Doc. no. N2482=07-0352
Date: 2007-12-09
Project: Programming Language C++
Reply to: Howard Hinnant <howard.hinnant@gmail.com>

C++ Standard Library Active Issues List (Revision R53)

Reference ISO/IEC IS 14882:1998(E)

Also see:

The purpose of this document is to record the status of issues which have come before the Library Working Group (LWG) of the ANSI (J16) and ISO (WG21) C++ Standards Committee. Issues represent potential defects in the ISO/IEC IS 14882:1998(E) document. Issues are not to be used to request new features.

This document contains only library issues which are actively being considered by the Library Working Group. That is, issues which have a status of New, Open, Ready, and Review. See Library Defect Reports List for issues considered defects and Library Closed Issues List for issues considered closed.

The issues in these lists are not necessarily formal ISO Defect Reports (DR's). While some issues will eventually be elevated to official Defect Report status, other issues will be disposed of in other ways. See Issue Status.

This document is in an experimental format designed for both viewing via a world-wide web browser and hard-copy printing. It is available as an HTML file for browsing or PDF file for printing.

Prior to Revision 14, library issues lists existed in two slightly different versions; a Committee Version and a Public Version. Beginning with Revision 14 the two versions were combined into a single version.

This document includes [bracketed italicized notes] as a reminder to the LWG of current progress on issues. Such notes are strictly unofficial and should be read with caution as they may be incomplete or incorrect. Be aware that LWG support for a particular resolution can quickly change if new viewpoints or killer examples are presented in subsequent discussions.

For the most current official version of this document see http://www.open-std.org/jtc1/sc22/wg21/. Requests for further information about this document should include the document number above, reference ISO/IEC 14882:1998(E), and be submitted to Information Technology Industry Council (ITI), 1250 Eye Street NW, Washington, DC 20005.

Public information as to how to obtain a copy of the C++ Standard, join the standards committee, submit an issue, or comment on an issue can be found in the comp.std.c++ FAQ. Public discussion of C++ Standard related issues occurs on news:comp.std.c++.

For committee members, files available on the committee's private web site include the HTML version of the Standard itself. HTML hyperlinks from this issues list to those files will only work for committee members who have downloaded them into the same disk directory as the issues list files.

Revision History

Issue Status

New - The issue has not yet been reviewed by the LWG. Any Proposed Resolution is purely a suggestion from the issue submitter, and should not be construed as the view of LWG.

Open - The LWG has discussed the issue but is not yet ready to move the issue forward. There are several possible reasons for open status:

A Proposed Resolution for an open issue is still not be construed as the view of LWG. Comments on the current state of discussions are often given at the end of open issues in an italic font. Such comments are for information only and should not be given undue importance.

Dup - The LWG has reached consensus that the issue is a duplicate of another issue, and will not be further dealt with. A Rationale identifies the duplicated issue's issue number.

NAD - The LWG has reached consensus that the issue is not a defect in the Standard.

NAD Editorial - The LWG has reached consensus that the issue can either be handled editorially, or is handled by a paper (usually linked to in the rationale).

NAD Future - In addition to the regular status, the LWG believes that this issue should be revisited at the next revision of the standard.

Review - Exact wording of a Proposed Resolution is now available for review on an issue for which the LWG previously reached informal consensus.

Tentatively Ready - The issue has been reviewed online, but not in a meeting, and some support has been formed for the proposed resolution. Tentatively Ready issues may be moved to Ready and forwarded to full committee within the same meeting. Unlike Ready issues they will be reviewed in subcommittee prior to forwarding to full committee.

Ready - The LWG has reached consensus that the issue is a defect in the Standard, the Proposed Resolution is correct, and the issue is ready to forward to the full committee for further action as a Defect Report (DR).

DR - (Defect Report) - The full J16 committee has voted to forward the issue to the Project Editor to be processed as a Potential Defect Report. The Project Editor reviews the issue, and then forwards it to the WG21 Convenor, who returns it to the full committee for final disposition. This issues list accords the status of DR to all these Defect Reports regardless of where they are in that process.

TC - (Technical Corrigenda) - The full WG21 committee has voted to accept the Defect Report's Proposed Resolution as a Technical Corrigenda. Action on this issue is thus complete and no further action is possible under ISO rules.

TRDec - (Decimal TR defect) - The LWG has voted to accept the Defect Report's Proposed Resolution into the Decimal TR. Action on this issue is thus complete and no further action is expected.

WP - (Working Paper) - The proposed resolution has not been accepted as a Technical Corrigendum, but the full WG21 committee has voted to apply the Defect Report's Proposed Resolution to the working paper.

Pending - This is a status qualifier. When prepended to a status this indicates the issue has been processed by the committee, and a decision has been made to move the issue to the associated unqualified status. However for logistical reasons the indicated outcome of the issue has not yet appeared in the latest working paper.

Issues are always given the status of New when they first appear on the issues list. They may progress to Open or Review while the LWG is actively working on them. When the LWG has reached consensus on the disposition of an issue, the status will then change to Dup, NAD, or Ready as appropriate. Once the full J16 committee votes to forward Ready issues to the Project Editor, they are given the status of Defect Report ( DR). These in turn may become the basis for Technical Corrigenda (TC), or are closed without action other than a Record of Response (RR ). The intent of this LWG process is that only issues which are truly defects in the Standard move to the formal ISO DR status.

Active Issues


23. Num_get overflow result

Section: 22.2.2.1.2 [facet.num.get.virtuals] Status: Open Submitter: Nathan Myers Date: 1998-08-06

View other active issues in [facet.num.get.virtuals].

View all other issues in [facet.num.get.virtuals].

View all issues with Open status.

Discussion:

The current description of numeric input does not account for the possibility of overflow. This is an implicit result of changing the description to rely on the definition of scanf() (which fails to report overflow), and conflicts with the documented behavior of traditional and current implementations.

Users expect, when reading a character sequence that results in a value unrepresentable in the specified type, to have an error reported. The standard as written does not permit this.

Further comments from Dietmar:

I don't feel comfortable with the proposed resolution to issue 23: It kind of simplifies the issue to much. Here is what is going on:

Currently, the behavior of numeric overflow is rather counter intuitive and hard to trace, so I will describe it briefly:

Further discussion from Redmond:

The basic problem is that we've defined our behavior, including our error-reporting behavior, in terms of C90. However, C90's method of reporting overflow in scanf is not technically an "input error". The strto_* functions are more precise.

There was general consensus that failbit should be set upon overflow. We considered three options based on this:

  1. Set failbit upon conversion error (including overflow), and don't store any value.
  2. Set failbit upon conversion error, and also set errno to indicated the precise nature of the error.
  3. Set failbit upon conversion error. If the error was due to overflow, store +-numeric_limits<T>::max() as an overflow indication.

Straw poll: (1) 5; (2) 0; (3) 8.

Discussed at Lillehammer. General outline of what we want the solution to look like: we want to say that overflow is an error, and provide a way to distinguish overflow from other kinds of errors. Choose candidate field the same way scanf does, but don't describe the rest of the process in terms of format. If a finite input field is too large (positive or negative) to be represented as a finite value, then set failbit and assign the nearest representable value. Bill will provide wording.

Discussed at Toronto: N2327 is in alignment with the direction we wanted to go with in Lillehammer. Bill to work on.

Proposed resolution:

Change 22.2.2.1.2 [facet.num.get.virtuals], end of p3:

Stage 3: The result of stage 2 processing can be one of The sequence of chars accumulated in stage 2 (the field) is converted to a numeric value by the rules of one of the functions declared in the header <cstdlib>:

The numeric value to be stored can be one of:

The resultant numeric value is stored in val.

Change 22.2.2.1.2 [facet.num.get.virtuals], p6-p7:

iter_type do_get(iter_type in, iter_type end, ios_base& str, 
                 ios_base::iostate& err, bool& val) const;

-6- Effects: If (str.flags()&ios_base::boolalpha)==0 then input proceeds as it would for a long except that if a value is being stored into val, the value is determined according to the following: If the value to be stored is 0 then false is stored. If the value is 1 then true is stored. Otherwise err|=ios_base::failbit is performed and no value true is stored. and ios_base::failbit is assigned to err.

-7- 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 23.1.1) 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. Otherwise false is stored and ios_base::failbit is assigned to err.


96. Vector<bool> is not a container

Section: 23.2.5 [vector] Status: Open Submitter: AFNOR Date: 1998-10-07

View other active issues in [vector].

View all other issues in [vector].

View all issues with Open status.

Discussion:

vector<bool> is not a container as its reference and pointer types are not references and pointers.

Also it forces everyone to have a space optimization instead of a speed one.

See also: 99-0008 == N1185 Vector<bool> is Nonconforming, Forces Optimization Choice.

[In Santa Cruz the LWG felt that this was Not A Defect.]

[In Dublin many present felt that failure to meet Container requirements was a defect. There was disagreement as to whether or not the optimization requirements constituted a defect.]

[The LWG looked at the following resolutions in some detail:
     * Not A Defect.
     * Add a note explaining that vector<bool> does not meet Container requirements.
     * Remove vector<bool>.
     * Add a new category of container requirements which vector<bool> would meet.
     * Rename vector<bool>.

No alternative had strong, wide-spread, support and every alternative had at least one "over my dead body" response.

There was also mention of a transition scheme something like (1) add vector_bool and deprecate vector<bool> in the next standard. (2) Remove vector<bool> in the following standard.]

[Modifying container requirements to permit returning proxies (thus allowing container requirements conforming vector<bool>) was also discussed.]

[It was also noted that there is a partial but ugly workaround in that vector<bool> may be further specialized with a customer allocator.]

[Kona: Herb Sutter presented his paper J16/99-0035==WG21/N1211, vector<bool>: More Problems, Better Solutions. Much discussion of a two step approach: a) deprecate, b) provide replacement under a new name. LWG straw vote on that: 1-favor, 11-could live with, 2-over my dead body. This resolution was mentioned in the LWG report to the full committee, where several additional committee members indicated over-my-dead-body positions.]

Discussed at Lillehammer. General agreement that we should deprecate vector<bool> and introduce this functionality under a different name, e.g. bit_vector. This might make it possible to remove the vector<bool> specialization in the standard that comes after C++0x. There was also a suggestion that in C++0x we could additional say that it's implementation defined whether vector<bool> refers to the specialization or to the primary template, but there wasn't general agreement that this was a good idea.

We need a paper for the new bit_vector class.

Proposed resolution:

We now have: N2050 and N2160.

[ Batavia: The LWG feels we need something closer to SGI's bitvector to ease migration from vector<bool>. Although some of the funcitonality from N2050 could well be used in such a template. The concern is easing the API migration for those users who want to continue using a bit-packed container. Alan and Beman to work. ]


255. Why do basic_streambuf<>::pbump() and gbump() take an int?

Section: 27.5.2 [streambuf] Status: Open Submitter: Martin Sebor Date: 2000-08-12

View all other issues in [streambuf].

View all issues with Open status.

Discussion:

The basic_streambuf members gbump() and pbump() are specified to take an int argument. This requirement prevents the functions from effectively manipulating buffers larger than std::numeric_limits<int>::max() characters. It also makes the common use case for these functions somewhat difficult as many compilers will issue a warning when an argument of type larger than int (such as ptrdiff_t on LLP64 architectures) is passed to either of the function. Since it's often the result of the subtraction of two pointers that is passed to the functions, a cast is necessary to silence such warnings. Finally, the usage of a native type in the functions signatures is inconsistent with other member functions (such as sgetn() and sputn()) that manipulate the underlying character buffer. Those functions take a streamsize argument.

Proposed resolution:

Change the signatures of these functions in the synopsis of template class basic_streambuf (27.5.2) and in their descriptions (27.5.2.3.1, p4 and 27.5.2.3.2, p4) to take a streamsize argument.

Although this change has the potential of changing the ABI of the library, the change will affect only platforms where int is different than the definition of streamsize. However, since both functions are typically inline (they are on all known implementations), even on such platforms the change will not affect any user code unless it explicitly relies on the existing type of the functions (e.g., by taking their address). Such a possibility is IMO quite remote.

Alternate Suggestion from Howard Hinnant, c++std-lib-7780:

This is something of a nit, but I'm wondering if streamoff wouldn't be a better choice than streamsize. The argument to pbump and gbump MUST be signed. But the standard has this to say about streamsize (27.4.1/2/Footnote):

[Footnote: streamsize is used in most places where ISO C would use size_t. Most of the uses of streamsize could use size_t, except for the strstreambuf constructors, which require negative values. It should probably be the signed type corresponding to size_t (which is what Posix.2 calls ssize_t). --- end footnote]

This seems a little weak for the argument to pbump and gbump. Should we ever really get rid of strstream, this footnote might go with it, along with the reason to make streamsize signed.

Rationale:

The LWG believes this change is too big for now. We may wish to reconsider this for a future revision of the standard. One possibility is overloading pbump, rather than changing the signature.

[ [2006-05-04: Reopened at the request of Chris (Krzysztof ?elechowski)] ]


290. Requirements to for_each and its function object

Section: 25.1.1 [alg.foreach] Status: Open Submitter: Angelika Langer Date: 2001-01-03

View all other issues in [alg.foreach].

View all issues with Open status.

Discussion:

The specification of the for_each algorithm does not have a "Requires" section, which means that there are no restrictions imposed on the function object whatsoever. In essence it means that I can provide any function object with arbitrary side effects and I can still expect a predictable result. In particular I can expect that the function object is applied exactly last - first times, which is promised in the "Complexity" section.

I don't see how any implementation can give such a guarantee without imposing requirements on the function object.

Just as an example: consider a function object that removes elements from the input sequence. In that case, what does the complexity guarantee (applies f exactly last - first times) mean?

One can argue that this is obviously a nonsensical application and a theoretical case, which unfortunately it isn't. I have seen programmers shooting themselves in the foot this way, and they did not understand that there are restrictions even if the description of the algorithm does not say so.

[Lillehammer: This is more general than for_each. We don't want the function object in transform invalidiating iterators either. There should be a note somewhere in clause 17 (17, not 25) saying that user code operating on a range may not invalidate iterators unless otherwise specified. Bill will provide wording.]

Proposed resolution:


299. Incorrect return types for iterator dereference

Section: 24.1.4 [bidirectional.iterators], 24.1.5 [random.access.iterators] Status: Open Submitter: John Potter Date: 2001-01-22

View all other issues in [bidirectional.iterators].

View all issues with Open status.

Discussion:

In section 24.1.4 [bidirectional.iterators], Table 75 gives the return type of *r-- as convertible to T. This is not consistent with Table 74 which gives the return type of *r++ as T&. *r++ = t is valid while *r-- = t is invalid.

In section 24.1.5 [random.access.iterators], Table 76 gives the return type of a[n] as convertible to T. This is not consistent with the semantics of *(a + n) which returns T& by Table 74. *(a + n) = t is valid while a[n] = t is invalid.

Discussion from the Copenhagen meeting: the first part is uncontroversial. The second part, operator[] for Random Access Iterators, requires more thought. There are reasonable arguments on both sides. Return by value from operator[] enables some potentially useful iterators, e.g. a random access "iota iterator" (a.k.a "counting iterator" or "int iterator"). There isn't any obvious way to do this with return-by-reference, since the reference would be to a temporary. On the other hand, reverse_iterator takes an arbitrary Random Access Iterator as template argument, and its operator[] returns by reference. If we decided that the return type in Table 76 was correct, we would have to change reverse_iterator. This change would probably affect user code.

History: the contradiction between reverse_iterator and the Random Access Iterator requirements has been present from an early stage. In both the STL proposal adopted by the committee (N0527==94-0140) and the STL technical report (HPL-95-11 (R.1), by Stepanov and Lee), the Random Access Iterator requirements say that operator[]'s return value is "convertible to T". In N0527 reverse_iterator's operator[] returns by value, but in HPL-95-11 (R.1), and in the STL implementation that HP released to the public, reverse_iterator's operator[] returns by reference. In 1995, the standard was amended to reflect the contents of HPL-95-11 (R.1). The original intent for operator[] is unclear.

In the long term it may be desirable to add more fine-grained iterator requirements, so that access method and traversal strategy can be decoupled. (See "Improved Iterator Categories and Requirements", N1297 = 01-0011, by Jeremy Siek.) Any decisions about issue 299 should keep this possibility in mind.

Further discussion: I propose a compromise between John Potter's resolution, which requires T& as the return type of a[n], and the current wording, which requires convertible to T. The compromise is to keep the convertible to T for the return type of the expression a[n], but to also add a[n] = t as a valid expression. This compromise "saves" the common case uses of random access iterators, while at the same time allowing iterators such as counting iterator and caching file iterators to remain random access iterators (iterators where the lifetime of the object returned by operator*() is tied to the lifetime of the iterator).

Note that the compromise resolution necessitates a change to reverse_iterator. It would need to use a proxy to support a[n] = t.

Note also there is one kind of mutable random access iterator that will no longer meet the new requirements. Currently, iterators that return an r-value from operator[] meet the requirements for a mutable random access iterartor, even though the expression a[n] = t will only modify a temporary that goes away. With this proposed resolution, a[n] = t will be required to have the same operational semantics as *(a + n) = t.

Proposed resolution:

In section 24.1.4 [lib.bidirectdional.iterators], change the return type in table 75 from "convertible to T" to T&.

In section 24.1.5 [lib.random.access.iterators], change the operational semantics for a[n] to " the r-value of a[n] is equivalent to the r-value of *(a + n)". Add a new row in the table for the expression a[n] = t with a return type of convertible to T and operational semantics of *(a + n) = t.

[Lillehammer: Real problem, but should be addressed as part of iterator redesign]


309. Does sentry catch exceptions?

Section: 27.6 [iostream.format] Status: Open Submitter: Martin Sebor Date: 2001-03-19

View other active issues in [iostream.format].

View all other issues in [iostream.format].

View all issues with Open status.

Discussion:

The descriptions of the constructors of basic_istream<>::sentry (27.6.1.1.3 [istream::sentry]) and basic_ostream<>::sentry (27.6.2.4 [ostream::sentry]) do not explain what the functions do in case an exception is thrown while they execute. Some current implementations allow all exceptions to propagate, others catch them and set ios_base::badbit instead, still others catch some but let others propagate.

The text also mentions that the functions may call setstate(failbit) (without actually saying on what object, but presumably the stream argument is meant). That may have been fine for basic_istream<>::sentry prior to issue 195, since the function performs an input operation which may fail. However, issue 195 amends 27.6.1.1.3 [istream::sentry], p2 to clarify that the function should actually call setstate(failbit | eofbit), so the sentence in p3 is redundant or even somewhat contradictory.

The same sentence that appears in 27.6.2.4 [ostream::sentry], p3 doesn't seem to be very meaningful for basic_istream<>::sentry which performs no input. It is actually rather misleading since it would appear to guide library implementers to calling setstate(failbit) when os.tie()->flush(), the only called function, throws an exception (typically, it's badbit that's set in response to such an event).

Additional comments from Martin, who isn't comfortable with the current proposed resolution (see c++std-lib-11530)

The istream::sentry ctor says nothing about how the function deals with exemptions (27.6.1.1.2, p1 says that the class is responsible for doing "exception safe"(*) prefix and suffix operations but it doesn't explain what level of exception safety the class promises to provide). The mockup example of a "typical implementation of the sentry ctor" given in 27.6.1.1.2, p6, removed in ISO/IEC 14882:2003, doesn't show exception handling, either. Since the ctor is not classified as a formatted or unformatted input function, the text in 27.6.1.1, p1 through p4 does not apply. All this would seem to suggest that the sentry ctor should not catch or in any way handle exceptions thrown from any functions it may call. Thus, the typical implementation of an istream extractor may look something like [1].

