset to_string() member function template is tedious to use since callers must explicitly specify the entire template argument list (3 arguments). At least two implementations provide a number of overloads of this template to make it easier to use.

Proposed resolution:

In order to allow callers to specify no template arguments at all, just the first one (charT), or the first 2 (charT and traits), in addition to all three template arguments, add the following three overloads to both the interface (declarations only) of the class template bitset as well as to section 23.3.5.2, immediately after p34, the Returns clause of the existing to_string() member function template:

    template <class charT, class traits>
    basic_string<charT, traits, allocator<charT> >
    to_string () const;

    -34.1- Returns: to_string<charT, traits, allocator<charT> >().

    template <class charT>
    basic_string<charT, char_traits<charT>, allocator<charT> >
    to_string () const;

    -34.2- Returns: to_string<charT, char_traits<charT>, allocator<charT> >().

    basic_string<char, char_traits<char>, allocator<char> >
    to_string () const;

    -34.3- Returns: to_string<char, char_traits<char>, allocator<char> >().

[Kona: the LWG agrees that this is an improvement over the status quo. Dietmar thought about an alternative using a proxy object but now believes that the proposed resolution above is the right choice. ]


435. bug in DR 25

Section: 21.3.7.9 [lib.string.io]  Status: WP  Submitter: Martin Sebor  Date: 15 Oct 2003

It has been pointed out that the proposed resolution in DR 25 may not be quite up to snuff:
http://gcc.gnu.org/ml/libstdc++/2003-09/msg00147.html http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/lwg-defects.html#25

It looks like Petur is right. The complete corrected text is copied below. I think we may have have been confused by the reference to 22.2.2.2.2 and the subsequent description of `n' which actually talks about the second argument to sputn(), not about the number of fill characters to pad with.

So the question is: was the original text correct? If the intent was to follow classic iostreams then it most likely wasn't, since setting width() to less than the length of the string doesn't truncate it on output. This is also the behavior of most implementations (except for SGI's standard iostreams where the operator does truncate).

Proposed resolution:

Change the text in 21.3.7.9, p4 from

If bool(k) is true, inserts characters as if by calling os.rdbuf()->sputn(str.data(), n), padding as described in stage 3 of lib.facet.num.put.virtuals, where n is the larger of os.width() and str.size();

to

If bool(k) is true, determines padding as described in lib.facet.num.put.virtuals, and then inserts the resulting sequence of characters seq as if by calling os.rdbuf()->sputn(seq, n), where n is the larger of os.width() and str.size();

[Kona: it appears that neither the original wording, DR25, nor the proposed resolution, is quite what we want. We want to say that the string will be output, padded to os.width() if necessary. We don't want to duplicate the padding rules in clause 22, because they're complicated, but we need to be careful because they weren't quite written with quite this case in mind. We need to say what the character sequence is, and then defer to clause 22. Post-Kona: Benjamin provided wording.]


436. are cv-qualified facet types valid facets?

Section: 22.1.1.1.2 [lib.locale.facet]  Status: WP  Submitter: Martin Sebor  Date: 15 Oct 2003

Is "const std::ctype<char>" a valid template argument to has_facet, use_facet, and the locale template ctor? And if so, does it designate the same Facet as the non-const "std::ctype<char>?" What about "volatile std::ctype<char>?" Different implementations behave differently: some fail to compile, others accept such types but behave inconsistently.

Proposed resolution:

Change 22.1.1.1.2, p1 to read:

Template parameters in this clause which are required to be facets are those named Facet in declarations. A program that passes a type that is not a facet, or a type that refers to volatile-qualified facet, as an (explicit or deduced) template parameter to a locale function expecting a facet, is ill-formed. A const-qualified facet is a valid template argument to any locale function that expects a Facet template parameter.

[Kona: changed the last sentence from a footnote to normative text.]


438. Ambiguity in the "do the right thing" clause

Section: 23.1.1 [lib.sequence.reqmts]  Status: DR  Submitter: Howard Hinnant  Date: 20 Oct 2003

Section 23.1.1 [lib.sequence.reqmts], paragraphs 9-11, fixed up the problem noticed with statements like:

vector<int> v(10, 1);

The intent of the above statement was to construct with:

vector(size_type, const value_type&);

but early implementations failed to compile as they bound to:

template <class InputIterator>
vector(InputIterator f, InputIterator l);

instead.