The problem with [1] is that while it correctly sets ios::badbit if an exception is thrown from one of the functions called from the sentry ctor, if the sentry ctor reaches EOF while extracting whitespace from a stream that has eofbit or failbit set in exceptions(), it will cause an ios::failure to be thrown, which will in turn cause the extractor to set ios::badbit.

The only straightforward way to prevent this behavior is to move the definition of the sentry object in the extractor above the try block (as suggested by the example in 22.2.8, p9 and also indirectly supported by 27.6.1.3, p1). See [2]. But such an implementation will allow exceptions thrown from functions called from the ctor to freely propagate to the caller regardless of the setting of ios::badbit in the stream object's exceptions().

So since neither [1] nor [2] behaves as expected, the only possible solution is to have the sentry ctor catch exceptions thrown from called functions, set badbit, and propagate those exceptions if badbit is also set in exceptions(). (Another solution exists that deals with both kinds of sentries, but the code is non-obvious and cumbersome -- see [3].)

Please note that, as the issue points out, current libraries do not behave consistently, suggesting that implementors are not quite clear on the exception handling in istream::sentry, despite the fact that some LWG members might feel otherwise. (As documented by the parenthetical comment here: http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/papers/2003/n1480.html#309)

Also please note that those LWG members who in Copenhagen felt that "a sentry's constructor should not catch exceptions, because sentries should only be used within (un)formatted input functions and that exception handling is the responsibility of those functions, not of the sentries," as noted here http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/papers/2001/n1310.html#309 would in effect be either arguing for the behavior described in [1] or for extractors implemented along the lines of [3].

The original proposed resolution (Revision 25 of the issues list) clarifies the role of the sentry ctor WRT exception handling by making it clear that extractors (both library or user-defined) should be implemented along the lines of [2] (as opposed to [1]) and that no exception thrown from the callees should propagate out of either function unless badbit is also set in exceptions().

[1] Extractor that catches exceptions thrown from sentry:

struct S { long i; };

istream& operator>> (istream &strm, S &s)
{
    ios::iostate err = ios::goodbit;
    try {
        const istream::sentry guard (strm, false);
        if (guard) {
            use_facet<num_get<char> >(strm.getloc ())
                .get (istreambuf_iterator<char>(strm),
                      istreambuf_iterator<char>(),
                      strm, err, s.i);
        }
    }
    catch (...) {
        bool rethrow;
        try {
            strm.setstate (ios::badbit);
            rethrow = false;
        }
        catch (...) {
            rethrow = true;
        }
        if (rethrow)
            throw;
    }
    if (err)
        strm.setstate (err);
    return strm;
}

[2] Extractor that propagates exceptions thrown from sentry:

istream& operator>> (istream &strm, S &s)
{
    istream::sentry guard (strm, false);
    if (guard) {
        ios::iostate err = ios::goodbit;
        try {
            use_facet<num_get<char> >(strm.getloc ())
                .get (istreambuf_iterator<char>(strm),
                      istreambuf_iterator<char>(),
                      strm, err, s.i);
        }
        catch (...) {
            bool rethrow;
            try {
                strm.setstate (ios::badbit);
                rethrow = false;
            }
            catch (...) {
                rethrow = true;
            }
            if (rethrow)
                throw;
        }
        if (err)
            strm.setstate (err);
    }
    return strm;
}

[3] Extractor that catches exceptions thrown from sentry but doesn't set badbit if the exception was thrown as a result of a call to strm.clear().

istream& operator>> (istream &strm, S &s)
{
    const ios::iostate state = strm.rdstate ();
    const ios::iostate except = strm.exceptions ();
    ios::iostate err = std::ios::goodbit;
    bool thrown = true;
    try {
        const istream::sentry guard (strm, false);
        thrown = false;
        if (guard) {
            use_facet<num_get<char> >(strm.getloc ())
                .get (istreambuf_iterator<char>(strm),
                      istreambuf_iterator<char>(),
                      strm, err, s.i);
        }
    }
    catch (...) {
        if (thrown && state & except)
            throw;
        try {
            strm.setstate (ios::badbit);
            thrown = false;
        }
        catch (...) {
            thrown = true;
        }
        if (thrown)
            throw;
    }
    if (err)
        strm.setstate (err);

    return strm;
}

[Pre-Berlin] Reopened at the request of Paolo Carlini and Steve Clamage.

[Pre-Portland] A relevant newsgroup post:

The current proposed resolution of issue #309 (http://www.open-std.org/jtc1/sc22/wg21/docs/lwg-active.html#309) is unacceptable. I write commerical software and coding around this makes my code ugly, non-intuitive, and requires comments referring people to this very issue. Following is the full explanation of my experience.

In the course of writing software for commercial use, I constructed std::ifstream's based on user-supplied pathnames on typical POSIX systems.

It was expected that some files that opened successfully might not read successfully -- such as a pathname which actually refered to a directory. Intuitively, I expected the streambuffer underflow() code to throw an exception in this situation, and recent implementations of libstdc++'s basic_filebuf do just that (as well as many of my own custom streambufs).

I also intuitively expected that the istream code would convert these exceptions to the "badbit' set on the stream object, because I had not requested exceptions. I refer to 27.6.1.1. P4.

However, this was not the case on at least two implementations -- if the first thing I did with an istream was call operator>>( T& ) for T among the basic arithmetic types and std::string. Looking further I found that the sentry's constructor was invoking the exception when it pre-scanned for whitespace, and the extractor function (operator>>()) was not catching exceptions in this situation.

So, I was in a situation where setting 'noskipws' would change the istream's behavior even though no characters (whitespace or not) could ever be successfully read.

Also, calling .peek() on the istream before calling the extractor() changed the behavior (.peek() had the effect of setting the badbit ahead of time).

I found this all to be so inconsistent and inconvenient for me and my code design, that I filed a bugzilla entry for libstdc++. I was then told that the bug cannot be fixed until issue #309 is resolved by the committee.

Proposed resolution:

Rationale:

The LWG agrees there is minor variation between implementations, but believes that it doesn't matter. This is a rarely used corner case. There is no evidence that this has any commercial importance or that it causes actual portability problems for customers trying to write code that runs on multiple implementations.


342. seek and eofbit

Section: 27.6.1.3 [istream.unformatted] Status: Open Submitter: Howard Hinnant Date: 2001-10-09

View all other issues in [istream.unformatted].

View all issues with Open status.

Discussion:

I think we have a defect.

According to lwg issue 60 which is now a dr, the description of seekg in 27.6.1.3 [istream.unformatted] paragraph 38 now looks like:

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() != true, executes rdbuf()->pubseekpos( pos).

And according to lwg issue 243 which is also now a dr, 27.6.1.3, paragraph 1 looks like:

Each unformatted input function begins execution by constructing an object of class sentry with the default argument noskipws (second) argument true. If the sentry object returns true, when converted to a value of type bool, the function endeavors to obtain the requested input. 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. If an exception is thrown during input then ios::badbit is turned on in *this'ss error state. If (exception()&badbit)!= 0 then the exception is rethrown. It also counts the number of characters extracted. If no exception has been thrown it ends by storing the count in a member object and returning the value specified. In any event the sentry object is destroyed before leaving the unformatted input function.

And finally 27.6.1.1.2/5 says this about sentry:

If, after any preparation is completed, is.good() is true, ok_ != false otherwise, ok_ == false.

So although the seekg paragraph says that the operation proceeds if !fail(), the behavior of unformatted functions says the operation proceeds only if good(). The two statements are contradictory when only eofbit is set. I don't think the current text is clear which condition should be respected.

Further discussion from Redmond:

PJP: It doesn't seem quite right to say that seekg is "unformatted". That makes specific claims about sentry that aren't quite appropriate for seeking, which has less fragile failure modes than actual input. If we do really mean that it's unformatted input, it should behave the same way as other unformatted input. On the other hand, "principle of least surprise" is that seeking from EOF ought to be OK.

Pre-Berlin: Paolo points out several problems with the proposed resolution in Ready state:

Proposed resolution:

Change 27.6.1.3 [istream.unformatted] 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, does not affect the value returned by subsequent calls to gcount(), and does not examine the value returned by the sentry object. After constructing a sentry object, if fail() != true, executes rdbuf()->pubseekpos(pos). In case of success, the function calls clear(). In case of failure, the function calls setstate(failbit) (which may throw ios_base::failure).

[Lillehammer: Matt provided wording.]

Rationale:

In C, fseek does clear EOF. This is probably what most users would expect. We agree that having eofbit set should not deter a seek, and that a successful seek should clear eofbit. Note that fail() is true only if failbit or badbit is set, so using !fail(), rather than good(), satisfies this goal.


343. Unspecified library header dependencies

Section: 21 [strings], 23 [containers], 27 [input.output] Status: Open Submitter: Martin Sebor Date: 2001-10-09

View all other issues in [strings].

View all issues with Open status.

Discussion:

The synopses of the C++ library headers clearly show which names are required to be defined in each header. Since in order to implement the classes and templates defined in these headers declarations of other templates (but not necessarily their definitions) are typically necessary the standard in 17.4.4, p1 permits library implementers to include any headers needed to implement the definitions in each header.

For instance, although it is not explicitly specified in the synopsis of <string>, at the point of definition of the std::basic_string template the declaration of the std::allocator template must be in scope. All current implementations simply include <memory> from within <string>, either directly or indirectly, to bring the declaration of std::allocator into scope.

Additionally, however, some implementation also include <istream> and <ostream> at the top of <string> to bring the declarations of std::basic_istream and std::basic_ostream into scope (which are needed in order to implement the string inserter and extractor operators (21.3.7.9 [lib.string.io])). Other implementations only include <iosfwd>, since strictly speaking, only the declarations and not the full definitions are necessary.

Obviously, it is possible to implement <string> without actually providing the full definitions of all the templates std::basic_string uses (std::allocator, std::basic_istream, and std::basic_ostream). Furthermore, not only is it possible, doing so is likely to have a positive effect on compile-time efficiency.

But while it may seem perfectly reasonable to expect a program that uses the std::basic_string insertion and extraction operators to also explicitly include <istream> or <ostream>, respectively, it doesn't seem reasonable to also expect it to explicitly include <memory>. Since what's reasonable and what isn't is highly subjective one would expect the standard to specify what can and what cannot be assumed. Unfortunately, that isn't the case.

The examples below demonstrate the issue.

Example 1:

It is not clear whether the following program is complete:

#include <string>

extern std::basic_ostream<char> &strm;

int main () {
    strm << std::string ("Hello, World!\n");
}

or whether one must explicitly include <memory> or <ostream> (or both) in addition to <string> in order for the program to compile.

Example 2:

Similarly, it is unclear whether the following program is complete:

#include <istream>

extern std::basic_iostream<char> &strm;

int main () {
    strm << "Hello, World!\n";
}

or whether one needs to explicitly include <ostream>, and perhaps even other headers containing the definitions of other required templates:

#include <ios>
#include <istream>
#include <ostream>
#include <streambuf>

extern std::basic_iostream<char> &strm;

int main () {
    strm << "Hello, World!\n";
}

Example 3:

Likewise, it seems unclear whether the program below is complete:

#include <iterator>

bool foo (std::istream_iterator<int> a, std::istream_iterator<int> b)
{
    return a == b;
}

int main () { }

or whether one should be required to include <istream>.

There are many more examples that demonstrate this lack of a requirement. I believe that in a good number of cases it would be unreasonable to require that a program explicitly include all the headers necessary for a particular template to be specialized, but I think that there are cases such as some of those above where it would be desirable to allow implementations to include only as much as necessary and not more.

Proposed resolution:

For every C++ library header, supply a minimum set of other C++ library headers that are required to be included by that header. The proposed list is below (C++ headers for C Library Facilities, table 12 in 17.4.1.2, p3, are omitted):

+------------+--------------------+
| C++ header |required to include |
+============+====================+
|<algorithm> |                    |
+------------+--------------------+
|<bitset>    |                    |
+------------+--------------------+
|<complex>   |                    |
+------------+--------------------+
|<deque>     |<memory>            |
+------------+--------------------+
|<exception> |                    |
+------------+--------------------+
|<fstream>   |<ios>               |
+------------+--------------------+
|<functional>|                    |
+------------+--------------------+
|<iomanip>   |<ios>               |
+------------+--------------------+
|<ios>       |<streambuf>         |
+------------+--------------------+
|<iosfwd>    |                    |
+------------+--------------------+
|<iostream>  |<istream>, <ostream>|
+------------+--------------------+
|<istream>   |<ios>               |
+------------+--------------------+
|<iterator>  |                    |
+------------+--------------------+
|<limits>    |                    |
+------------+--------------------+
|<list>      |<memory>            |
+------------+--------------------+
|<locale>    |                    |
+------------+--------------------+
|<map>       |<memory>            |
+------------+--------------------+
|<memory>    |                    |
+------------+--------------------+
|<new>       |<exception>         |
+------------+--------------------+
|<numeric>   |                    |
+------------+--------------------+
|<ostream>   |<ios>               |
+------------+--------------------+
|<queue>     |<deque>             |
+------------+--------------------+
|<set>       |<memory>            |
+------------+--------------------+
|<sstream>   |<ios>, <string>     |
+------------+--------------------+
|<stack>     |<deque>             |
+------------+--------------------+
|<stdexcept> |                    |
+------------+--------------------+
|<streambuf> |<ios>               |
+------------+--------------------+
|<string>    |<memory>            |
+------------+--------------------+
|<strstream> |                    |
+------------+--------------------+
|<typeinfo>  |<exception>         |
+------------+--------------------+
|<utility>   |                    |
+------------+--------------------+
|<valarray>  |                    |
+------------+--------------------+
|<vector>    |<memory>            |
+------------+--------------------+

Rationale:

The portability problem is real. A program that works correctly on one implementation might fail on another, because of different header dependencies. This problem was understood before the standard was completed, and it was a conscious design choice.

One possible way to deal with this, as a library extension, would be an <all> header.

Hinnant: It's time we dealt with this issue for C++0X. Reopened.


382. codecvt do_in/out result

Section: 22.2.1.4 [locale.codecvt] Status: Open Submitter: Martin Sebor Date: 2002-08-30

View all other issues in [locale.codecvt].

View all issues with Open status.

Discussion:

It seems that the descriptions of codecvt do_in() and do_out() leave sufficient room for interpretation so that two implementations of codecvt may not work correctly with the same filebuf. Specifically, the following seems less than adequately specified:

  1. the conditions under which the functions terminate
  2. precisely when the functions return ok
  3. precisely when the functions return partial
  4. the full set of conditions when the functions return error
  1. 22.2.1.4.2 [locale.codecvt.virtuals], p2 says this about the effects of the function: ...Stops if it encounters a character it cannot convert... This assumes that there *is* a character to convert. What happens when there is a sequence that doesn't form a valid source character, such as an unassigned or invalid UNICODE character, or a sequence that cannot possibly form a character (e.g., the sequence "\xc0\xff" in UTF-8)?
  2. Table 53 says that the function returns codecvt_base::ok to indicate that the function(s) "completed the conversion." Suppose that the source sequence is "\xc0\x80" in UTF-8, with from pointing to '\xc0' and (from_end==from + 1). It is not clear whether the return value should be ok or partial (see below).
  3. Table 53 says that the function returns codecvt_base::partial if "not all source characters converted." With the from pointers set up the same way as above, it is not clear whether the return value should be partial or ok (see above).
  4. Table 53, in the row describing the meaning of error mistakenly refers to a "from_type" character, without the symbol from_type having been defined. Most likely, the word "source" character is intended, although that is not sufficient. The functions may also fail when they encounter an invalid source sequence that cannot possibly form a valid source character (e.g., as explained in bullet 1 above).

Finally, the conditions described at the end of 22.2.1.4.2 [locale.codecvt.virtuals], p4 don't seem to be possible:

"A return value of partial, if (from_next == from_end), indicates that either the destination sequence has not absorbed all the available destination elements, or that additional source elements are needed before another destination element can be produced."

If the value is partial, it's not clear to me that (from_next ==from_end) could ever hold if there isn't enough room in the destination buffer. In order for (from_next==from_end) to hold, all characters in that range must have been successfully converted (according to 22.2.1.4.2 [locale.codecvt.virtuals], p2) and since there are no further source characters to convert, no more room in the destination buffer can be needed.

It's also not clear to me that (from_next==from_end) could ever hold if additional source elements are needed to produce another destination character (not element as incorrectly stated in the text). partial is returned if "not all source characters have been converted" according to Table 53, which also implies that (from_next==from) does NOT hold.

Could it be that the intended qualifying condition was actually (from_next != from_end), i.e., that the sentence was supposed to read

"A return value of partial, if (from_next != from_end),..."

which would make perfect sense, since, as far as I understand it, partial can only occur if (from_next != from_end)?

[Lillehammer: Defer for the moment, but this really needs to be fixed. Right now, the description of codecvt is too vague for it to be a useful contract between providers and clients of codecvt facets. (Note that both vendors and users can be both providers and clients of codecvt facets.) The major philosophical issue is whether the standard should only describe mappings that take a single wide character to multiple narrow characters (and vice versa), or whether it should describe fully general N-to-M conversions. When the original standard was written only the former was contemplated, but today, in light of the popularity of utf8 and utf16, that doesn't seem sufficient for C++0x. Bill supports general N-to-M conversions; we need to make sure Martin and Howard agree.]

Proposed resolution:


387. std::complex over-encapsulated

Section: 26.3 [complex.numbers] Status: Open Submitter: Gabriel Dos Reis Date: 2002-11-08

View all other issues in [complex.numbers].

View all issues with Open status.

Discussion:

The absence of explicit description of std::complex<T> layout makes it imposible to reuse existing software developed in traditional languages like Fortran or C with unambigous and commonly accepted layout assumptions. There ought to be a way for practitioners to predict with confidence the layout of std::complex<T> whenever T is a numerical datatype. The absence of ways to access individual parts of a std::complex<T> object as lvalues unduly promotes severe pessimizations. For example, the only way to change, independently, the real and imaginary parts is to write something like

complex<T> z;
// ...
// set the real part to r
z = complex<T>(r, z.imag());
// ...
// set the imaginary part to i
z = complex<T>(z.real(), i);

At this point, it seems appropriate to recall that a complex number is, in effect, just a pair of numbers with no particular invariant to maintain. Existing practice in numerical computations has it that a complex number datatype is usually represented by Cartesian coordinates. Therefore the over-encapsulation put in the specification of std::complex<> is not justified.

Proposed resolution:

Add the following requirements to 26.3 [complex.numbers] as 26.3/4:

If z is an lvalue expression of type cv std::complex<T> then

Moreover, if a is an expression of pointer type cv complex<T>* and the expression a[i] is well-defined for an integer expression i then:

In the header synopsis in 26.3.1 [complex.synopsis], replace

  template<class T> T real(const complex<T>&);
  template<class T> T imag(const complex<T>&);

with

  template<class T> const T& real(const complex<T>&);
  template<class T>       T& real(      complex<T>&);
  template<class T> const T& imag(const complex<T>&);
  template<class T>       T& imag(      complex<T>&);

In 26.3.7 [complex.value.ops] paragraph 1, change

  template<class T> T real(const complex<T>&);

to

  template<class T> const T& real(const complex<T>&);
  template<class T>       T& real(      complex<T>&);

and change the Returns clause to "Returns: The real part of x.

In 26.3.7 [complex.value.ops] paragraph 2, change

  template<class T> T imag(const complex<T>&);

to

  template<class T> const T& imag(const complex<T>&);
  template<class T>       T& imag(      complex<T>&);

and change the Returns clause to "Returns: The imaginary part of x.

[Kona: The layout guarantee is absolutely necessary for C compatibility. However, there was disagreement about the other part of this proposal: retrieving elements of the complex number as lvalues. An alternative: continue to have real() and imag() return rvalues, but add set_real() and set_imag(). Straw poll: return lvalues - 2, add setter functions - 5. Related issue: do we want reinterpret_cast as the interface for converting a complex to an array of two reals, or do we want to provide a more explicit way of doing it? Howard will try to resolve this issue for the next meeting.]

[pre-Sydney: Howard summarized the options in n1589.]

Rationale:

The LWG believes that C99 compatibility would be enough justification for this change even without other considerations. All existing implementations already have the layout proposed here.


394. behavior of formatted output on failure

Section: 27.6.2.6.1 [ostream.formatted.reqmts] Status: Open Submitter: Martin Sebor Date: 2002-12-27

View all issues with Open status.

Discussion:

There is a contradiction in Formatted output about what bit is supposed to be set if the formatting fails. On sentence says it's badbit and another that it's failbit.

27.6.2.5.1, p1 says in the Common Requirements on Formatted output functions:

     ... If the generation fails, then the formatted output function
     does setstate(ios::failbit), which might throw an exception.

27.6.2.5.2, p1 goes on to say this about Arithmetic Inserters:

... 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();

     ... If failed is true then does setstate(badbit) ...

The original intent of the text, according to Jerry Schwarz (see c++std-lib-10500), is captured in the following paragraph:

In general "badbit" should mean that the stream is unusable because of some underlying failure, such as disk full or socket closure; "failbit" should mean that the requested formatting wasn't possible because of some inconsistency such as negative widths. So typically if you clear badbit and try to output something else you'll fail again, but if you clear failbit and try to output something else you'll succeed.

In the case of the arithmetic inserters, since num_put cannot report failure by any means other than exceptions (in response to which the stream must set badbit, which prevents the kind of recoverable error reporting mentioned above), the only other detectable failure is if the iterator returned from num_put returns true from failed().

Since that can only happen (at least with the required iostream specializations) under such conditions as the underlying failure referred to above (e.g., disk full), setting badbit would seem to be the appropriate response (indeed, it is required in 27.6.2.5.2, p1). It follows that failbit can never be directly set by the arithmetic (it can only be set by the sentry object under some unspecified conditions).

The situation is different for other formatted output functions which can fail as a result of the streambuf functions failing (they may do so by means other than exceptions), and which are then required to set failbit.

The contradiction, then, is that ostream::operator<<(int) will set badbit if the disk is full, while operator<<(ostream&, char) will set failbit under the same conditions. To make the behavior consistent, the Common requirements sections for the Formatted output functions should be changed as proposed below.

[Kona: There's agreement that this is a real issue. What we decided at Kona: 1. An error from the buffer (which can be detected either directly from streambuf's member functions or by examining a streambuf_iterator) should always result in badbit getting set. 2. There should never be a circumstance where failbit gets set. That represents a formatting error, and there are no circumstances under which the output facets are specified as signaling a formatting error. (Even more so for string output that for numeric because there's nothing to format.) If we ever decide to make it possible for formatting errors to exist then the facets can signal the error directly, and that should go in clause 22, not clause 27. 3. The phrase "if generation fails" is unclear and should be eliminated. It's not clear whether it's intended to mean a buffer error (e.g. a full disk), a formatting error, or something else. Most people thought it was supposed to refer to buffer errors; if so, we should say so. Martin will provide wording.]

Proposed resolution:

Rationale:


396. what are characters zero and one

Section: 23.3.5.1 [bitset.cons] Status: Open Submitter: Martin Sebor Date: 2003-01-05

View all other issues in [bitset.cons].

View all issues with Open status.

Discussion:

23.3.5.1, p6 [lib.bitset.cons] talks about a generic character having the value of 0 or 1 but there is no definition of what that means for charT other than char and wchar_t. And even for those two types, the values 0 and 1 are not actually what is intended -- the values '0' and '1' are. This, along with the converse problem in the description of to_string() in 23.3.5.2, p33, looks like a defect remotely related to DR 303.

http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/lwg-defects.html#303

23.3.5.1:
  -6-  An element of the constructed string has value zero if the
       corresponding character in str, beginning at position pos,
       is 0. Otherwise, the element has the value one.
    
23.3.5.2:
  -33-  Effects: Constructs a string object of the appropriate
        type and initializes it to a string of length N characters.
        Each character is determined by the value of its
        corresponding bit position in *this. Character position N
        ?- 1 corresponds to bit position zero. Subsequent decreasing
        character positions correspond to increasing bit positions.
        Bit value zero becomes the character 0, bit value one becomes
        the character 1.
    

Also note the typo in 23.3.5.1, p6: the object under construction is a bitset, not a string.

Proposed resolution:

Change the constructor's function declaration immediately before 23.3.5.1 [bitset.cons] p3 to:

    template <class charT, class traits, class Allocator>
    explicit
    bitset(const basic_string<charT, traits, Allocator>& str,
           typename basic_string<charT, traits, Allocator>::size_type pos = 0,
           typename basic_string<charT, traits, Allocator>::size_type n =
             basic_string<charT, traits, Allocator>::npos,
           charT zero = charT('0'), charT one = charT('1'))

Change the first two sentences of 23.3.5.1 [bitset.cons] p6 to: "An element of the constructed string has value 0 if the corresponding character in str, beginning at position pos, is zero. Otherwise, the element has the value 1.

Change the text of the second sentence in 23.3.5.1, p5 to read: "The function then throws invalid_argument if any of the rlen characters in str beginning at position pos is other than zero or one. The function uses traits::eq() to compare the character values."

Change the declaration of the to_string member function immediately before 23.3.5.2 [bitset.members] p33 to:

    template <class charT, class traits, class Allocator>
    basic_string<charT, traits, Allocator> 
    to_string(charT zero = charT('0'), charT one = charT('1')) const;

Change the last sentence of 23.3.5.2 [bitset.members] p33 to: "Bit value 0 becomes the character zero, bit value 1 becomes the character one.

Change 23.3.5.3 [bitset.operators] p8 to:

Returns:

  os << x.template to_string<charT,traits,allocator<charT> >(
      use_facet<ctype<charT> >(os.getloc()).widen('0'),
      use_facet<ctype<charT> >(os.getloc()).widen('1'));

Rationale:

There is a real problem here: we need the character values of '0' and '1', and we have no way to get them since strings don't have imbued locales. In principle the "right" solution would be to provide an extra object, either a ctype facet or a full locale, which would be used to widen '0' and '1'. However, there was some discomfort about using such a heavyweight mechanism. The proposed resolution allows those users who care about this issue to get it right.

We fix the inserter to use the new arguments. Note that we already fixed the analogous problem with the extractor in issue 303.


397. ostream::sentry dtor throws exceptions

Section: 27.6.2.4 [ostream::sentry] Status: Open Submitter: Martin Sebor Date: 2003-01-05

View other active issues in [ostream::sentry].

View all other issues in [ostream::sentry].

View all issues with Open status.

Discussion:

17.4.4.8, p3 prohibits library dtors from throwing exceptions.

27.6.2.3, p4 says this about the ostream::sentry dtor:

    -4- If ((os.flags() & ios_base::unitbuf) && !uncaught_exception())
        is true, calls os.flush().
    

27.6.2.6, p7 that describes ostream::flush() says:

    -7- If rdbuf() is not a null pointer, calls rdbuf()->pubsync().
        If that function returns ?-1 calls setstate(badbit) (which
        may throw ios_base::failure (27.4.4.3)).
    

That seems like a defect, since both pubsync() and setstate() can throw an exception.

[ The contradiction is real. Clause 17 says destructors may never throw exceptions, and clause 27 specifies a destructor that does throw. In principle we might change either one. We're leaning toward changing clause 17: putting in an "unless otherwise specified" clause, and then putting in a footnote saying the sentry destructor is the only one that can throw. PJP suggests specifying that sentry::~sentry() should internally catch any exceptions it might cause. ]

[ See 418 and 622 for related issues. ]

Proposed resolution:


398. effects of end-of-file on unformatted input functions

Section: 27.6.2.4 [ostream::sentry] Status: Open Submitter: Martin Sebor Date: 2003-01-05

View other active issues in [ostream::sentry].

View all other issues in [ostream::sentry].

View all issues with Open status.

Discussion:

While reviewing unformatted input member functions of istream for their behavior when they encounter end-of-file during input I found that the requirements vary, sometimes unexpectedly, and in more than one case even contradict established practice (GNU libstdc++ 3.2, IBM VAC++ 6.0, STLPort 4.5, SunPro 5.3, HP aCC 5.38, Rogue Wave libstd 3.1, and Classic Iostreams).

The following unformatted input member functions set eofbit if they encounter an end-of-file (this is the expected behavior, and also the behavior of all major implementations):

    basic_istream<charT, traits>&
    get (char_type*, streamsize, char_type);
    

Also sets failbit if it fails to extract any characters.

    basic_istream<charT, traits>&
    get (char_type*, streamsize);
    

Also sets failbit if it fails to extract any characters.

    basic_istream<charT, traits>&
    getline (char_type*, streamsize, char_type);
    

Also sets failbit if it fails to extract any characters.

    basic_istream<charT, traits>&
    getline (char_type*, streamsize);
    

Also sets failbit if it fails to extract any characters.

    basic_istream<charT, traits>&
    ignore (int, int_type);
    
    basic_istream<charT, traits>&
    read (char_type*, streamsize);
    

Also sets failbit if it encounters end-of-file.

    streamsize readsome (char_type*, streamsize);
    

The following unformated input member functions set failbit but not eofbit if they encounter an end-of-file (I find this odd since the functions make it impossible to distinguish a general failure from a failure due to end-of-file; the requirement is also in conflict with all major implementation which set both eofbit and failbit):

    int_type get();
    
    basic_istream<charT, traits>&
    get (char_type&);
    

These functions only set failbit of they extract no characters, otherwise they don't set any bits, even on failure (I find this inconsistency quite unexpected; the requirement is also in conflict with all major implementations which set eofbit whenever they encounter end-of-file):

    basic_istream<charT, traits>&
    get (basic_streambuf<charT, traits>&, char_type);
    
    basic_istream<charT, traits>&
    get (basic_streambuf<charT, traits>&);
    

This function sets no bits (all implementations except for STLport and Classic Iostreams set eofbit when they encounter end-of-file):

    int_type peek ();
    

Informally, what we want is a global statement of intent saying that eofbit gets set if we trip across EOF, and then we can take away the specific wording for individual functions. A full review is necessary. The wording currently in the standard is a mishmash, and changing it on an individual basis wouldn't make things better. Dietmar will do this work.

Proposed resolution:


408. Is vector<reverse_iterator<char*> > forbidden?

Section: 24.1 [iterator.requirements] Status: Open Submitter: Nathan Myers Date: 2003-06-03

View other active issues in [iterator.requirements].

View all other issues in [iterator.requirements].

View all issues with Open status.

Discussion:

I've been discussing iterator semantics with Dave Abrahams, and a surprise has popped up. I don't think this has been discussed before.

24.1 [iterator.requirements] says that the only operation that can be performed on "singular" iterator values is to assign a non-singular value to them. (It doesn't say they can be destroyed, and that's probably a defect.) Some implementations have taken this to imply that there is no need to initialize the data member of a reverse_iterator<> in the default constructor. As a result, code like

  std::vector<std::reverse_iterator<char*> > v(7);
  v.reserve(1000);

invokes undefined behavior, because it must default-initialize the vector elements, and then copy them to other storage. Of course many other vector operations on these adapters are also left undefined, and which those are is not reliably deducible from the standard.

I don't think that 24.1 was meant to make standard-library iterator types unsafe. Rather, it was meant to restrict what operations may be performed by functions which take general user- and standard iterators as arguments, so that raw pointers would qualify as iterators. However, this is not clear in the text, others have come to the opposite conclusion.

One question is whether the standard iterator adaptors have defined copy semantics. Another is whether they have defined destructor semantics: is

  { std::vector<std::reverse_iterator<char*> >  v(7); }

undefined too?

Note this is not a question of whether algorithms are allowed to rely on copy semantics for arbitrary iterators, just whether the types we actually supply support those operations. I believe the resolution must be expressed in terms of the semantics of the adapter's argument type. It should make clear that, e.g., the reverse_iterator<T> constructor is actually required to execute T(), and so copying is defined if the result of T() is copyable.

Issue 235, which defines reverse_iterator's default constructor more precisely, has some relevance to this issue. However, it is not the whole story.

The issue was whether

  reverse_iterator() { }

is allowed, vs.

  reverse_iterator() : current() { }

The difference is when T is char*, where the first leaves the member uninitialized, and possibly equal to an existing pointer value, or (on some targets) may result in a hardware trap when copied.

8.5 paragraph 5 seems to make clear that the second is required to satisfy DR 235, at least for non-class Iterator argument types.

But that only takes care of reverse_iterator, and doesn't establish a policy for all iterators. (The reverse iterator adapter was just an example.) In particular, does my function

  template <typename Iterator>
    void f() { std::vector<Iterator>  v(7); } 

evoke undefined behavior for some conforming iterator definitions? I think it does, now, because vector<> will destroy those singular iterator values, and that's explicitly disallowed.

24.1 shouldn't give blanket permission to copy all singular iterators, because then pointers wouldn't qualify as iterators. However, it should allow copying of that subset of singular iterator values that are default-initialized, and it should explicitly allow destroying any iterator value, singular or not, default-initialized or not.

Related issue: 407

[ We don't want to require all singular iterators to be copyable, because that is not the case for pointers. However, default construction may be a special case. Issue: is it really default construction we want to talk about, or is it something like value initialization? We need to check with core to see whether default constructed pointers are required to be copyable; if not, it would be wrong to impose so strict a requirement for iterators. ]

Proposed resolution:


417. what does ctype::do_widen() return on failure

Section: 22.2.1.1.2 [locale.ctype.virtuals] Status: Open Submitter: Martin Sebor Date: 2003-09-18

View all other issues in [locale.ctype.virtuals].

View all issues with Open status.

Discussion:

The Effects and Returns clauses of the do_widen() member function of the ctype facet fail to specify the behavior of the function on failure. That the function may not be able to simply cast the narrow character argument to the type of the result since doing so may yield the wrong value for some wchar_t encodings. Popular implementations of ctype<wchar_t> that use mbtowc() and UTF-8 as the native encoding (e.g., GNU glibc) will fail when the argument's MSB is set. There is no way for the the rest of locale and iostream to reliably detect this failure.

[Kona: This is a real problem. Widening can fail. It's unclear what the solution should be. Returning WEOF works for the wchar_t specialization, but not in general. One option might be to add a default, like narrow. But that's an incompatible change. Using traits::eof might seem like a good idea, but facets don't have access to traits (a recurring problem). We could have widen throw an exception, but that's a scary option; existing library components aren't written with the assumption that widen can throw.]

Proposed resolution:


418. exceptions thrown during iostream cleanup

Section: 27.4.2.1.6 [ios::Init] Status: Open Submitter: Martin Sebor Date: 2003-09-18

View all issues with Open status.

Discussion:

The dtor of the ios_base::Init object is supposed to call flush() on the 6 standard iostream objects cout, cerr, clog, wcout, wcerr, and wclog. This call may cause an exception to be thrown.

17.4.4.8, p3 prohibits all library destructors from throwing exceptions.

The question is: What should this dtor do if one or more of these calls to flush() ends up throwing an exception? This can happen quite easily if one of the facets installed in the locale imbued in the iostream object throws.

[Kona: We probably can't do much better than what we've got, so the LWG is leaning toward NAD. At the point where the standard stream objects are being cleaned up, the usual error reporting mechanism are all unavailable. And exception from flush at this point will definitely cause problems. A quality implementation might reasonably swallow the exception, or call abort, or do something even more drastic.]

[ See 397 and 622 for related issues. ]

Proposed resolution:


419. istream extractors not setting failbit if eofbit is already set

Section: 27.6.1.1.3 [istream::sentry] Status: Open Submitter: Martin Sebor Date: 2003-09-18

View all other issues in [istream::sentry].

View all issues with Open status.

Discussion:

27.6.1.1.3 [istream::sentry], p2 says that istream::sentry ctor prepares for input if is.good() is true. p4 then goes on to say that the ctor sets the sentry::ok_ member to true if the stream state is good after any preparation. 27.6.1.2.1 [istream.formatted.reqmts], p1 then says that a formatted input function endeavors to obtain the requested input if the sentry's operator bool() returns true. Given these requirements, no formatted extractor should ever set failbit if the initial stream rdstate() == eofbit. That is contrary to the behavior of all implementations I tested. The program below prints out eof = 1, fail = 0 eof = 1, fail = 1 on all of them.


#include <sstream>
#include <cstdio>

int main()
{
    std::istringstream strm ("1");

    int i = 0;

    strm >> i;

    std::printf ("eof = %d, fail = %d\n",
                 !!strm.eof (), !!strm.fail ());

    strm >> i;

    std::printf ("eof = %d, fail = %d\n",
                 !!strm.eof (), !!strm.fail ());
}


Comments from Jerry Schwarz (c++std-lib-11373):
Jerry Schwarz wrote:
I don't know where (if anywhere) it says it in the standard, but the formatted extractors are supposed to set failbit if they don't extract any characters. If they didn't then simple loops like
while (cin >> x);
would loop forever.
Further comments from Martin Sebor:
The question is which part of the extraction should prevent this from happening by setting failbit when eofbit is already set. It could either be the sentry object or the extractor. It seems that most implementations have chosen to set failbit in the sentry [...] so that's the text that will need to be corrected.

Pre Berlin: This issue is related to 342. If the sentry sets failbit when it finds eofbit already set, then you can never seek away from the end of stream.

Kona: Possibly NAD. If eofbit is set then good() will return false. We then set ok to false. We believe that the sentry's constructor should always set failbit when ok is false, and we also think the standard already says that. Possibly it could be clearer.

Proposed resolution:

Change 27.6.1.1.3 [istream::sentry], p2 to:

explicit sentry(basic_istream<charT,traits>& is , bool noskipws = false);

-2- Effects: If is.good() is true false, calls is.setstate(failbit). Otherwise prepares for formatted or unformatted input. ...


421. is basic_streambuf copy-constructible?

Section: 27.5.2.1 [streambuf.cons] Status: Open Submitter: Martin Sebor Date: 2003-09-18

View all other issues in [streambuf.cons].

View all issues with Open status.

Discussion:

The reflector thread starting with c++std-lib-11346 notes that the class template basic_streambuf, along with basic_stringbuf and basic_filebuf, is copy-constructible but that the semantics of the copy constructors are not defined anywhere. Further, different implementations behave differently in this respect: some prevent copy construction of objects of these types by declaring their copy ctors and assignment operators private, others exhibit undefined behavior, while others still give these operations well-defined semantics.

Note that this problem doesn't seem to be isolated to just the three types mentioned above. A number of other types in the library section of the standard provide a compiler-generated copy ctor and assignment operator yet fail to specify their semantics. It's believed that the only types for which this is actually a problem (i.e. types where the compiler-generated default may be inappropriate and may not have been intended) are locale facets. See issue 439.

Proposed resolution:

27.5.2 [lib.streambuf]: Add into the synopsis, public section, just above the destructor declaration:

basic_streambuf(const basic_streambuf& sb);
basic_streambuf& operator=(const basic_streambuf& sb);

Insert after 27.5.2.1, paragraph 2:

basic_streambuf(const basic_streambuf& sb);

Constructs a copy of sb.