Paragraphs 9-11 say that if InputIterator is an integral type, then the member template constructor will have the same effect as:

vector<static_cast<size_type>(f), static_cast<value_type>(l));

(and similarly for the other member template functions of sequences).

There is also a note that describes one implementation technique:

One way that sequence implementors can satisfy this requirement is to specialize the member template for every integral type.

This might look something like:

template <class T>
struct vector
{
     typedef unsigned size_type;

     explicit vector(size_type) {}
     vector(size_type, const T&) {}

     template <class I>
     vector(I, I);

     // ...
};

template <class T>
template <class I>
vector<T>::vector(I, I) { ... }

template <>
template <>
vector<int>::vector(int, int) { ... }

template <>
template <>
vector<int>::vector(unsigned, unsigned) { ... }

//  ...

Label this solution 'A'.

The standard also says:

Less cumbersome implementation techniques also exist.

A popular technique is to not specialize as above, but instead catch every call with the member template, detect the type of InputIterator, and then redirect to the correct logic. Something like:

template <class T>
template <class I>
vector<T>::vector(I f, I l)
{
     choose_init(f, l, int2type<is_integral<I>::value>());
}

template <class T>
template <class I>
vector<T>::choose_init(I f, I l, int2type<false>)
{
    // construct with iterators
}

template <class T>
template <class I>
vector<T>::choose_init(I f, I l, int2type<true>)
{
    size_type sz = static_cast<size_type>(f);
    value_type v = static_cast<value_type>(l);
    // construct with sz,v
}

Label this solution 'B'.

Both of these solutions solve the case the standard specifically mentions:

vector<int> v(10, 1);  // ok, vector size 10, initialized to 1

However, (and here is the problem), the two solutions have different behavior in some cases where the value_type of the sequence is not an integral type. For example consider:

     pair<char, char>                     p('a', 'b');
     vector<vector<pair<char, char> > >   d('a', 'b');

The second line of this snippet is likely an error. Solution A catches the error and refuses to compile. The reason is that there is no specialization of the member template constructor that looks like:

template <>
template <>
vector<vector<pair<char, char> > >::vector(char, char) { ... }

So the expression binds to the unspecialized member template constructor, and then fails (compile time) because char is not an InputIterator.

Solution B compiles the above example though. 'a' is casted to an unsigned integral type and used to size the outer vector. 'b' is static casted to the inner vector using it's explicit constructor:

explicit vector(size_type n);

and so you end up with a static_cast<size_type>('a') by static_cast<size_type>('b') matrix.

It is certainly possible that this is what the coder intended. But the explicit qualifier on the inner vector has been thwarted at any rate.

The standard is not clear whether the expression:

     vector<vector<pair<char, char> > >   d('a', 'b');

(and similar expressions) are:

  1. undefined behavior.
  2. illegal and must be rejected.
  3. legal and must be accepted.

My preference is listed in the order presented.

There are still other techniques for implementing the requirements of paragraphs 9-11, namely the "restricted template technique" (e.g. enable_if). This technique is the most compact and easy way of coding the requirements, and has the behavior of #2 (rejects the above expression).

Choosing 1 would allow all implementation techniques I'm aware of. Choosing 2 would allow only solution 'A' and the enable_if technique. Choosing 3 would allow only solution 'B'.

Possible wording for a future standard if we wanted to actively reject the expression above would be to change "static_cast" in paragraphs 9-11 to "implicit_cast" where that is defined by:

template <class T, class U>
inline
T implicit_cast(const U& u)
{
     return u;
}

Proposed resolution:

Replace 23.1.1 [lib.sequence.reqmts] paragraphs 9 - 11 with:

For every sequence defined in this clause and in clause lib.strings:

In the previous paragraph the alternative binding will fail if f is not implicitly convertible to X::size_type or if l is not implicitly convertible to X::value_type.

The extent to which an implementation determines that a type cannot be an input iterator is unspecified, except that as a minimum integral types shall not qualify as input iterators.

[ Kona: agreed that the current standard requires v('a', 'b') to be accepted, and also agreed that this is surprising behavior. The LWG considered several options, including something like implicit_cast, which doesn't appear to be quite what we want. We considered Howards three options: allow acceptance or rejection, require rejection as a compile time error, and require acceptance. By straw poll (1-6-1), we chose to require a compile time error. Post-Kona: Howard provided wording. ]

[ Sydney: The LWG agreed with this general direction, but there was some discomfort with the wording in the original proposed resolution. Howard submitted new wording, and we will review this again in Redmond. ]

[Redmond: one very small change in wording: the first argument is cast to size_t. This fixes the problem of something like vector<vector<int> >(5, 5), where int is not implicitly convertible to the value type.]