Postcondtions:

                eback() == sb.eback()
                gptr()  == sb.gptr()
                egptr() == sb.egptr()
                pbase() == sb.pbase()
                pptr()  == sb.pptr()
                epptr() == sb.epptr()
                getloc() == sb.getloc()
basic_streambuf& operator=(const basic_streambuf& sb);

Assigns the data members of sb to this.

Postcondtions:

                eback() == sb.eback()
                gptr()  == sb.gptr()
                egptr() == sb.egptr()
                pbase() == sb.pbase()
                pptr()  == sb.pptr()
                epptr() == sb.epptr()
                getloc() == sb.getloc()

Returns: *this.

27.7.1 [lib.stringbuf]:

Option A:

Insert into the basic_stringbuf synopsis in the private section:

basic_stringbuf(const basic_stringbuf&);             // not defined
basic_stringbuf& operator=(const basic_stringbuf&);  // not defined

Option B:

Insert into the basic_stringbuf synopsis in the public section:

basic_stringbuf(const basic_stringbuf& sb);
basic_stringbuf& operator=(const basic_stringbuf& sb);

27.7.1.1, insert after paragraph 4:

basic_stringbuf(const basic_stringbuf& sb);

Constructs an independent copy of sb as if with sb.str(), and with the openmode that sb was constructed with.

Postcondtions:

               str() == sb.str()
               gptr()  - eback() == sb.gptr()  - sb.eback()
               egptr() - eback() == sb.egptr() - sb.eback()
               pptr()  - pbase() == sb.pptr()  - sb.pbase()
               getloc() == sb.getloc()

Note: The only requirement on epptr() is that it point beyond the initialized range if an output sequence exists. There is no requirement that epptr() - pbase() == sb.epptr() - sb.pbase().

basic_stringbuf& operator=(const basic_stringbuf& sb);

After assignment the basic_stringbuf has the same state as if it were initially copy constructed from sb, except that the basic_stringbuf is allowed to retain any excess capacity it might have, which may in turn effect the value of epptr().

27.8.1.1 [lib.filebuf]

Insert at the bottom of the basic_filebuf synopsis:

private:
  basic_filebuf(const basic_filebuf&);             // not defined
  basic_filebuf& operator=(const basic_filebuf&);  // not defined

[Kona: this is an issue for basic_streambuf itself and for its derived classes. We are leaning toward allowing basic_streambuf to be copyable, and specifying its precise semantics. (Probably the obvious: copying the buffer pointers.) We are less sure whether the streambuf derived classes should be copyable. Howard will write up a proposal.]

[Sydney: Dietmar presented a new argument against basic_streambuf being copyable: it can lead to an encapsulation violation. Filebuf inherits from streambuf. Now suppose you inhert a my_hijacking_buf from streambuf. You can copy the streambuf portion of a filebuf to a my_hijacking_buf, giving you access to the pointers into the filebuf's internal buffer. Perhaps not a very strong argument, but it was strong enough to make people nervous. There was weak preference for having streambuf not be copyable. There was weak preference for having stringbuf not be copyable even if streambuf is. Move this issue to open for now. ]

[ 2007-01-12, Howard: Rvalue Reference Recommendations for Chapter 27 recommends protected copy constructor and assignment for basic_streambuf with the same semantics as would be generated by the compiler. These members aid in derived classes implementing move semantics. A protected copy constructor and copy assignment operator do not expose encapsulation more so than it is today as each data member of a basic_streambuf is already both readable and writable by derived classes via various get/set protected member functions (eback(), setp(), etc.). Rather a protected copy constructor and copy assignment operator simply make the job of derived classes implementing move semantics less tedious and error prone. ]

Rationale:

27.5.2 [lib.streambuf]: The proposed basic_streambuf copy constructor and assignment operator are the same as currently implied by the lack of declarations: public and simply copies the data members. This resolution is not a change but a clarification of the current standard.

27.7.1 [lib.stringbuf]: There are two reasonable options: A) Make basic_stringbuf not copyable. This is likely the status-quo of current implementations. B) Reasonable copy semantics of basic_stringbuf can be defined and implemented. A copyable basic_streambuf is arguably more useful than a non-copyable one. This should be considered as new functionality and not the fixing of a defect. If option B is chosen, ramifications from issue 432 are taken into account.

27.8.1.1 [lib.filebuf]: There are no reasonable copy semantics for basic_filebuf.


423. effects of negative streamsize in iostreams

Section: 27 [input.output] Status: Open Submitter: Martin Sebor Date: 2003-09-18

View all other issues in [input.output].

View all issues with Open status.

Discussion:

A third party test suite tries to exercise istream::ignore(N) with a negative value of N and expects that the implementation will treat N as if it were 0. Our implementation asserts that (N >= 0) holds and aborts the test.

I can't find anything in section 27 that prohibits such values but I don't see what the effects of such calls should be, either (this applies to a number of unformatted input functions as well as some member functions of the basic_streambuf template).

Proposed resolution:

I propose that we add to each function in clause 27 that takes an argument, say N, of type streamsize a Requires clause saying that "N >= 0." The intent is to allow negative streamsize values in calls to precision() and width() but disallow it in calls to streambuf::sgetn(), istream::ignore(), or ostream::write().

[Kona: The LWG agreed that this is probably what we want. However, we need a review to find all places where functions in clause 27 take arguments of type streamsize that shouldn't be allowed to go negative. Martin will do that review.]


424. normative notes

Section: 17.3.1.1 [structure.summary] Status: Open Submitter: Martin Sebor Date: 2003-09-18

View all issues with Open status.

Discussion:

The text in 17.3.1.1, p1 says:
"Paragraphs labelled "Note(s):" or "Example(s):" are informative, other paragraphs are normative."
The library section makes heavy use of paragraphs labeled "Notes(s)," some of which are clearly intended to be normative (see list 1), while some others are not (see list 2). There are also those where the intent is not so clear (see list 3).

List 1 -- Examples of (presumably) normative Notes:
20.6.1.1 [allocator.members], p3,
20.6.1.1 [allocator.members], p10,
21.3.2 [string.cons], p11,
22.1.1.2 [locale.cons], p11,
23.2.2.3 [deque.modifiers], p2,
25.3.7 [alg.min.max], p3,
26.3.6 [complex.ops], p15,
27.5.2.4.3 [streambuf.virt.get], p7.

List 2 -- Examples of (presumably) informative Notes:
18.5.1.3 [new.delete.placement], p3,
21.3.6.6 [string::replace], p14,
22.2.1.4.2 [locale.codecvt.virtuals], p3,
25.1.1 [alg.foreach], p4,
26.3.5 [complex.member.ops], p1,
27.4.2.5 [ios.base.storage], p6.

List 3 -- Examples of Notes that are not clearly either normative or informative:
22.1.1.2 [locale.cons], p8,
22.1.1.5 [locale.statics], p6,
27.5.2.4.5 [streambuf.virt.put], p4.

None of these lists is meant to be exhaustive.

[Definitely a real problem. The big problem is there's material that doesn't quite fit any of the named paragraph categories (e.g. Effects). Either we need a new kind of named paragraph, or we need to put more material in unnamed paragraphs jsut after the signature. We need to talk to the Project Editor about how to do this. ]

Proposed resolution:

[Pete: I changed the paragraphs marked "Note" and "Notes" to use "Remark" and "Remarks". Fixed as editorial. This change has been in the WD since the post-Redmond mailing, in 2004. Recommend NAD.]

[ Batavia: We feel that the references in List 2 above should be changed from Remarks to Notes. We also feel that those items in List 3 need to be double checked for the same change. Alan and Pete to review. ]


427. stage 2 and rationale of DR 221

Section: 22.2.2.1.2 [facet.num.get.virtuals] Status: Open Submitter: Martin Sebor Date: 2003-09-18

View other active issues in [facet.num.get.virtuals].

View all other issues in [facet.num.get.virtuals].

View all issues with Open status.

Discussion:

The requirements specified in Stage 2 and reiterated in the rationale of DR 221 (and echoed again in DR 303) specify that num_get<charT>:: do_get() compares characters on the stream against the widened elements of "012...abc...ABCX+-"

An implementation is required to allow programs to instantiate the num_get template on any charT that satisfies the requirements on a user-defined character type. These requirements do not include the ability of the character type to be equality comparable (the char_traits template must be used to perform tests for equality). Hence, the num_get template cannot be implemented to support any arbitrary character type. The num_get template must either make the assumption that the character type is equality-comparable (as some popular implementations do), or it may use char_traits<charT> to do the comparisons (some other popular implementations do that). This diversity of approaches makes it difficult to write portable programs that attempt to instantiate the num_get template on user-defined types.

[Kona: the heart of the problem is that we're theoretically supposed to use traits classes for all fundamental character operations like assignment and comparison, but facets don't have traits parameters. This is a fundamental design flaw and it appears all over the place, not just in this one place. It's not clear what the correct solution is, but a thorough review of facets and traits is in order. The LWG considered and rejected the possibility of changing numeric facets to use narrowing instead of widening. This may be a good idea for other reasons (see issue 459), but it doesn't solve the problem raised by this issue. Whether we use widen or narrow the num_get facet still has no idea which traits class the user wants to use for the comparison, because only streams, not facets, are passed traits classes. The standard does not require that two different traits classes with the same char_type must necessarily have the same behavior.]

Informally, one possibility: require that some of the basic character operations, such as eq, lt, and assign, must behave the same way for all traits classes with the same char_type. If we accept that limitation on traits classes, then the facet could reasonably be required to use char_traits<charT>.

Proposed resolution:


430. valarray subset operations

Section: 26.5.2.4 [valarray.sub] Status: Open Submitter: Martin Sebor Date: 2003-09-18

View all issues with Open status.

Discussion:

The standard fails to specify the behavior of valarray::operator[](slice) and other valarray subset operations when they are passed an "invalid" slice object, i.e., either a slice that doesn't make sense at all (e.g., slice (0, 1, 0) or one that doesn't specify a valid subset of the valarray object (e.g., slice (2, 1, 1) for a valarray of size 1).

[Kona: the LWG believes that invalid slices should invoke undefined behavior. Valarrays are supposed to be designed for high performance, so we don't want to require specific checking. We need wording to express this decision.]

Proposed resolution:


431. Swapping containers with unequal allocators

Section: 20.1.2 [allocator.requirements], 25 [algorithms] Status: Open Submitter: Matt Austern Date: 2003-09-20

View other active issues in [allocator.requirements].

View all other issues in [allocator.requirements].

View all issues with Open status.

Discussion:

Clause 20.1.2 [allocator.requirements] paragraph 4 says that implementations are permitted to supply containers that are unable to cope with allocator instances and that container implementations may assume that all instances of an allocator type compare equal. We gave implementers this latitude as a temporary hack, and eventually we want to get rid of it. What happens when we're dealing with allocators that don't compare equal?

In particular: suppose that v1 and v2 are both objects of type vector<int, my_alloc> and that v1.get_allocator() != v2.get_allocator(). What happens if we write v1.swap(v2)? Informally, three possibilities:

1. This operation is illegal. Perhaps we could say that an implementation is required to check and to throw an exception, or perhaps we could say it's undefined behavior.

2. The operation performs a slow swap (i.e. using three invocations of operator=, leaving each allocator with its original container. This would be an O(N) operation.

3. The operation swaps both the vectors' contents and their allocators. This would be an O(1) operation. That is:

    my_alloc a1(...);
    my_alloc a2(...);
    assert(a1 != a2);

    vector<int, my_alloc> v1(a1);
    vector<int, my_alloc> v2(a2);
    assert(a1 == v1.get_allocator());
    assert(a2 == v2.get_allocator());

    v1.swap(v2);
    assert(a1 == v2.get_allocator());
    assert(a2 == v1.get_allocator());
  

[Kona: This is part of a general problem. We need a paper saying how to deal with unequal allocators in general.]

[pre-Sydney: Howard argues for option 3 in N1599. ]

[ 2007-01-12, Howard: This issue will now tend to come up more often with move constructors and move assignment operators. For containers, these members transfer resources (i.e. the allocated memory) just like swap. ]

[ Batavia: There is agreement to overload the container swap on the allocator's Swappable requirement using concepts. If the allocator supports Swappable, then container's swap will swap allocators, else it will perform a "slow swap" using copy construction and copy assignment. ]

Proposed resolution:


446. Iterator equality between different containers

Section: 24.1 [iterator.requirements], 23.1 [container.requirements] Status: Open Submitter: Andy Koenig Date: 2003-12-16

View other active issues in [iterator.requirements].

View all other issues in [iterator.requirements].

View all issues with Open status.

Discussion:

What requirements does the standard place on equality comparisons between iterators that refer to elements of different containers. For example, if v1 and v2 are empty vectors, is v1.end() == v2.end() allowed to yield true? Is it allowed to throw an exception?

The standard appears to be silent on both questions.

[Sydney: The intention is that comparing two iterators from different containers is undefined, but it's not clear if we say that, or even whether it's something we should be saying in clause 23 or in clause 24. Intuitively we might want to say that equality is defined only if one iterator is reachable from another, but figuring out how to say it in any sensible way is a bit tricky: reachability is defined in terms of equality, so we can't also define equality in terms of reachability. ]

Proposed resolution:


454. basic_filebuf::open should accept wchar_t names

Section: 27.8.1.4 [filebuf.members] Status: Open Submitter: Bill Plauger Date: 2004-01-30

View other active issues in [filebuf.members].

View all other issues in [filebuf.members].

View all issues with Open status.

Discussion:

    basic_filebuf *basic_filebuf::open(const char *, ios_base::open_mode);

should be supplemented with the overload:

    basic_filebuf *basic_filebuf::open(const wchar_t *, ios_base::open_mode);

Depending on the operating system, one of these forms is fundamental and the other requires an implementation-defined mapping to determine the actual filename.

[Sydney: Yes, we want to allow wchar_t filenames. Bill will provide wording.]

[ In Toronto we noted that this is issue 5 from N1569. ]

How does this interact with the newly-defined character types, and how do we avoid interface explosion considering std::string overloads that were added? Propose another solution that is different than the suggestion proposed by PJP.

Suggestion is to make a member template function for basic_string (for char, wchar_t, u16char, u32char instantiations), and then just keep a const char* member.

Goal is to do implicit conversion between character string literals to appropriate basic_string type. Not quite sure if this is possible.

Implementors are free to add specific overloads for non-char character types.

Proposed resolution:

Change from:

basic_filebuf<charT,traits>* open(
	const char* s,
	ios_base::openmode mode );

Effects: If is_open() != false, returns a null pointer. Otherwise, initializes the filebuf as required. It then opens a file, if possible, whose name is the NTBS s ("as if" by calling std::fopen(s,modstr)).

to:

basic_filebuf<charT,traits>* open(
	const char* s,
	ios_base::openmode mode );

basic_filebuf<charT,traits>* open(
	const wchar_t* ws,
	ios_base::openmode mode );

Effects: If is_open() != false, returns a null pointer. Otherwise, initializes the filebuf as required. It then opens a file, if possible, whose name is the NTBS s ("as if" by calling std::fopen(s,modstr)). For the second signature, the NTBS s is determined from the WCBS ws in an implementation-defined manner.

(NOTE: For a system that "naturally" represents a filename as a WCBS, the NTBS s in the first signature may instead be mapped to a WCBS; if so, it follows the same mapping rules as the first argument to open.)

Rationale:

Slightly controversial, but by a 7-1 straw poll the LWG agreed to move this to Ready. The controversy was because the mapping between wide names and files in a filesystem is implementation defined. The counterargument, which most but not all LWG members accepted, is that the mapping between narrow files names and files is also implemenation defined.

[Lillehammer: Moved back to "open" status, at Beman's urging. (1) Why just basic_filebuf, instead of also basic_fstream (and possibly other things too). (2) Why not also constructors that take std::basic_string? (3) We might want to wait until we see Beman's filesystem library; we might decide that it obviates this.]


458. 24.1.5 contains unintented limitation for operator-

Section: 24.1.5 [random.access.iterators] Status: Open Submitter: Daniel Frey Date: 2004-02-27

View all other issues in [random.access.iterators].

View all issues with Open status.

Discussion:

In 24.1.5 [lib.random.access.iterators], table 76 the operational semantics for the expression "r -= n" are defined as "return r += -n". This means, that the expression -n must be valid, which is not the case for unsigned types.

[ Sydney: Possibly not a real problem, since difference type is required to be a signed integer type. However, the wording in the standard may be less clear than we would like. ]

Proposed resolution:

To remove this limitation, I suggest to change the operational semantics for this column to:

    { Distance m = n; 
      if (m >= 0) 
        while (m--) --r; 
      else 
        while (m++) ++r;
      return r; }

459. Requirement for widening in stage 2 is overspecification

Section: 22.2.2.1.2 [facet.num.get.virtuals] Status: Open Submitter: Martin Sebor Date: 2004-03-16

View other active issues in [facet.num.get.virtuals].

View all other issues in [facet.num.get.virtuals].

View all issues with Open status.

Discussion:

When parsing strings of wide-character digits, the standard requires the library to widen narrow-character "atoms" and compare the widened atoms against the characters that are being parsed. Simply narrowing the wide characters would be far simpler, and probably more efficient. The two choices are equivalent except in convoluted test cases, and many implementations already ignore the standard and use narrow instead of widen.

First, I disagree that using narrow() instead of widen() would necessarily have unfortunate performance implications. A possible implementation of narrow() that allows num_get to be implemented in a much simpler and arguably comparably efficient way as calling widen() allows, i.e. without making a virtual call to do_narrow every time, is as follows:

  inline char ctype<wchar_t>::narrow (wchar_t wc, char dflt) const
  {
      const unsigned wi = unsigned (wc);

      if (wi > UCHAR_MAX)
          return typeid (*this) == typeid (ctype<wchar_t>) ?
                 dflt : do_narrow (wc, dflt);

      if (narrow_ [wi] < 0) {
         const char nc = do_narrow (wc, dflt);
         if (nc == dflt)
             return dflt;
         narrow_ [wi] = nc;
      }

      return char (narrow_ [wi]);
  }

Second, I don't think the change proposed in the issue (i.e., to use narrow() instead of widen() during Stage 2) would be at all drastic. Existing implementations with the exception of libstdc++ currently already use narrow() so the impact of the change on programs would presumably be isolated to just a single implementation. Further, since narrow() is not required to translate alternate wide digit representations such as those mentioned in issue 303 to their narrow equivalents (i.e., the portable source characters '0' through '9'), the change does not necessarily imply that these alternate digits would be treated as ordinary digits and accepted as part of numbers during parsing. In fact, the requirement in 22.2.1.1.2 [locale.ctype.virtuals], p13 forbids narrow() to translate an alternate digit character, wc, to an ordinary digit in the basic source character set unless the expression (ctype<charT>::is(ctype_base::digit, wc) == true) holds. This in turn is prohibited by the C standard (7.25.2.1.5, 7.25.2.1.5, and 5.2.1, respectively) for charT of either char or wchar_t.

[Sydney: To a large extent this is a nonproblem. As long as you're only trafficking in char and wchar_t we're only dealing with a stable character set, so you don't really need either 'widen' or 'narrow': can just use literals. Finally, it's not even clear whether widen-vs-narrow is the right question; arguably we should be using codecvt instead.]

Proposed resolution:

Change stage 2 so that implementations are permitted to use either technique to perform the comparison:

  1. call widen on the atoms and compare (either by using operator== or char_traits<charT>::eq) the input with the widened atoms, or
  2. call narrow on the input and compare the narrow input with the atoms
  3. do (1) or (2) only if charT is not char or wchar_t, respectively; i.e., avoid calling widen or narrow if it the source and destination types are the same

462. Destroying objects with static storage duration

Section: 3.6.3 [basic.start.term], 18.3 [cstdint] Status: Open Submitter: Bill Plauger Date: 2004-03-23

View all issues with Open status.

Discussion:

3.6.3 Termination spells out in detail the interleaving of static destructor calls and calls to functions registered with atexit. To match this behavior requires intimate cooperation between the code that calls destructors and the exit/atexit machinery. The former is tied tightly to the compiler; the latter is a primitive mechanism inherited from C that traditionally has nothing to do with static construction and destruction. The benefits of intermixing destructor calls with atexit handler calls is questionable at best, and very difficult to get right, particularly when mixing third-party C++ libraries with different third-party C++ compilers and C libraries supplied by still other parties.

I believe the right thing to do is defer all static destruction until after all atexit handlers are called. This is a change in behavior, but one that is likely visible only to perverse test suites. At the very least, we should permit deferred destruction even if we don't require it.

[If this is to be changed, it should probably be changed by CWG. At this point, however, the LWG is leaning toward NAD. Implementing what the standard says is hard work, but it's not impossible and most vendors went through that pain years ago. Changing this behavior would be a user-visible change, and would break at least one real application.]

[ Batavia: Send to core with our recommendation that we should permit deferred destruction but not require it. ]

[ Howard: The course of action recommended in Batavia would undo LWG issue 3 and break current code implementing the "phoenix singleton". Search the net for "phoenix singleton atexit" to get a feel for the size of the adverse impact this change would have. Below is sample code which implements the phoenix singleton and would break if atexit is changed in this way: ]

#include <cstdlib>
#include <iostream>
#include <type_traits>
#include <new>

class A
{
    bool alive_;
    A(const A&);
    A& operator=(const A&);
public:
    A() : alive_(true) {std::cout << "A()\n";}
    ~A() {alive_ = false; std::cout << "~A()\n";}
    void use()
    {
        if (alive_)
            std::cout << "A is alive\n";
        else
            std::cout << "A is dead\n";
    }
};

void deallocate_resource();

// This is the phoenix singleton pattern
A& get_resource(bool create = true)
{
    static std::aligned_storage<sizeof(A), std::alignment_of<A>::value>::type buf;
    static A* a;
    if (create)
    {
        if (a != (A*)&buf)
        {
            a = ::new (&buf) A;
            std::atexit(deallocate_resource);
        }
    }
    else
    {
        a->~A();
        a = (A*)&buf + 1;
    }
    return *a;
}

void deallocate_resource()
{
    get_resource(false);
}

void use_A(const char* message)
{
    A& a = get_resource();
    std::cout << "Using A " << message << "\n";
    a.use();
}

struct B
{
    ~B() {use_A("from ~B()");}
};

B b;

int main()
{
    use_A("from main()");
}

The correct output is:

A()
Using A from main()
A is alive
~A()
A()
Using A from ~B()
A is alive
~A()

Proposed resolution:


463. auto_ptr usability issues

Section: D.9.1 [auto.ptr] Status: Open Submitter: Rani Sharoni Date: 2003-12-07

View all other issues in [auto.ptr].

View all issues with Open status.

Discussion:

TC1 CWG DR #84 effectively made the template<class Y> operator auto_ptr<Y>() member of auto_ptr (20.4.5.3/4) obsolete.

The sole purpose of this obsolete conversion member is to enable copy initialization base from r-value derived (or any convertible types like cv-types) case:

#include <memory>
using std::auto_ptr;

struct B {};
struct D : B {};

auto_ptr<D> source();
int sink(auto_ptr<B>);
int x1 = sink( source() ); // #1 EDG - no suitable copy constructor

The excellent analysis of conversion operations that was given in the final auto_ptr proposal (http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/papers/1997/N1128.pdf) explicitly specifies this case analysis (case 4). DR #84 makes the analysis wrong and actually comes to forbid the loophole that was exploited by the auto_ptr designers.

I didn't encounter any compliant compiler (e.g. EDG, GCC, BCC and VC) that ever allowed this case. This is probably because it requires 3 user defined conversions and in fact current compilers conform to DR #84.

I was surprised to discover that the obsolete conversion member actually has negative impact of the copy initialization base from l-value derived case:

auto_ptr<D> dp;
int x2 = sink(dp); // #2 EDG - more than one user-defined conversion applies

I'm sure that the original intention was allowing this initialization using the template<class Y> auto_ptr(auto_ptr<Y>& a) constructor (20.4.5.1/4) but since in this copy initialization it's merely user defined conversion (UDC) and the obsolete conversion member is UDC with the same rank (for the early overloading stage) there is an ambiguity between them.

Removing the obsolete member will have impact on code that explicitly invokes it:

int y = sink(source().operator auto_ptr<B>());

IMHO no one ever wrote such awkward code and the reasonable workaround for #1 is:

int y = sink( auto_ptr<B>(source()) );

I was even more surprised to find out that after removing the obsolete conversion member the initialization was still ill-formed: int x3 = sink(dp); // #3 EDG - no suitable copy constructor

This copy initialization semantically requires copy constructor which means that both template conversion constructor and the auto_ptr_ref conversion member (20.4.5.3/3) are required which is what was explicitly forbidden in DR #84. This is a bit amusing case in which removing ambiguity results with no candidates.

I also found exception safety issue with auto_ptr related to auto_ptr_ref:

int f(auto_ptr<B>, std::string);
auto_ptr<B> source2();

// string constructor throws while auto_ptr_ref
// "holds" the pointer
int x4 = f(source2(), "xyz"); // #4

The theoretic execution sequence that will cause a leak:

  1. call auto_ptr<B>::operator auto_ptr_ref<B>()
  2. call string::string(char const*) and throw

According to 20.4.5.3/3 and 20.4.5/2 the auto_ptr_ref conversion member returns auto_ptr_ref<Y> that holds *this and this is another defect since the type of *this is auto_ptr<X> where X might be different from Y. Several library vendors (e.g. SGI) implement auto_ptr_ref<Y> with Y* as member which is much more reasonable. Other vendor implemented auto_ptr_ref as defectively required and it results with awkward and catastrophic code: int oops = sink(auto_ptr<B>(source())); // warning recursive on all control paths

Dave Abrahams noticed that there is no specification saying that auto_ptr_ref copy constructor can't throw.

My proposal comes to solve all the above issues and significantly simplify auto_ptr implementation. One of the fundamental requirements from auto_ptr is that it can be constructed in an intuitive manner (i.e. like ordinary pointers) but with strict ownership semantics which yield that source auto_ptr in initialization must be non-const. My idea is to add additional constructor template with sole propose to generate ill-formed, diagnostic required, instance for const auto_ptr arguments during instantiation of declaration. This special constructor will not be instantiated for other types which is achievable using 14.8.2/2 (SFINAE). Having this constructor in hand makes the constructor template<class Y> auto_ptr(auto_ptr<Y> const&) legitimate since the actual argument can't be const yet non const r-value are acceptable.

This implementation technique makes the "private auxiliary class" auto_ptr_ref obsolete and I found out that modern C++ compilers (e.g. EDG, GCC and VC) consume the new implementation as expected and allow all intuitive initialization and assignment cases while rejecting illegal cases that involve const auto_ptr arguments.

The proposed auto_ptr interface:

namespace std {
    template<class X> class auto_ptr {
    public:
        typedef X element_type;

        // 20.4.5.1 construct/copy/destroy:
        explicit auto_ptr(X* p=0) throw();
        auto_ptr(auto_ptr&) throw();
        template<class Y> auto_ptr(auto_ptr<Y> const&) throw();
        auto_ptr& operator=(auto_ptr&) throw();
        template<class Y> auto_ptr& operator=(auto_ptr<Y>) throw();
        ~auto_ptr() throw();

        // 20.4.5.2 members:
        X& operator*() const throw();
        X* operator->() const throw();
        X* get() const throw();
        X* release() throw();
        void reset(X* p=0) throw();

    private:
        template<class U>
        auto_ptr(U& rhs, typename
unspecified_error_on_const_auto_ptr<U>::type = 0);
    };
}

One compliant technique to implement the unspecified_error_on_const_auto_ptr helper class is using additional private auto_ptr member class template like the following:

template<typename T> struct unspecified_error_on_const_auto_ptr;

template<typename T>
struct unspecified_error_on_const_auto_ptr<auto_ptr<T> const>
{ typedef typename auto_ptr<T>::const_auto_ptr_is_not_allowed type; };

There are other techniques to implement this helper class that might work better for different compliers (i.e. better diagnostics) and therefore I suggest defining its semantic behavior without mandating any specific implementation. IMO, and I didn't found any compiler that thinks otherwise, 14.7.1/5 doesn't theoretically defeat the suggested technique but I suggest verifying this with core language experts.

Further changes in standard text:

Remove section 20.4.5.3

Change 20.4.5/2 to read something like: Initializing auto_ptr<X> from const auto_ptr<Y> will result with unspecified ill-formed declaration that will require unspecified diagnostic.

Change 20.4.5.1/4,5,6 to read:

template<class Y> auto_ptr(auto_ptr<Y> const& a) throw();

4 Requires: Y* can be implicitly converted to X*.

5 Effects: Calls const_cast<auto_ptr<Y>&>(a).release().

6 Postconditions: *this holds the pointer returned from a.release().

Change 20.4.5.1/10

template<class Y> auto_ptr& operator=(auto_ptr<Y> a) throw();

10 Requires: Y* can be implicitly converted to X*. The expression delete get() is well formed.

LWG TC DR #127 is obsolete.

Notice that the copy constructor and copy assignment operator should remain as before and accept non-const auto_ptr& since they have effect on the form of the implicitly declared copy constructor and copy assignment operator of class that contains auto_ptr as member per 12.8/5,10:

struct X {
    // implicit X(X&)
    // implicit X& operator=(X&)
    auto_ptr<D> aptr_;
};

In most cases this indicates about sloppy programming but preserves the current auto_ptr behavior.

Dave Abrahams encouraged me to suggest fallback implementation in case that my suggestion that involves removing of auto_ptr_ref will not be accepted. In this case removing the obsolete conversion member to auto_ptr<Y> and 20.4.5.3/4,5 is still required in order to eliminate ambiguity in legal cases. The two constructors that I suggested will co exist with the current members but will make auto_ptr_ref obsolete in initialization contexts. auto_ptr_ref will be effective in assignment contexts as suggested in DR #127 and I can't see any serious exception safety issues in those cases (although it's possible to synthesize such). auto_ptr_ref<X> semantics will have to be revised to say that it strictly holds pointer of type X and not reference to an auto_ptr for the favor of cases in which auto_ptr_ref<Y> is constructed from auto_ptr<X> in which X is different from Y (i.e. assignment from r-value derived to base).

[Redmond: punt for the moment. We haven't decided yet whether we want to fix auto_ptr for C++-0x, or remove it and replace it with move_ptr and unique_ptr.]

[ Oxford 2007: Recommend NAD. We're just going to deprecate it. It still works for simple use cases and people know how to deal with it. Going forward unique_ptr is the recommended tool. ]

[ 2007-11-09: Reopened at the request of David Abrahams, Alisdair Meredith and Gabriel Dos Reis. ]

Proposed resolution:

Change the synopsis in D.9.1 [auto.ptr]:

namespace std { 
  template <class Y> struct auto_ptr_ref {};

  // exposition only
  template <class T> struct constant_object;

  // exposition only
  template <class T>
  struct cannot_transfer_ownership_from
    : constant_object<T> {};

  template <class X> class auto_ptr { 
  public: 
    typedef X element_type; 

    // D.9.1.1 construct/copy/destroy: 
    explicit auto_ptr(X* p =0) throw(); 
    auto_ptr(auto_ptr&) throw(); 
    template<class Y> auto_ptr(auto_ptr<Y> const&) throw(); 
    auto_ptr& operator=(auto_ptr&) throw(); 
    template<class Y> auto_ptr& operator=(auto_ptr<Y>&) throw();
    auto_ptr& operator=(auto_ptr_ref<X> r) throw();
    ~auto_ptr() throw(); 

    // D.9.1.2 members: 
    X& operator*() const throw();
    X* operator->() const throw();
    X* get() const throw();
    X* release() throw();
    void reset(X* p =0) throw();

    // D.9.1.3 conversions:
    auto_ptr(auto_ptr_ref<X>) throw();
    template<class Y> operator auto_ptr_ref<Y>() throw();
    template<class Y> operator auto_ptr<Y>() throw();

    // exposition only
    template<class U>
    auto_ptr(U& rhs, typename cannot_transfer_ownership_from<U>::error = 0);
  }; 

  template <> class auto_ptr<void> 
  { 
  public: 
    typedef void element_type; 
  }; 

}

Remove D.9.1.3 [auto.ptr.conv].

Change D.9.1 [auto.ptr], p3:

The auto_ptr provides a semantics of strict ownership. An auto_ptr owns the object it holds a pointer to. Copying an auto_ptr copies the pointer and transfers ownership to the destination. If more than one auto_ptr owns the same object at the same time the behavior of the program is undefined. Templates constant_object and cannot_transfer_ownership_from, and the final constructor of auto_ptr are for exposition only. For any types X and Y, initializing auto_ptr<X> from const auto_ptr<Y> is ill-formed, diagnostic required. [Note: The uses of auto_ptr include providing temporary exception-safety for dynamically allocated memory, passing ownership of dynamically allocated memory to a function, and returning dynamically allocated memory from a function. auto_ptr does not meet the CopyConstructible and Assignable requirements for Standard Library container elements and thus instantiating a Standard Library container with an auto_ptr results in undefined behavior. -- end note]

Change D.9.1.1 [auto.ptr.cons], p5:

template<class Y> auto_ptr(auto_ptr<Y> const& a) throw();

Requires: Y* can be implicitly converted to X*.

Effects: Calls const_cast<auto_ptr<Y>&>(a).release().

Postconditions: *this holds the pointer returned from a.release().

Change D.9.1.1 [auto.ptr.cons], p10:

template<class Y> auto_ptr& operator=(auto_ptr<Y>& a) throw();

Requires: Y* can be implicitly converted to X*. The expression delete get() is well formed.

Effects: Calls reset(a.release()).

Returns: *this.


471. result of what() implementation-defined

Section: 18.6.1 [type.info] Status: Open Submitter: Martin Sebor Date: 2004-06-28

View all other issues in [type.info].

View all issues with Open status.

Discussion:

[lib.exception] specifies the following:

    exception (const exception&) throw();
    exception& operator= (const exception&) throw();

    -4- Effects: Copies an exception object.
    -5- Notes: The effects of calling what() after assignment
        are implementation-defined.

First, does the Note only apply to the assignment operator? If so, what are the effects of calling what() on a copy of an object? Is the returned pointer supposed to point to an identical copy of the NTBS returned by what() called on the original object or not?

Second, is this Note intended to extend to all the derived classes in section 19? I.e., does the standard provide any guarantee for the effects of what() called on a copy of any of the derived class described in section 19?

Finally, if the answer to the first question is no, I believe it constitutes a defect since throwing an exception object typically implies invoking the copy ctor on the object. If the answer is yes, then I believe the standard ought to be clarified to spell out exactly what the effects are on the copy (i.e., after the copy ctor was called).

[Redmond: Yes, this is fuzzy. The issue of derived classes is fuzzy too.]

[ Batavia: Howard provided wording. ]

Proposed resolution:

Change 18.7.1 [exception] to:

exception(const exception& e) throw();
exception& operator=(const exception& e) throw();

-4- Effects: Copies an exception object.

-5- Remarks: The effects of calling what() after assignment are implementation-defined.

-5- Throws: Nothing. This also applies to all standard library-defined classes that derive from exception.

-7- Postcondition: strcmp(what(), e.what()) == 0. This also applies to all standard library-defined classes that derive from exception.


473. underspecified ctype calls

Section: 22.2.1.1 [locale.ctype] Status: Open Submitter: Martin Sebor Date: 2004-07-01

View all issues with Open status.

Discussion:

Most ctype member functions come in two forms: one that operates on a single character at a time and another form that operates on a range of characters. Both forms are typically described by a single Effects and/or Returns clause.

The Returns clause of each of the single-character non-virtual forms suggests that the function calls the corresponding single character virtual function, and that the array form calls the corresponding virtual array form. Neither of the two forms of each virtual member function is required to be implemented in terms of the other.

There are three problems:

1. One is that while the standard does suggest that each non-virtual member function calls the corresponding form of the virtual function, it doesn't actually explicitly require it.

Implementations that cache results from some of the virtual member functions for some or all values of their arguments might want to call the array form from the non-array form the first time to fill the cache and avoid any or most subsequent virtual calls. Programs that rely on each form of the virtual function being called from the corresponding non-virtual function will see unexpected behavior when using such implementations.

2. The second problem is that either form of each of the virtual functions can be overridden by a user-defined function in a derived class to return a value that is different from the one produced by the virtual function of the alternate form that has not been overriden.

Thus, it might be possible for, say, ctype::widen(c) to return one value, while for ctype::widen(&c, &c + 1, &wc) to set wc to another value. This is almost certainly not intended. Both forms of every function should be required to return the same result for the same character, otherwise the same program using an implementation that calls one form of the functions will behave differently than when using another implementation that calls the other form of the function "under the hood."

3. The last problem is that the standard text fails to specify whether one form of any of the virtual functions is permitted to be implemented in terms of the other form or not, and if so, whether it is required or permitted to call the overridden virtual function or not.

Thus, a program that overrides one of the virtual functions so that it calls the other form which then calls the base member might end up in an infinite loop if the called form of the base implementation of the function in turn calls the other form.

Lillehammer: Part of this isn't a real problem. We already talk about caching. 22.1.1/6 But part is a real problem. ctype virtuals may call each other, so users don't know which ones to override to avoid avoid infinite loops.

This is a problem for all facet virtuals, not just ctype virtuals, so we probably want a blanket statement in clause 22 for all facets. The LWG is leaning toward a blanket prohibition, that a facet's virtuals may never call each other. We might want to do that in clause 27 too, for that matter. A review is necessary. Bill will provide wording.

Proposed resolution:


484. Convertible to T

Section: 24.1.1 [input.iterators] Status: Open Submitter: Chris Jefferson Date: 2004-09-16

View all other issues in [input.iterators].

View all issues with Open status.

Discussion:

From comp.std.c++:

I note that given an input iterator a for type T, then *a only has to be "convertable to T", not actually of type T.

Firstly, I can't seem to find an exact definition of "convertable to T". While I assume it is the obvious definition (an implicit conversion), I can't find an exact definition. Is there one?

Slightly more worryingly, there doesn't seem to be any restriction on the this type, other than it is "convertable to T". Consider two input iterators a and b. I would personally assume that most people would expect *a==*b would perform T(*a)==T(*b), however it doesn't seem that the standard requires that, and that whatever type *a is (call it U) could have == defined on it with totally different symantics and still be a valid inputer iterator.

Is this a correct reading? When using input iterators should I write T(*a) all over the place to be sure that the object i'm using is the class I expect?

This is especially a nuisance for operations that are defined to be "convertible to bool". (This is probably allowed so that implementations could return say an int and avoid an unnessary conversion. However all implementations I have seen simply return a bool anyway. Typical implemtations of STL algorithms just write things like while(a!=b && *a!=0). But strictly speaking, there are lots of types that are convertible to T but that also overload the appropriate operators so this doesn't behave as expected.

If we want to make code like this legal (which most people seem to expect), then we'll need to tighten up what we mean by "convertible to T".

[Lillehammer: The first part is NAD, since "convertible" is well-defined in core. The second part is basically about pathological overloads. It's a minor problem but a real one. So leave open for now, hope we solve it as part of iterator redesign.]