Rationale:

The proposed resolution fixes:

  vector<int> v(10, 1);

since as integral types 10 and 1 must be disqualified as input iterators and therefore the (size,value) constructor is called (as if).

The proposed resolution breaks:

  vector<vector<T> > v(10, 1);

because the integral type 1 is not *implicitly* convertible to vector<T>. The wording above requires a diagnostic.

The proposed resolution leaves the behavior of the following code unspecified.

  struct A
  {
    operator int () const {return 10;}
  };

  struct B
  {
    B(A) {}
  };

  vector<B> v(A(), A());

The implementation may or may not detect that A is not an input iterator and employee the (size,value) constructor. Note though that in the above example if the B(A) constructor is qualified explicit, then the implementation must reject the constructor as A is no longer implicitly convertible to B.


441. Is fpos::state const?

Section: 27.4.3 [lib.fpos]  Status: WP  Submitter: Vincent Leloup  Date: 17 Nov 2003

In section 27.4.3.1 [lib.fpos.members] fpos<stateT>::state() is declared non const, but in section 27.4.3 [lib.fpos] it is declared const.

Proposed resolution:

In section 27.4.3.1 [lib.fpos.members], change the declaration of fpos<stateT>::state() to const.


442. sentry::operator bool() inconsistent signature

Section: 27.6.2.3 [lib.ostream::sentry]  Status: WP  Submitter: Vincent Leloup  Date: 18 Nov 2003

In section 27.6.2.3 [lib.ostream::sentry] paragraph 4, in description part basic_ostream<charT, traits>::sentry::operator bool() is declared as non const, but in section 27.6.2.3, in synopsis it is declared const.

Proposed resolution:

In section 27.6.2.3 [lib.ostream::sentry] paragraph 4, change the declaration of sentry::operator bool() to const.


443. filebuf::close() inconsistent use of EOF

Section: 27.8.1.3 [lib.filebuf.members]  Status: WP  Submitter: Vincent Leloup  Date: 20 Nov 2003

In section 27.8.1.3 [lib.filebuf.members] par6, in effects description of basic_filebuf<charT, traits>::close(), overflow(EOF) is used twice; should be overflow(traits::eof()).

Proposed resolution:

Change overflow(EOF) to overflow(traits::eof()).


444. Bad use of casts in fstream

Section: 27.8.1 [lib.fstreams]  Status: DR  Submitter: Vincent Leloup  Date: 20 Nov 2003

27.8.1.7 [lib.ifstream.members] p1, 27.8.1.10 [lib.ofstream.members] p1, 27.8.1.13 [lib.fstream.members] p1 seems have same problem as exposed in LWG issue 252.

Proposed resolution:

[Sydney: Genuine defect. 27.8.1.13 needs a cast to cast away constness. The other two places are stylistic: we could change the C-style casts to const_cast. Post-Sydney: Howard provided wording. ]

Change 27.8.1.7/1 from:

Returns: (basic_filebuf<charT,traits>*)&sb.

to:

Returns: const_cast<basic_filebuf<charT,traits>*>(&sb).

Change 27.8.1.10/1 from:

Returns: (basic_filebuf<charT,traits>*)&sb.

to:

Returns: const_cast<basic_filebuf<charT,traits>*>(&sb).

Change 27.8.1.13/1 from:

Returns: &sb.

to:

Returns: const_cast<basic_filebuf<charT,traits>*>(&sb).

445. iterator_traits::reference unspecified for some iterator categories

Section: 24.3.1 [lib.iterator.traits]  Status: DR  Submitter: Dave Abrahams  Date: 9 Dec 2003

The standard places no restrictions at all on the reference type of input, output, or forward iterators (for forward iterators it only specifies that *x must be value_type& and doesn't mention the reference type). Bidirectional iterators' reference type is restricted only by implication, since the base iterator's reference type is used as the return type of reverse_iterator's operator*, which must be T& in order to be a conforming forward iterator.