Proposed resolution:


485. output iterator insufficently constrained

Section: 24.1.2 [output.iterators] Status: Open Submitter: Chris Jefferson Date: 2004-10-13

View all other issues in [output.iterators].

View all issues with Open status.

Discussion:

The note on 24.1.2 Output iterators insufficently limits what can be performed on output iterators. While it requires that each iterator is progressed through only once and that each iterator is written to only once, it does not require the following things:

Note: Here it is assumed that x is an output iterator of type X which has not yet been assigned to.

a) That each value of the output iterator is written to: The standard allows: ++x; ++x; ++x;

b) That assignments to the output iterator are made in order X a(x); ++a; *a=1; *x=2; is allowed

c) Chains of output iterators cannot be constructed: X a(x); ++a; X b(a); ++b; X c(b); ++c; is allowed, and under the current wording (I believe) x,a,b,c could be written to in any order.

I do not believe this was the intension of the standard?

[Lillehammer: Real issue. There are lots of constraints we intended but didn't specify. Should be solved as part of iterator redesign.]

Proposed resolution:


492. Invalid iterator arithmetic expressions

Section: 23 [containers], 24 [iterators], 25 [algorithms] Status: Open Submitter: Thomas Mang Date: 2004-12-12

View all other issues in [containers].

View all issues with Open status.

Discussion:

Various clauses other than clause 25 make use of iterator arithmetic not supported by the iterator category in question. Algorithms in clause 25 are exceptional because of 25 [lib.algorithms], paragraph 9, but this paragraph does not provide semantics to the expression "iterator - n", where n denotes a value of a distance type between iterators.

1) Examples of current wording:

Current wording outside clause 25:

23.2.2.4 [lib.list.ops], paragraphs 19-21: "first + 1", "(i - 1)", "(last - first)" 23.3.1.1 [lib.map.cons], paragraph 4: "last - first" 23.3.2.1 [lib.multimap.cons], paragraph 4: "last - first" 23.3.3.1 [lib.set.cons], paragraph 4: "last - first" 23.3.4.1 [lib.multiset.cons], paragraph 4: "last - first" 24.4.1 [lib.reverse.iterators], paragraph 1: "(i - 1)"

[Important note: The list is not complete, just an illustration. The same issue might well apply to other paragraphs not listed here.]

None of these expressions is valid for the corresponding iterator category.

Current wording in clause 25:

25.1.1 [lib.alg.foreach], paragraph 1: "last - 1" 25.1.3 [lib.alg.find.end], paragraph 2: "[first1, last1 - (last2-first2))" 25.2.8 [lib.alg.unique], paragraph 1: "(i - 1)" 25.2.8 [lib.alg.unique], paragraph 5: "(i - 1)"

However, current wording of 25 [lib.algorithms], paragraph 9 covers neither of these four cases:

Current wording of 25 [lib.algorithms], paragraph 9:

"In the description of the algorithms operator + 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+n is the same as that of

{X tmp = a;
advance(tmp, n);
return tmp;
}

and that of b-a is the same as of return distance(a, b)"

This paragrpah does not take the expression "iterator - n" into account, where n denotes a value of a distance type between two iterators [Note: According to current wording, the expression "iterator - n" would be resolved as equivalent to "return distance(n, iterator)"]. Even if the expression "iterator - n" were to be reinterpreted as equivalent to "iterator + -n" [Note: This would imply that "a" and "b" were interpreted implicitly as values of iterator types, and "n" as value of a distance type], then 24.3.4/2 interfers because it says: "Requires: n may be negative only for random access and bidirectional iterators.", and none of the paragraphs quoted above requires the iterators on which the algorithms operate to be of random access or bidirectional category.

2) Description of intended behavior:

For the rest of this Defect Report, it is assumed that the expression "iterator1 + n" and "iterator1 - iterator2" has the semantics as described in current 25 [lib.algorithms], paragraph 9, but applying to all clauses. The expression "iterator1 - n" is equivalent to an result-iterator for which the expression "result-iterator + n" yields an iterator denoting the same position as iterator1 does. The terms "iterator1", "iterator2" and "result-iterator" shall denote the value of an iterator type, and the term "n" shall denote a value of a distance type between two iterators.

All implementations known to the author of this Defect Report comply with these assumptions. No impact on current code is expected.

3) Proposed fixes:

Change 25 [lib.algorithms], paragraph 9 to:

"In the description of the algorithms operator + and - are used for some of the iterator categories for which they do not have to be defined. In this paragraph, a and b denote values of an iterator type, and n denotes a value of a distance type between two iterators. In these cases the semantics of a+n is the same as that of

{X tmp = a;
advance(tmp, n);
return tmp;
}

,the semantics of a-n denotes the value of an iterator i for which the following condition holds: advance(i, n) == a, and that of b-a is the same as of return distance(a, b)".

Comments to the new wording:

a) The wording " In this paragraph, a and b denote values of an iterator type, and n denotes a value of a distance type between two iterators." was added so the expressions "b-a" and "a-n" are distinguished regarding the types of the values on which they operate. b) The wording ",the semantics of a-n denotes the value of an iterator i for which the following condition holds: advance(i, n) == a" was added to cover the expression 'iterator - n'. The wording "advance(i, n) == a" was used to avoid a dependency on the semantics of a+n, as the wording "i + n == a" would have implied. However, such a dependency might well be deserved. c) DR 225 is not considered in the new wording.

Proposed fixes regarding invalid iterator arithmetic expressions outside clause 25:

Either a) Move modified 25 [lib.algorithms], paragraph 9 (as proposed above) before any current invalid iterator arithmetic expression. In that case, the first sentence of 25 [lib.algorithms], paragraph 9, need also to be modified and could read: "For the rest of this International Standard, ...." / "In the description of the following clauses including this ...." / "In the description of the text below ..." etc. - anyways substituting the wording "algorithms", which is a straight reference to clause 25. In that case, 25 [lib.algorithms] paragraph 9 will certainly become obsolete. Alternatively, b) Add an appropiate paragraph similar to resolved 25 [lib.algorithms], paragraph 9, to the beginning of each clause containing invalid iterator arithmetic expressions. Alternatively, c) Fix each paragraph (both current wording and possible resolutions of DRs) containing invalid iterator arithmetic expressions separately.

5) References to other DRs:

See DR 225. See DR 237. The resolution could then also read "Linear in last - first".

Proposed resolution:

[Lillehammer: Minor issue, but real. We have a blanket statement about this in 25/11. But (a) it should be in 17, not 25; and (b) it's not quite broad enough, because there are some arithmetic expressions it doesn't cover. Bill will provide wording.]


498. Requirements for partition() and stable_partition() too strong

Section: 25.2.13 [alg.partitions] Status: Open Submitter: Sean Parent, Joe Gottman Date: 2005-05-04

View all issues with Open status.

Discussion:

Problem: The iterator requirements for partition() and stable_partition() [25.2.12] are listed as BidirectionalIterator, however, there are efficient algorithms for these functions that only require ForwardIterator that have been known since before the standard existed. The SGI implementation includes these (see http://www.sgi.com/tech/stl/partition.html and http://www.sgi.com/tech/stl/stable_partition.html).

Proposed resolution:

Change 25.2.12 from

template<class BidirectionalIterator, class Predicate> 
BidirectionalIterator partition(BidirectionalIterato r first, 
                                BidirectionalIterator last, 
                                Predicate pred); 

to

template<class ForwardIterator, class Predicate> 
ForwardIterator partition(ForwardIterator first, 
                          ForwardIterator last, 
                          Predicate pred); 

Change the complexity from

At most (last - first)/2 swaps are done. Exactly (last - first) applications of the predicate are done.

to

If ForwardIterator is a bidirectional_iterator, at most (last - first)/2 swaps are done; otherwise at most (last - first) swaps are done. Exactly (last - first) applications of the predicate are done.

Rationale:

Partition is a "foundation" algorithm useful in many contexts (like sorting as just one example) - my motivation for extending it to include forward iterators is slist - without this extension you can't partition an slist (without writing your own partition). Holes like this in the standard library weaken the argument for generic programming (ideally I'd be able to provide a library that would refine std::partition() to other concepts without fear of conflicting with other libraries doing the same - but that is a digression). I consider the fact that partition isn't defined to work for ForwardIterator a minor embarrassment.

[Mont Tremblant: Moved to Open, request motivation and use cases by next meeting. Sean provided further rationale by post-meeting mailing.]


502. Proposition: Clarification of the interaction between a facet and an iterator

Section: 22.1.1.1.1 [locale.category] Status: Open Submitter: Christopher Conrade Zseleghovski Date: 2005-06-07

View all other issues in [locale.category].

View all issues with Open status.

Discussion:

Motivation:

This requirement seems obvious to me, it is the essence of code modularity. I have complained to Mr. Plauger that the Dinkumware library does not observe this principle but he objected that this behaviour is not covered in the standard.

Proposed resolution:

Append the following point to 22.1.1.1.1:

6. The implementation of a facet of Table 52 parametrized with an InputIterator/OutputIterator should use that iterator only as character source/sink respectively. For a *_get facet, it means that the value received depends only on the sequence of input characters and not on how they are accessed. For a *_put facet, it means that the sequence of characters output depends only on the value to be formatted and not of how the characters are stored.

[ Berlin: Moved to Open, Need to clean up this area to make it clear locales don't have to contain open ended sets of facets. Jack, Howard, Bill. ]


503. more on locales

Section: 22.2 [locale.categories] Status: Open Submitter: P.J. Plauger Date: 2005-06-20

View other active issues in [locale.categories].

View all other issues in [locale.categories].

View all issues with Open status.

Discussion:

a) In 22.2.1.1 para. 2 we refer to "the instantiations required in Table 51" to refer to the facet *objects* associated with a locale. And we almost certainly mean just those associated with the default or "C" locale. Otherwise, you can't switch to a locale that enforces a different mapping between narrow and wide characters, or that defines additional uppercase characters.

b) 22.2.1.5 para. 3 (codecvt) has the same issues.

c) 22.2.1.5.2 (do_unshift) is even worse. It *forbids* the generation of a homing sequence for the basic character set, which might very well need one.

d) 22.2.1.5.2 (do_length) likewise dictates that the default mapping between wide and narrow characters be taken as one-for-one.

e) 22.2.2 para. 2 (num_get/put) is both muddled and vacuous, as far as I can tell. The muddle is, as before, calling Table 51 a list of instantiations. But the constraint it applies seems to me to cover *all* defined uses of num_get/put, so why bother to say so?

f) 22.2.3.1.2 para. 1(do_decimal_point) says "The required instantiations return '.' or L'.'.) Presumably this means "as appropriate for the character type. But given the vague definition of "required" earlier, this overrules *any* change of decimal point for non "C" locales. Surely we don't want to do that.

g) 22.2.3.1.2 para. 2 (do_thousands_sep) says "The required instantiations return ',' or L','.) As above, this probably means "as appropriate for the character type. But this overrules the "C" locale, which requires *no* character ('\0') for the thousands separator. Even if we agree that we don't mean to block changes in decimal point or thousands separator, we should also eliminate this clear incompatibility with C.

h) 22.2.3.1.2 para. 2 (do_grouping) says "The required instantiations return the empty string, indicating no grouping." Same considerations as for do_decimal_point.

i) 22.2.4.1 para. 1 (collate) refers to "instantiations required in Table 51". Same bad jargon.

j) 22.2.4.1.2 para. 1 (do_compare) refers to "instantiations required in Table 51". Same bad jargon.

k) 22.2.5 para. 1 (time_get/put) uses the same muddled and vacuous as num_get/put.

l) 22.2.6 para. 2 (money_get/put) uses the same muddled and vacuous as num_get/put.

m) 22.2.6.3.2 (do_pos/neg_format) says "The instantiations required in Table 51 ... return an object of type pattern initialized to {symbol, sign, none, value}." This once again *overrides* the "C" locale, as well as any other locale."

3) We constrain the use_facet calls that can be made by num_get/put, so why don't we do the same for money_get/put? Or for any of the other facets, for that matter?

4) As an almost aside, we spell out when a facet needs to use the ctype facet, but several also need to use a codecvt facet and we don't say so.

[ Berlin: Bill to provide wording. ]

Proposed resolution:


518. Are insert and erase stable for unordered_multiset and unordered_multimap?

Section: 23.1.3 [unord.req], TR1 6.3.1 [tr.unord.req] Status: Review Submitter: Matt Austern Date: 2005-07-03

View other active issues in [unord.req].

View all other issues in [unord.req].

View all issues with Review status.

Discussion:

Issue 371 deals with stability of multiset/multimap under insert and erase (i.e. do they preserve the relative order in ranges of equal elements). The same issue applies to unordered_multiset and unordered_multimap.

[ Moved to open (from review): There is no resolution. ]

[ Toronto: We have a resolution now. Moved to Review. Some concern was noted as to whether this conflicted with existing practice or not. An additional concern was in specifying (partial) ordering for an unordered container. ]

Proposed resolution:

Wording for the proposed resolution is taken from the equivalent text for associative containers.

Change 23.1.3 [unord.req], Unordered associative containers, paragraph 6 to:

An unordered associative container supports unique keys if it may contain at most one element for each key. Otherwise, it supports equivalent keys. unordered_set and unordered_map support unique keys. unordered_multiset and unordered_multimap support equivalent keys. In containers that support equivalent keys, elements with equivalent keys are adjacent to each other. For unordered_multiset and unordered_multimap, insert and erase preserve the relative ordering of equivalent elements.

Change 23.1.3 [unord.req], Unordered associative containers, paragraph 8 to:

The elements of an unordered associative container are organized into buckets. Keys with the same hash code appear in the same bucket. The number of buckets is automatically increased as elements are added to an unordered associative container, so that the average number of elements per bucket is kept below a bound. Rehashing invalidates iterators, changes ordering between elements, and changes which buckets elements appear in, but does not invalidate pointers or references to elements. For unordered_multiset and unordered_multimap, rehashing preserves the relative ordering of equivalent elements.


522. Tuple doesn't define swap

Section: 20.3 [tuple], TR1 6.1 [tr.tuple] Status: Open Submitter: Andy Koenig Date: 2005-07-03

View all issues with Open status.

Discussion:

Tuple doesn't define swap(). It should.

[ Berlin: Doug to provide wording. ]

[ Batavia: Howard to provide wording. ]

[ Toronto: Howard to provide wording (really this time). ]

Proposed resolution:


523. regex case-insensitive character ranges are unimplementable as specified

Section: 28 [re] Status: Open Submitter: Eric Niebler Date: 2005-07-01

View all other issues in [re].

View all issues with Open status.

Discussion:

A problem with TR1 regex is currently being discussed on the Boost developers list. It involves the handling of case-insensitive matching of character ranges such as [Z-a]. The proper behavior (according to the ECMAScript standard) is unimplementable given the current specification of the TR1 regex_traits<> class template. John Maddock, the author of the TR1 regex proposal, agrees there is a problem. The full discussion can be found at http://lists.boost.org/boost/2005/06/28850.php (first message copied below). We don't have any recommendations as yet.

-- Begin original message --

The situation of interest is described in the ECMAScript specification (ECMA-262), section 15.10.2.15:

"Even if the pattern ignores case, the case of the two ends of a range is significant in determining which characters belong to the range. Thus, for example, the pattern /[E-F]/i matches only the letters E, F, e, and f, while the pattern /[E-f]/i matches all upper and lower-case ASCII letters as well as the symbols [, \, ], ^, _, and `."

A more interesting case is what should happen when doing a case-insentitive match on a range such as [Z-a]. It should match z, Z, a, A and the symbols [, \, ], ^, _, and `. This is not what happens with Boost.Regex (it throws an exception from the regex constructor).

The tough pill to swallow is that, given the specification in TR1, I don't think there is any effective way to handle this situation. According to the spec, case-insensitivity is handled with regex_traits<>::translate_nocase(CharT) -- two characters are equivalent if they compare equal after both are sent through the translate_nocase function. But I don't see any way of using this translation function to make character ranges case-insensitive. Consider the difficulty of detecting whether "z" is in the range [Z-a]. Applying the transformation to "z" has no effect (it is essentially std::tolower). And we're not allowed to apply the transformation to the ends of the range, because as ECMA-262 says, "the case of the two ends of a range is significant."

So AFAICT, TR1 regex is just broken, as is Boost.Regex. One possible fix is to redefine translate_nocase to return a string_type containing all the characters that should compare equal to the specified character. But this function is hard to implement for Unicode, and it doesn't play nice with the existing ctype facet. What a mess!

-- End original message --

[ John Maddock adds: ]

One small correction, I have since found that ICU's regex package does implement this correctly, using a similar mechanism to the current TR1.Regex.

Given an expression [c1-c2] that is compiled as case insensitive it:

Enumerates every character in the range c1 to c2 and converts it to it's case folded equivalent. That case folded character is then used a key to a table of equivalence classes, and each member of the class is added to the list of possible matches supported by the character-class. This second step isn't possible with our current traits class design, but isn't necessary if the input text is also converted to a case-folded equivalent on the fly.

ICU applies similar brute force mechanisms to character classes such as [[:lower:]] and [[:word:]], however these are at least cached, so the impact is less noticeable in this case.

Quick and dirty performance comparisons show that expressions such as "[X-\\x{fff0}]+" are indeed very slow to compile with ICU (about 200 times slower than a "normal" expression). For an application that uses a lot of regexes this could have a noticeable performance impact. ICU also has an advantage in that it knows the range of valid characters codes: code points outside that range are assumed not to require enumeration, as they can not be part of any equivalence class. I presume that if we want the TR1.Regex to work with arbitrarily large character sets enumeration really does become impractical.

Finally note that Unicode has:

Three cases (upper, lower and title). One to many, and many to one case transformations. Character that have context sensitive case translations - for example an uppercase sigma has two different lowercase forms - the form chosen depends on context(is it end of a word or not), a caseless match for an upper case sigma should match either of the lower case forms, which is why case folding is often approximated by tolower(toupper(c)).

Probably we need some way to enumerate character equivalence classes, including digraphs (either as a result or an input), and some way to tell whether the next character pair is a valid digraph in the current locale.

Hoping this doesn't make this even more complex that it was already,

[ Portland: Alisdair: Detect as invalid, throw an exception. Pete: Possible general problem with case insensitive ranges. ]

Proposed resolution:


527. tr1::bind has lost its Throws clause

Section: 20.5.11.1.3 [func.bind.bind], TR1 3.6.3 [tr.func.bind.bind] Status: Open Submitter: Peter Dimov Date: 2005-10-01

View all issues with Open status.

Discussion:

The original bind proposal gives the guarantee that tr1::bind(f, t1, ..., tN) does not throw when the copy constructors of f, t1, ..., tN don't.

This guarantee is not present in the final version of TR1.

I'm pretty certain that we never removed it on purpose. Editorial omission? :-)

[ Berlin: not quite editorial, needs proposed wording. ]

[ Batavia: Doug to translate wording to variadic templates. ]

[ Toronto: We agree but aren't quite happy with the wording. The "t"'s no longer refer to anything. Alan to provide improved wording. ]

Proposed resolution:

In 20.5.10.1.3 [lib.func.bind.bind] ([tr.func.bind.bind]), add a new paragraph after p2:

Throws: Nothing unless one of the copy constructors of f, t1, t2, ..., tN throws an exception.

Add a new paragraph after p4:

Throws: nothing unless one of the copy constructors of f, t1, t2, ..., tN throws an exception.


529. The standard encourages redundant and confusing preconditions

Section: 17.4.3.8 [res.on.required] Status: Open Submitter: David Abrahams Date: 2005-10-25

View all issues with Open status.

Discussion:

17.4.3.8/1 says:

Violation of the preconditions specified in a function's Required behavior: paragraph results in undefined behavior unless the function's Throws: paragraph specifies throwing an exception when the precondition is violated.

This implies that a precondition violation can lead to defined behavior. That conflicts with the only reasonable definition of precondition: that a violation leads to undefined behavior. Any other definition muddies the waters when it comes to analyzing program correctness, because precondition violations may be routinely done in correct code (e.g. you can use std::vector::at with the full expectation that you'll get an exception when your index is out of range, catch the exception, and continue). Not only is it a bad example to set, but it encourages needless complication and redundancy in the standard. For example:

  21 Strings library 
  21.3.3 basic_string capacity

  void resize(size_type n, charT c);

  5 Requires: n <= max_size()
  6 Throws: length_error if n > max_size().
  7 Effects: Alters the length of the string designated by *this as follows:

The Requires clause is entirely redundant and can be dropped. We could make that simplifying change (and many others like it) even without changing 17.4.3.8/1; the wording there just seems to encourage the redundant and error-prone Requires: clause.

[ Batavia: Alan and Pete to work. ]

Proposed resolution:

1. Change 17.4.3.8/1 to read:

Violation of the preconditions specified in a function's Required behavior: paragraph results in undefined behavior unless the function's Throws: paragraph specifies throwing an exception when the precondition is violated.

2. Go through and remove redundant Requires: clauses. Specifics to be provided by Dave A.

[ Berlin: The LWG requests a detailed survey of part 2 of the proposed resolution. ]

[ Alan provided the survey N2121. ]


539. partial_sum and adjacent_difference should mention requirements

Section: 26.6.3 [partial.sum] Status: Open Submitter: Marc Schoolderman Date: 2006-02-06

View all issues with Open status.

Discussion:

There are some problems in the definition of partial_sum and adjacent_difference in 26.4 [lib.numeric.ops]

Unlike accumulate and inner_product, these functions are not parametrized on a "type T", instead, 26.4.3 [lib.partial.sum] simply specifies the effects clause as;

Assigns to every element referred to by iterator i in the range [result,result + (last - first)) a value correspondingly equal to

((...(* first + *( first + 1)) + ...) + *( first + ( i - result )))

And similarly for BinaryOperation. Using just this definition, it seems logical to expect that:

char i_array[4] = { 100, 100, 100, 100 };
int  o_array[4];

std::partial_sum(i_array, i_array+4, o_array);

Is equivalent to

int o_array[4] = { 100, 100+100, 100+100+100, 100+100+100+100 };

i.e. 100, 200, 300, 400, with addition happening in the result type, int.

Yet all implementations I have tested produce 100, -56, 44, -112, because they are using an accumulator of the InputIterator's value_type, which in this case is char, not int.

The issue becomes more noticeable when the result of the expression *i + *(i+1) or binary_op(*i, *i-1) can't be converted to the value_type. In a contrived example:

enum not_int { x = 1, y = 2 };
...
not_int e_array[4] = { x, x, y, y };
std::partial_sum(e_array, e_array+4, o_array);

Is it the intent that the operations happen in the input type, or in the result type?

If the intent is that operations happen in the result type, something like this should be added to the "Requires" clause of 26.4.3/4 [lib.partial.sum]:

The type of *i + *(i+1) or binary_op(*i, *(i+1)) shall meet the requirements of CopyConstructible (20.1.3) and Assignable (23.1) types.

(As also required for T in 26.4.1 [lib.accumulate] and 26.4.2 [lib.inner.product].)

The "auto initializer" feature proposed in N1894 is not required to implement partial_sum this way. The 'narrowing' behaviour can still be obtained by using the std::plus<> function object.

If the intent is that operations happen in the input type, then something like this should be added instead;

The type of *first shall meet the requirements of CopyConstructible (20.1.3) and Assignable (23.1) types. The result of *i + *(i+1) or binary_op(*i, *(i+1)) shall be convertible to this type.

The 'widening' behaviour can then be obtained by writing a custom proxy iterator, which is somewhat involved.

In both cases, the semantics should probably be clarified.

26.4.4 [lib.adjacent.difference] is similarly underspecified, although all implementations seem to perform operations in the 'result' type:

unsigned char i_array[4] = { 4, 3, 2, 1 };
int o_array[4];

std::adjacent_difference(i_array, i_array+4, o_array);

o_array is 4, -1, -1, -1 as expected, not 4, 255, 255, 255.

In any case, adjacent_difference doesn't mention the requirements on the value_type; it can be brought in line with the rest of 26.4 [lib.numeric.ops] by adding the following to 26.4.4/2 [lib.adjacent.difference]:

The type of *first shall meet the requirements of CopyConstructible (20.1.3) and Assignable (23.1) types."

[ Berlin: Giving output iterator's value_types very controversial. Suggestion of adding signatures to allow user to specify "accumulator". ]

Proposed resolution:


546. _Longlong and _ULonglong are integer types

Section: TR1 5.1.1 [tr.rand.req] Status: Open Submitter: Matt Austern Date: 2006-01-10

View all issues with Open status.

Discussion:

The TR sneaks in two new integer types, _Longlong and _Ulonglong, in [tr.c99]. The rest of the TR should use that type. I believe this affects two places. First, the random number requirements, 5.1.1/10-11, lists all of the types with which template parameters named IntType and UIntType may be instantiated. _Longlong (or "long long", assuming it is added to C++0x) should be added to the IntType list, and UIntType (again, or "unsigned long long") should be added to the UIntType list. Second, 6.3.2 lists the types for which hash<> is required to be instantiable. _Longlong and _Ulonglong should be added to that list, so that people may use long long as a hash key.

Proposed resolution:


550. What should the return type of pow(float,int) be?

Section: 26.7 [c.math] Status: Open Submitter: Howard Hinnant Date: 2006-01-12

View other active issues in [c.math].

View all other issues in [c.math].

View all issues with Open status.

Discussion:

Assuming we adopt the C compatibility package from C99 what should be the return type of the following signature be:

?  pow(float, int);

C++03 says that the return type should be float. TR1 and C90/99 say the return type should be double. This can put clients into a situation where C++03 provides answers that are not as high quality as C90/C99/TR1. For example:

#include <math.h>

int main()
{
    float x = 2080703.375F;
    double y = pow(x, 2);
}

Assuming an IEEE 32 bit float and IEEE 64 bit double, C90/C99/TR1 all suggest:

y = 4329326534736.390625

which is exactly right. While C++98/C++03 demands:

y = 4329326510080.

which is only approximately right.

I recommend that C++0X adopt the mixed mode arithmetic already adopted by Fortran, C and TR1 and make the return type of pow(float,int) be double.

[ Kona (2007): Other functions that are affected by this issue include ldexp, scalbln, and scalbn. We also believe that there is a typo in 26.7/10: float nexttoward(float, long double); [sic] should be float nexttoward(float, float); Proposed Disposition: Review (the proposed resolution appears above, rather than below, the heading "Proposed resolution") ]

[

Howard, post Kona:

Unfortunately I strongly disagree with a part of the resolution from Kona. I am moving from New to Open instead of to Review because I do not believe we have consensus on the intent of the resolution.

This issue does not include ldexp, scalbln, and scalbn because the second integral parameter in each of these signatures (from C99) is not a generic parameter according to C99 7.22p2. The corresponding C++ overloads are intended (as far as I know) to correspond directly to C99's definition of generic parameter.

For similar reasons, I do not believe that the second long double parameter of nexttoward, nor the return type of this function, is in error. I believe the correct signature is:

float nexttoward(float, long double);

which is what both the C++0X working paper and C99 state (as far as I currently understand).

This is really only about pow(float, int). And this is because C++98 took one route (with pow only) and C99 took another (with many math functions in <tgmath.h>. The proposed resolution basically says: C++98 got it wrong and C99 got it right; let's go with C99.

]

Proposed resolution:

Change 26.7 [c.math]

float pow(float, float); 
float double pow(float, int);

556. is Compare a BinaryPredicate?

Section: 25.3 [alg.sorting] Status: Open Submitter: Martin Sebor Date: 2006-02-05

View all other issues in [alg.sorting].

View all issues with Open status.

Discussion:

In 25, p8 we allow BinaryPredicates to return a type that's convertible to bool but need not actually be bool. That allows predicates to return things like proxies and requires that implementations be careful about what kinds of expressions they use the result of the predicate in (e.g., the expression in if (!pred(a, b)) need not be well-formed since the negation operator may be inaccessible or return a type that's not convertible to bool).

Here's the text for reference:

...if an algorithm takes BinaryPredicate binary_pred as its argument and first1 and first2 as its iterator arguments, it should work correctly in the construct if (binary_pred(*first1, first2)){...}.

In 25.3, p2 we require that the Compare function object return true of false, which would seem to preclude such proxies. The relevant text is here:

Compare is used as a function object which returns true if the first argument is less than the second, and false otherwise...

Proposed resolution:

I think we could fix this by rewording 25.3, p2 to read somthing like:

-2- Compare is used as a function object which returns true if the first argument a BinaryPredicate. The return value of the function call operator applied to an object of type Compare, when converted to type bool, yields true if the first argument of the call is less than the second, and false otherwise. Compare comp is used throughout for algorithms assuming an ordering relation. It is assumed that comp will not apply any non-constant function through the dereferenced iterator.

[ Portland: Jack to define "convertible to bool" such that short circuiting isn't destroyed. ]


557. TR1: div(_Longlong, _Longlong) vs div(intmax_t, intmax_t)

Section: 18.3 [cstdint], TR1 8.22 [tr.c99.cstdint] Status: Open Submitter: Paolo Carlini Date: 2006-02-06

View all other issues in [cstdint].

View all issues with Open status.

Discussion:

I'm seeing a problem with such overloads: when, _Longlong == intmax_t == long long we end up, essentially, with the same arguments and different return types (lldiv_t and imaxdiv_t, respectively). Similar issue with abs(_Longlong) and abs(intmax_t), of course.

Comparing sections 8.25 and 8.11, I see an important difference, however: 8.25.3 and 8.25.4 carefully describe div and abs for _Longlong types (rightfully, because not moved over directly from C99), whereas there is no equivalent in 8.11: the abs and div overloads for intmax_t types appear only in the synopsis and are not described anywhere, in particular no mention in 8.11.2 (at variance with 8.25.2).

I'm wondering whether we really, really, want div and abs for intmax_t...

Proposed resolution:

[ Portland: no consensus. ]

Rationale:

[ Batavia, Bill: The <cstdint> synopsis in TR1 8.11.1 [tr.c99.cinttypes.syn] contains: ]

intmax_t imaxabs(intmax_t i);
intmax_t abs(intmax_t i);

imaxdiv_t imaxdiv(intmax_t numer, intmax_t denom);
imaxdiv_t div(intmax_t numer, intmax_t denom);

[ and in TR1 8.11.2 [tr.c99.cinttypes.def]: ]

The header defines all functions, types, and macros the same as C99 subclause 7.8.

[ This is as much definition as we give for most other C99 functions, so nothing need change. We might, however, choose to add the footnote: ]

[Note: These overloads for abs and div may well be equivalent to those that take long long arguments. If so, the implementation is responsible for avoiding conflicting declarations. -- end note]


561. inserter overly generic

Section: 24.4.2.6.5 [inserter] Status: Ready Submitter: Howard Hinnant Date: 2006-02-21

View all issues with Ready status.

Discussion:

The declaration of std::inserter is:

template <class Container, class Iterator>
insert_iterator<Container>
inserter(Container& x, Iterator i);

The template parameter Iterator in this function is completely unrelated to the template parameter Container when it doesn't need to be. This causes the code to be overly generic. That is, any type at all can be deduced as Iterator, whether or not it makes sense. Now the same is true of Container. However, for every free (unconstrained) template parameter one has in a signature, the opportunity for a mistaken binding grows geometrically.

It would be much better if inserter had the following signature instead:

template <class Container>
insert_iterator<Container>
inserter(Container& x, typename Container::iterator i);

Now there is only one free template parameter. And the second argument to inserter must be implicitly convertible to the container's iterator, else the call will not be a viable overload (allowing other functions in the overload set to take precedence). Furthermore, the first parameter must have a nested type named iterator, or again the binding to std::inserter is not viable. Contrast this with the current situation where any type can bind to Container or Iterator and those types need not be anything closely related to containers or iterators.

This can adversely impact well written code. Consider:

#include <iterator>
#include <string>

namespace my
{

template <class String>
struct my_type {};

struct my_container
{
template <class String>
void push_back(const my_type<String>&);
};

template <class String>
void inserter(const my_type<String>& m, my_container& c) {c.push_back(m);}

}  // my

int main()
{
    my::my_container c;
    my::my_type<std::string> m;
    inserter(m, c);
}

Today this code fails because the call to inserter binds to std::inserter instead of to my::inserter. However with the proposed change std::inserter will no longer be a viable function which leaves only my::inserter in the overload resolution set. Everything works as the client intends.

To make matters a little more insidious, the above example works today if you simply change the first argument to an rvalue:

    inserter(my::my_type(), c);

It will also work if instantiated with some string type other than std::string (or any other std type). It will also work if <iterator> happens to not get included.

And it will fail again for such inocuous reaons as my_type or my_container privately deriving from any std type.

It seems unfortunate that such simple changes in the client's code can result in such radically differing behavior.

Proposed resolution:

Change 24.2:

24.2 Header <iterator> synopsis

...
template <class Container, class Iterator>
   insert_iterator<Container> inserter(Container& x, Iterator typename Container::iterator i);
...

Change 24.4.2.5:

24.4.2.5 Class template insert_iterator

...
template <class Container, class Iterator>
   insert_iterator<Container> inserter(Container& x, Iterator typename Container::iterator i);
...

Change 24.4.2.6.5:

24.4.2.6.5 inserter

template <class Container, class Inserter>
   insert_iterator<Container> inserter(Container& x, Inserter typename Container::iterator i);

-1- Returns: insert_iterator<Container>(x,typename Container::iterator(i)).

[ Kona (2007): This issue will probably be addressed as a part of the concepts overhaul of the library anyway, but the proposed resolution is correct in the absence of concepts. Proposed Disposition: Ready ]


562. stringbuf ctor inefficient

Section: 27.7 [string.streams] Status: Ready Submitter: Martin Sebor Date: 2006-02-23

View all other issues in [string.streams].

View all issues with Ready status.

Discussion:

For better efficiency, the requirement on the stringbuf ctor that takes a string argument should be loosened up to let it set epptr() beyond just one past the last initialized character just like overflow() has been changed to be allowed to do (see issue 432). That way the first call to sputc() on an object won't necessarily cause a call to overflow. The corresponding change should be made to the string overload of the str() member function.

Proposed resolution:

Change 27.7.1.1, p3 of the Working Draft, N1804, as follows:

explicit basic_stringbuf(const basic_string<charT,traits,Allocator>& str,
               ios_base::openmode which = ios_base::in | ios_base::out);

-3- Effects: Constructs an object of class basic_stringbuf, initializing the base class with basic_streambuf() (27.5.2.1), and initializing mode with which. Then calls str(s). 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 pptr() is equal to epptr() if which & ios_base::ate is true, otherwise pptr() is 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.

Change the Effects clause of the str() in 27.7.1.2, p2 to read:

-2- Effects: Copies the contents of s into the basic_stringbuf underlying character sequence and initializes the input and output sequences according to mode. 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 pptr() is equal to epptr() if mode & ios_base::in is true, otherwise pptr() is 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.

-3- Postconditions: If mode & ios_base::out is true, pbase() points to the first underlying character and (epptr() >= pbase() + s.size()) holds; in addition, if mode & ios_base::in is true, (pptr() == pbase() + s.data()) holds, otherwise (pptr() == pbase()) is true. If mode & ios_base::in is true, eback() points to the first underlying character, and (gptr() == eback()) and (egptr() == eback() + s.size()) hold.

[ Kona (2007) Moved to Ready. ]


563. stringbuf seeking from end

Section: 27.7.1.4 [stringbuf.virtuals] Status: Ready Submitter: Martin Sebor Date: 2006-02-23

View other active issues in [stringbuf.virtuals].

View all other issues in [stringbuf.virtuals].

View all issues with Ready status.

Discussion:

According to Table 92 (unchanged by issue 432), when (way == end) the newoff value in out mode is computed as the difference between epptr() and pbase().

This value isn't meaningful unless the value of epptr() can be precisely controlled by a program. That used to be possible until we accepted the resolution of issue 432, but since then the requirements on overflow() have been relaxed to allow it to make more than 1 write position available (i.e., by setting epptr() to some unspecified value past pptr()). So after the first call to overflow() positioning the output sequence relative to end will have unspecified results.

In addition, in in|out mode, since (egptr() == epptr()) need not hold, there are two different possible values for newoff: epptr() - pbase() and egptr() - eback().

Proposed resolution:

Change the newoff column in the last row of Table 94 to read:

the end high mark pointer minus the beginning pointer (xend high_mark - xbeg).

[ Kona (2007) Moved to Ready. ]


564. stringbuf seekpos underspecified

Section: 27.7.1.4 [stringbuf.virtuals] Status: Open Submitter: Martin Sebor Date: 2006-02-23

View other active issues in [stringbuf.virtuals].

View all other issues in [stringbuf.virtuals].

View all issues with Open status.

Discussion:

The effects of the seekpos() member function of basic_stringbuf simply say that the function positions the input and/or output sequences but fail to spell out exactly how. This is in contrast to the detail in which seekoff() is described.

Proposed resolution:

Change 27.7.1.3, p13 to read:

-13- Effects: Same as seekoff(off_type(sp), ios_base::beg, which). Alters the stream position within the controlled sequences, if possible, to correspond to the stream position stored in sp (as described below).

[ Kona (2007): A pos_type is a position in a stream by definition, so there is no ambiguity as to what it means. Proposed Disposition: NAD ]

[ Post-Kona Martin adds: I'm afraid I disagree with the Kona '07 rationale for marking it NAD. The only text that describes precisely what it means to position the input or output sequence is in seekoff(). The seekpos() Effects clause is inadequate in comparison and the proposed resolution plugs the hole by specifying seekpos() in terms of seekoff(). ]


565. xsputn inefficient

Section: 27.5.2.4.5 [streambuf.virt.put] Status: Open Submitter: Martin Sebor Date: 2006-02-23

View all issues with Open status.

Discussion:

streambuf::xsputn() is specified to have the effect of "writing up to n characters to the output sequence as if by repeated calls to sputc(c)."

Since sputc() is required to call overflow() when (pptr() == epptr()) is true, strictly speaking xsputn() should do the same. However, doing so would be suboptimal in some interesting cases, such as in unbuffered mode or when the buffer is basic_stringbuf.

Assuming calling overflow() is not really intended to be required and the wording is simply meant to describe the general effect of appending to the end of the sequence it would be worthwhile to mention in xsputn() that the function is not actually required to cause a call to overflow().

Proposed resolution:

Add the following sentence to the xsputn() Effects clause in 27.5.2.4.5, p1 (N1804):

-1- Effects: Writes up to n characters to the output sequence as if by repeated calls to sputc(c). The characters written are obtained from successive elements of the array whose first element is designated by s. Writing stops when either n characters have been written or a call to sputc(c) would return traits::eof(). It is uspecified whether the function calls overflow() when (pptr() == epptr()) becomes true or whether it achieves the same effects by other means.

In addition, I suggest to add a footnote to this function with the same text as Footnote 292 to make it extra clear that derived classes are permitted to override xsputn() for efficiency.