Here's what I think we ought to be able to expect from an input or forward iterator's reference type R, where a is an iterator and V is its value_type

A mutable forward iterator ought to satisfy, for x of type V:

  • { R r = *a; r = x; } is equivalent to *a = x;
  • I think these requirements capture existing container iterators (including vector<bool>'s), but render istream_iterator invalid; its reference type would have to be changed to a constant reference.

    (Jeremy Siek) During the discussion in Sydney, it was felt that a simpler long term solution for this was needed. The solution proposed was to require reference to be the same type as *a and pointer to be the same type as a->. Most iterators in the Standard Library already meet this requirement. Some iterators are output iterators, and do not need to meet the requirement, and others are only specified through the general iterator requirements (which will change with this resolution). The sole case where there is an explicit definition of the reference type that will need to change is istreambuf_iterator which returns charT from operator* but has a reference type of charT&. We propose changing the reference type of istreambuf_iterator to charT.

    The other option for resolving the issue with pointer, mentioned in the note below, is to remove pointer altogether. I prefer placing requirements on pointer to removing it for two reasons. First, pointer will become useful for implementing iterator adaptors and in particular, reverse_iterator will become more well defined. Second, removing pointer is a rather drastic and publicly-visible action to take.

    The proposed resolution technically enlarges the requirements for iterators, which means there are existing iterators (such as istreambuf_iterator, and potentially some programmer-defined iterators) that will no longer meet the requirements. Will this break existing code? The scenario in which it would is if an algorithm implementation (say in the Standard Library) is changed to rely on iterator_traits::reference, and then is used with one of the iterators that do not have an appropriately defined iterator_traits::reference.

    The proposed resolution makes one other subtle change. Previously, it was required that output iterators have a difference_type and value_type of void, which means that a forward iterator could not be an output iterator. This is clearly a mistake, so I've changed the wording to say that those types may be void.

    Proposed resolution:

    In 24.3.1 [lib.iterator.traits], after:

    be defined as the iterator's difference type, value type and iterator category, respectively.

    add

    In addition, the types
    iterator_traits<Iterator>::reference
    iterator_traits<Iterator>::pointer
    
    must be defined as the iterator's reference and pointer types, that is, the same type as the type of *a and a->, respectively.

    In 24.3.1 [lib.iterator.traits], change:

    In the case of an output iterator, the types
    iterator_traits<Iterator>::difference_type
    iterator_traits<Iterator>::value_type
    
    are both defined as void.

    to:

    In the case of an output iterator, the types
    iterator_traits<Iterator>::difference_type
    iterator_traits<Iterator>::value_type
    iterator_traits<Iterator>::reference
    iterator_traits<Iterator>::pointer
    
    may be defined as void.

    In 24.5.3 [lib.istreambuf.iterator], change:

    typename traits::off_type, charT*, charT&>
    

    to:

    typename traits::off_type, charT*, charT>
    

    [ Redmond: there was concern in Sydney that this might not be the only place where things were underspecified and needed to be changed. Jeremy reviewed iterators in the standard and confirmed that nothing else needed to be changed. ]


    448. Random Access Iterators over abstract classes

    Section: 24.1.5 [lib.random.access.iterators]  Status: WP  Submitter: Dave Abrahams  Date: 7 Jan 2004

    Table 76, the random access iterator requirement table, says that the return type of a[n] must be "convertible to T". When an iterator's value_type T is an abstract class, nothing is convertible to T. Surely this isn't an intended restriction?

    Proposed resolution:

    Change the return type to "convertible to T const&".


    449. Library Issue 306 Goes Too Far

    Section: 18.1 [lib.support.types]  Status: WP  Submitter: Pete Becker  Date: 15 Jan 2004

    Original text:

    The macro offsetof accepts a restricted set of type arguments in this International Standard. type shall be a POD structure or a POD union (clause 9). The result of applying the offsetof macro to a field that is a static data member or a function member is undefined."

    Revised text:

    "If type is not a POD structure or a POD union the results are undefined."

    Looks to me like the revised text should have replaced only the second sentence. It doesn't make sense standing alone.

    Proposed resolution:

    Change 18.1, paragraph 5, to:

    The macro offsetof accepts a restricted set of type arguments in this International Standard. If type is not a POD structure or a POD union the results are undefined. The result of applying the offsetof macro to a field that is a static data member or a function member is undefined."

    453. basic_stringbuf::seekoff need not always fail for an empty stream

    Section: 27.7.1.3 [lib.stringbuf.virtuals]  Status: DR  Submitter: Bill Plauger  Date: 30 Jan 2004

      pos_type basic_stringbuf::seekoff(off_type, ios_base::seekdir,
                                        ios_base::openmode);
    

    is obliged to fail if nothing has been inserted into the stream. This is unnecessary and undesirable. It should be permissible to seek to an effective offset of zero.

    [ Sydney: Agreed that this is an annoying problem: seeking to zero should be legal. Bill will provide wording. ]

    Proposed resolution:

    Change the sentence from:

    For a sequence to be positioned, if its next pointer (either gptr() or pptr()) is a null pointer, the positioning operation fails.

    to:

    For a sequence to be positioned, if its next pointer (either gptr() or pptr()) is a null pointer and the new offset newoff is nonzero, the positioning operation fails.

    455. cerr::tie() and wcerr::tie() are overspecified

    Section: 27.3 [lib.iostream.objects]  Status: DR  Submitter: Bill Plauger  Date: 30 Jan 2004

    Both cerr::tie() and wcerr::tie() are obliged to be null at program startup. This is overspecification and overkill. It is both traditional and useful to tie cerr to cout, to ensure that standard output is drained whenever an error message is written. This behavior should at least be permitted if not required. Same for wcerr::tie().

    Proposed resolution:

    Add to the description of cerr:

    After the object cerr is initialized, cerr.tie() returns &cout. Its state is otherwise the same as required for basic_ios<char>::init (lib.basic.ios.cons).

    Add to the description of wcerr:

    After the object wcerr is initialized, wcerr.tie() returns &wcout. Its state is otherwise the same as required for basic_ios<wchar_t>::init (lib.basic.ios.cons).

    [Sydney: straw poll (3-1): we should require, not just permit, cout and cerr to be tied on startup. Pre-Redmond: Bill will provide wording.]


    457. bitset constructor: incorrect number of initialized bits

    Section: 23.3.5.1 [lib.bitset.cons]  Status: DR  Submitter: Dag Henriksson  Date: 30 Jan 2004

    The constructor from unsigned long says it initializes "the first M bit positions to the corresponding bit values in val. M is the smaller of N and the value CHAR_BIT * sizeof(unsigned long)."

    Object-representation vs. value-representation strikes again. CHAR_BIT * sizeof (unsigned long) does not give us the number of bits an unsigned long uses to hold the value. Thus, the first M bit position above is not guaranteed to have any corresponding bit values in val.

    Proposed resolution:

    In 23.3.5.1 [lib.bitset.cons] paragraph 2, change "M is the smaller of N and the value CHAR_BIT * sizeof (unsigned long). (249)" to "M is the smaller of N and the number of bits in the value representation (section 3.9 [basic.types]) of unsigned long."


    460. Default modes missing from basic_fstream member specifications

    Section: 27.8.1 [lib.fstreams]  Status: DR  Submitter: Ben Hutchings  Date: 1 Apr 2004

    The second parameters of the non-default constructor and of the open member function for basic_fstream, named "mode", are optional according to the class declaration in 27.8.1.11 [lib.fstream]. The specifications of these members in 27.8.1.12 [lib.fstream.cons] and 27.8.1.13 lib.fstream.members] disagree with this, though the constructor declaration has the "explicit" function-specifier implying that it is intended to be callable with one argument.

    Proposed resolution:

    In 27.8.1.12 [lib.fstream.cons], change

      explicit basic_fstream(const char* s, ios_base::openmode mode); 
    

    to

      explicit basic_fstream(const char* s,
                             ios_base::openmode mode = ios_base::in|ios_base::out);
    

    In 27.8.1.13 [lib.fstream.members], change

      void open(const char*s, ios_base::openmode mode); 
    

    to

    void open(const char*s, ios_base::openmode mode = ios_base::in|ios_base::out);

    469. vector<bool> ill-formed relational operators

    Section: 23.2.5 [lib.vector.bool]  Status: DR  Submitter: Martin Sebor  Date: 28 Jun 2004

    The overloads of relational operators for vector<bool> specified in [lib.vector.bool] are redundant (they are semantically identical to those provided for the vector primary template) and may even be diagnosed as ill-formed (refer to Daveed Vandevoorde's explanation in c++std-lib-13647).

    Proposed resolution:

    Remove all overloads of overloads of relational operators for vector<bool> from [lib.vector.bool].

    ----- End of document -----