[ Kona (2007): We want to permit a streambuf that streams output directly to a device without making calls to sputc or overflow. We believe that has always been the intention of the committee. We believe that the proposed wording doesn't accomplish that. Proposed Disposition: Open ]


567. streambuf inserter and extractor should be unformatted

Section: 27.6 [iostream.format] Status: Ready Submitter: Martin Sebor Date: 2006-02-25

View other active issues in [iostream.format].

View all other issues in [iostream.format].

View all issues with Ready status.

Discussion:

Issue 60 explicitly made the extractor and inserter operators that take a basic_streambuf* argument formatted input and output functions, respectively. I believe that's wrong, certainly in the case of the extractor, since formatted functions begin by extracting and discarding whitespace. The extractor should not discard any characters.

Proposed resolution:

I propose to change each operator to behave as unformatted input and output function, respectively. The changes below are relative to the working draft document number N1804.

Specifically, change 27.6.1.2.3, p14 as follows:

Effects: Behaves as an unformatted input function (as described in 27.6.1.2.127.6.1.3, paragraph 1).

And change 27.6.2.5.3, p7 as follows:

Effects: Behaves as an unformatted output function (as described in 27.6.2.5.127.6.2.6, paragraph 1).

[ Kona (2007): Proposed Disposition: Ready ]


568. log2 overloads missing

Section: TR1 8.16.4 [tr.c99.cmath.over] Status: New Submitter: Paolo Carlini Date: 2006-03-07

View all issues with New status.

Discussion:

log2 is missing from the list of "additional overloads" in TR1 8.16.4 [tr.c99.cmath.over] p1.

Hinnant: This is a TR1 issue only. It is fixed in the current (N2135) WD.

Proposed resolution:

Add log2 to the list of functions in TR1 8.16.4 [tr.c99.cmath.over] p1.


570. Request adding additional explicit specializations of char_traits

Section: 21.1 [char.traits] Status: Open Submitter: Jack Reeves Date: 2006-04-06

View all issues with Open status.

Discussion:

Currently, the Standard Library specifies only a declaration for template class char_traits<> and requires the implementation provide two explicit specializations: char_traits<char> and char_traits<wchar_t>. I feel the Standard should require explicit specializations for all built-in character types, i.e. char, wchar_t, unsigned char, and signed char.

I have put together a paper (N1985) that describes this in more detail and includes all the necessary wording.

[ Portland: Jack will rewrite N1985 to propose a primary template that will work with other integral types. ]

[ Toronto: issue has grown with addition of char16_t and char32_t. ]

Proposed resolution:


573. C++0x file positioning should handle modern file sizes

Section: 27.4.3 [fpos] Status: Open Submitter: Beman Dawes Date: 2006-04-12

View all other issues in [fpos].

View all issues with Open status.

Discussion:

There are two deficiencies related to file sizes:

  1. It doesn't appear that the Standard Library is specified in a way that handles modern file sizes, which are often too large to be represented by an unsigned long.
  2. The std::fpos class does not currently have the ability to set/get file positions.

The Dinkumware implementation of the Standard Library as shipped with the Microsoft compiler copes with these issues by:

  1. Defining fpos_t be long long, which is large enough to represent any file position likely in the foreseeable future.
  2. Adding member functions to class fpos. For example,
    fpos_t seekpos() const;
    

Because there are so many types relating to file positions and offsets (fpos_t, fpos, pos_type, off_type, streamoff, streamsize, streampos, wstreampos, and perhaps more), it is difficult to know if the Dinkumware extensions are sufficient. But they seem a useful starting place for discussions, and they do represent existing practice.

[ Kona (2007): We need a paper. It would be nice if someone proposed clarifications to the definitions of pos_type and off_type. Currently these definitions are horrible. Proposed Disposition: Open ]

Proposed resolution:


574. DR 369 Contradicts Text

Section: 27.3 [iostream.objects] Status: Review Submitter: Pete Becker Date: 2006-04-18

View all other issues in [iostream.objects].

View all issues with Review status.

Discussion:

lib.iostream.objects requires that the standard stream objects are never destroyed, and it requires that they be destroyed.

DR 369 adds words to say that we really mean for ios_base::Init objects to force construction of standard stream objects. It ends, though, with the phrase "these stream objects shall be destroyed after the destruction of dynamically ...". However, the rule for destruction is stated in the standard: "The objects are not destroyed during program execution."

Proposed resolution:

Change 27.3 [iostream.objects]/1:

-2- The objects are constructed and the associations are established at some time prior to or during the first time an object of class ios_base::Init is constructed, and in any case before the body of main begins execution.290) The objects are not destroyed during program execution.291) If a translation unit includes <iostream&t; or explicitly constructs an ios_base::Init object, these stream objects shall be constructed before dynamic initialization of non-local objects defined later in that translation unit, and these stream objects shall be destroyed after the destruction of dynamically initialized non-local objects defined later in that translation unit.

[ Kona (2007): From 27.3 [iostream.objects]/2, strike the words "...and these stream objects shall be destroyed after the destruction of dynamically initialized non-local objects defined later in that translation unit." Proposed Disposition: Review ]


579. erase(iterator) for unordered containers should not return an iterator

Section: 23.1.3 [unord.req] Status: Open Submitter: Joaquín M López Muñoz Date: 2006-06-13

View other active issues in [unord.req].

View all other issues in [unord.req].

View all issues with Open status.

Discussion:

See N2023 for full discussion.

Proposed resolution:

Option 1:

The problem can be eliminated by omitting the requirement that a.erase(q) return an iterator. This is, however, in contrast with the equivalent requirements for other standard containers.

Option 2:

a.erase(q) can be made to compute the next iterator only when explicitly requested: the technique consists in returning a proxy object implicitly convertible to iterator, so that

iterator q1=a.erase(q);

works as expected, while

a.erase(q);

does not ever invoke the conversion-to-iterator operator, thus avoiding the associated computation. To allow this technique, some sections of TR1 along the line "return value is an iterator..." should be changed to "return value is an unspecified object implicitly convertible to an iterator..." Although this trick is expected to work transparently, it can have some collateral effects when the expression a.erase(q) is used inside generic code.

Rationale:

N2023 was discussed in Portland and the consensus was that there appears to be no need for either change proposed in the paper. The consensus opinion was that since the iterator could serve as its own proxy, there appears to be no need for the change. In general, "converts to" is undesirable because it interferes with template matching.

Post Toronto: There does not at this time appear to be consensus with the Portland consensus.


580. unused allocator members

Section: 20.1.2 [allocator.requirements] Status: Open Submitter: Martin Sebor Date: 2006-06-14

View other active issues in [allocator.requirements].

View all other issues in [allocator.requirements].

View all issues with Open status.

Duplicate of: 479

Discussion:

C++ Standard Library templates that take an allocator as an argument are required to call the allocate() and deallocate() members of the allocator object to obtain storage. However, they do not appear to be required to call any other allocator members such as construct(), destroy(), address(), and max_size(). This makes these allocator members less than useful in portable programs.

It's unclear to me whether the absence of the requirement to use these allocator members is an unintentional omission or a deliberate choice. However, since the functions exist in the standard allocator and since they are required to be provided by any user-defined allocator I believe the standard ought to be clarified to explictly specify whether programs should or should not be able to rely on standard containers calling the functions.

I propose that all containers be required to make use of these functions.

[ Batavia: We support this resolution. Martin to provide wording. ]

[ pre-Oxford: Martin provided wording. ]

Proposed resolution:

       

Specifically, I propose to change 23.1 [container.requirements], p9 as follows:        

           

-9- Copy constructors  for all container types defined  in this clause that   are  parametrized  on   Allocator  copy anthe  allocator argument from  their respective first parameters. All other  constructors for  these container types  take an const  Allocator&  argument  (20.1.6),  an allocator whose value_type is the same as the container's value_type. A copy of this  argument isshall be used for any memory  allocation and  deallocation performed, by these  constructors and by all  member functions, during the  lifetime  of each  container  object.   Allocation shall  be performed  "as  if"  by  calling  the  allocate()  member function on  a copy  of the allocator  object of the  appropriate type New  Footnote),   and  deallocation  "as   if"  by  calling deallocate() on  a copy of  the same allocator  object of the corresponding type. A  copy of  this argument  shall also  be used  to  construct and destroy objects whose lifetime  is managed by the container, including but not  limited to those of  the container's value_type, and  to  obtain  their  address.   All  objects  residing  in  storage allocated by a  container's allocator shall be constructed  "as if" by calling the construct() member  function on a copy of the allocator object of  the appropriate type.  The same  objects shall be destroyed "as if"  by calling destroy() on a  copy of the same allocator object  of the same type.  The  address of such objects shall be obtained "as if" by calling the address() member function  on  a  copy  of  the allocator  object  of  the  appropriate type. Finally, a copy  of this argument shall be  used by its container object to determine  the maximum number of objects  of the container's value_type the container may  store at the same time. The container  member function max_size() obtains  this number from      the      value      returned      by     a      call      to get_allocator().max_size(). In   all  container   types  defined   in  this   clause that  are parametrized     on    Allocator,     the    member get_allocator()     returns     a     copy     of     the Allocator     object     used     to    construct     the container.258)

New Footnote: This type  may be different from Allocator: it     may    be     derived    from     Allocator    via Allocator::rebind<U>::other   for  the  appropriate type U.

           
       

The proposed wording seems cumbersome but I couldn't think of a better way   to  describe   the   requirement  that   containers  use   their Allocator  to manage  only objects  (regardless  of their type)  that  persist  over  their  lifetimes  and  not,  for  example, temporaries  created on the  stack. That  is, containers  shouldn't be required  to  call  Allocator::construct(Allocator::allocate(1), elem)  just to  construct a  temporary copy  of an  element, or Allocator::destroy(Allocator::address(temp),     1)    to destroy temporaries.        

[ Howard: This same paragraph will need some work to accommodate 431. ]

[ post Oxford: This would be rendered NAD Editorial by acceptance of N2257. ]


581. flush() not unformatted function

Section: 27.6.2.7 [ostream.unformatted] Status: Ready Submitter: Martin Sebor Date: 2006-06-14

View all other issues in [ostream.unformatted].

View all issues with Ready status.

Discussion:

The resolution of issue 60 changed basic_ostream::flush() so as not to require it to behave as an unformatted output function. That has at least two in my opinion problematic consequences:

First, flush() now calls rdbuf()->pubsync() unconditionally, without regard to the state of the stream. I can't think of any reason why flush() should behave differently from the vast majority of stream functions in this respect.

Second, flush() is not required to catch exceptions from pubsync() or set badbit in response to such events. That doesn't seem right either, as most other stream functions do so.

Proposed resolution:

I propose to revert the resolution of issue 60 with respect to flush(). Specifically, I propose to change 27.6.2.6, p7 as follows:

Effects: Behaves as an unformatted output function (as described in 27.6.2.6, paragraph 1). If rdbuf() is not a null pointer, constructs a sentry object. If this object returns true when converted to a value of type bool the function calls rdbuf()->pubsync(). If that function returns -1 calls setstate(badbit) (which may throw ios_base::failure (27.4.4.3)). Otherwise, if the sentry object returns false, does nothing.Does not behave as an unformatted output function (as described in 27.6.2.6, paragraph 1).

[ Kona (2007): Proposed Disposition: Ready ]


582. specialized algorithms and volatile storage

Section: 20.6.4.1 [uninitialized.copy] Status: Open Submitter: Martin Sebor Date: 2006-06-14

View other active issues in [uninitialized.copy].

View all other issues in [uninitialized.copy].

View all issues with Open status.

Discussion:

The specialized algorithms [lib.specialized.algorithms] are specified as having the general effect of invoking the following expression:


new (static_cast<void*>(&*i))
    typename iterator_traits<ForwardIterator>::value_type (x)

            

This expression is ill-formed when the type of the subexpression &*i is some volatile-qualified T.

[ Batavia: Lack of support for proposed resolution but agree there is a defect. Howard to look at wording. Concern that move semantics properly expressed if iterator returns rvalue. ]

Proposed resolution:

In order to allow these algorithms to operate on volatile storage I propose to change the expression so as to make it well-formed even for pointers to volatile types. Specifically, I propose the following changes to clauses 20 and 24. Change 20.6.4.1, p1 to read:


Effects:

typedef typename iterator_traits<ForwardIterator>::pointer    pointer;
typedef typename iterator_traits<ForwardIterator>::value_type value_type;

for (; first != last; ++result, ++first)
    new (static_cast<void*>(const_cast<pointer>(&*result))
        value_type (*first);

            

change 20.6.4.2, p1 to read


Effects:

typedef typename iterator_traits<ForwardIterator>::pointer    pointer;
typedef typename iterator_traits<ForwardIterator>::value_type value_type;

for (; first != last; ++result, ++first)
    new (static_cast<void*>(const_cast<pointer>(&*first))
        value_type (*x);

            

and change 20.6.4.3, p1 to read


Effects:

typedef typename iterator_traits<ForwardIterator>::pointer    pointer;
typedef typename iterator_traits<ForwardIterator>::value_type value_type;

for (; n--; ++first)
    new (static_cast<void*>(const_cast<pointer>(&*first))
        value_type (*x);

            

In addition, since there is no partial specialization for iterator_traits<volatile T*> I propose to add one to parallel such specialization for <const T*>. Specifically, I propose to add the following text to the end of 24.3.1, p3:

and for pointers to volatile as


namespace std {
template<class T> struct iterator_traits<volatile T*> {
typedef ptrdiff_t difference_type;
typedef T value_type;
typedef volatile T* pointer;
typedef volatile T& reference;
typedef random_access_iterator_tag iterator_category;
};
}

            

Note that the change to iterator_traits isn't necessary in order to implement the specialized algorithms in a way that allows them to operate on volatile strorage. It is only necesassary in order to specify their effects in terms of iterator_traits as is done here. Implementations can (and some do) achieve the same effect by means of function template overloading.


585. facet error reporting

Section: 22.2 [locale.categories] Status: Open Submitter: Martin Sebor, Paolo Carlini Date: 2006-06-22

View other active issues in [locale.categories].

View all other issues in [locale.categories].

View all issues with Open status.

Discussion:

Section 22.2, paragraph 2 requires facet get() members that take an ios_base::iostate& argument, err, to ignore the (initial) value of the argument, but to set it to ios_base::failbit in case of a parse error.

We believe there are a few minor problems with this blanket requirement in conjunction with the wording specific to each get() member function.

First, besides get() there are other member functions with a slightly different name (for example, get_date()). It's not completely clear that the intent of the paragraph is to include those as well, and at least one implementation has interpreted the requirement literally.

Second, the requirement to "set the argument to ios_base::failbit suggests that the functions are not permitted to set it to any other value (such as ios_base::eofbit, or even ios_base::eofbit | ios_base::failbit).

However, 22.2.2.1.2, p5 (Stage 3 of num_get parsing) and p6 (bool parsing) specifies that the do_get functions perform err |= ios_base::eofbit, which contradicts the earlier requirement to ignore err's initial value.

22.2.6.1.2, p1 (the Effects clause of the money_get facet's do_get member functions) also specifies that err's initial value be used to compute the final value by ORing it with either ios_base::failbit or withios_base::eofbit | ios_base::failbit.

Proposed resolution:

We believe the intent is for all facet member functions that take an ios_base::iostate& argument to:

To that effect we propose to change 22.2, p2 as follows:

The put() members make no provision for error reporting. (Any failures of the OutputIterator argument must be extracted from the returned iterator.) Unless otherwise specified, the get() members that take an ios_base::iostate& argument whose value they ignore, but set to ios_base::failbit in case of a parse error., err, start by evaluating err = ios_base::goodbit, and may subsequently set err to either ios_base::eofbit, or ios_base::failbit, or ios_base::eofbit | ios_base::failbit in response to reaching the end-of-file or in case of a parse error, or both, respectively.

[ Kona (2007): We need to change the proposed wording to clarify that the phrase "the get members" actually denotes get(), get_date(), etc. Proposed Disposition: Open ]


588. requirements on zero sized tr1::arrays and other details

Section: 23.2.1 [array] Status: Open Submitter: Gennaro Prota Date: 2006-07-18

View other active issues in [array].

View all other issues in [array].

View all issues with Open status.

Discussion:

The wording used for section 23.2.1 [lib.array] seems to be subtly ambiguous about zero sized arrays (N==0). Specifically:

* "An instance of array<T, N> stores N elements of type T, so that [...]"

Does this imply that a zero sized array object stores 0 elements, i.e. that it cannot store any element of type T? The next point clarifies the rationale behind this question, basically how to implement begin() and end():

* 23.2.1.5 [lib.array.zero], p2: "In the case that N == 0, begin() == end() == unique value."

What does "unique" mean in this context? Let's consider the following possible implementations, all relying on a partial specialization:

a)
    template< typename T >
    class array< T, 0 > {
    
        ....

        iterator begin()
        { return iterator( reinterpret_cast< T * >( this ) ); }
        ....

    };

This has been used in boost, probably intending that the return value had to be unique to the specific array object and that array couldn't store any T. Note that, besides relying on a reinterpret_cast, has (more than potential) alignment problems.

b)
    template< typename T >
    class array< T, 0 > {
    
        T t;

        iterator begin()
        { return iterator( &t ); }
        ....

    };

This provides a value which is unique to the object and to the type of the array, but requires storing a T. Also, it would allow the user to mistakenly provide an initializer list with one element.

A slight variant could be returning *the* null pointer of type T

    return static_cast<T*>(0);

In this case the value would be unique to the type array<T, 0> but not to the objects (all objects of type array<T, 0> with the same value for T would yield the same pointer value).

Furthermore this is inconsistent with what the standard requires from allocation functions (see library issue 9).

c) same as above but with t being a static data member; again, the value would be unique to the type, not to the object.

d) to avoid storing a T *directly* while disallowing the possibility to use a one-element initializer list a non-aggregate nested class could be defined

    struct holder { holder() {} T t; } h;

and then begin be defined as

 iterator begin() { return &h.t; }

But then, it's arguable whether the array stores a T or not. Indirectly it does.

-----------------------------------------------------

Now, on different issues:

* what's the effect of calling assign(T&) on a zero-sized array? There seems to be only mention of front() and back(), in 23.2.1 [lib.array] p4 (I would also suggest to move that bullet to section 23.2.1.5 [lib.array.zero], for locality of reference)

* (minor) the opening paragraph of 23.2.1 [lib.array] wording is a bit inconsistent with that of other sequences: that's not a problem in itself, but compare it for instance with "A vector is a kind of sequence that supports random access iterators"; though the intent is obvious one might argue that the wording used for arrays doesn't tell what an array is, and relies on the reader to infer that it is what the <array> header defines.

* it would be desiderable to have a static const data member of type std::size_t, with value N, for usage as integral constant expression

* section 23.1 [lib.container.requirements] seem not to consider fixed-size containers at all, as it says: "[containers] control allocation and deallocation of these objects [the contained objects] through constructors, destructors, *insert and erase* operations"

* max_size() isn't specified: the result is obvious but, technically, it relies on table 80: "size() of the largest possible container" which, again, doesn't seem to consider fixed size containers

Proposed resolution:

[ Kona (2007): requirements on zero sized tr1::arrays and other details Issue 617: std::array is a sequence that doesn't satisfy the sequence requirements? Alisdair will prepare a paper. Proposed Disposition: Open ]


595. TR1/C++0x: fabs(complex<T>) redundant / wrongly specified

Section: 26.3.7 [complex.value.ops] Status: Ready Submitter: Stefan Große Pawig Date: 2006-09-24

View all issues with Ready status.

Discussion:

TR1 introduced, in the C compatibility chapter, the function fabs(complex<T>):

----- SNIP -----
8.1.1 Synopsis                                [tr.c99.cmplx.syn]

  namespace std {
  namespace tr1 {
[...]
  template<class T> complex<T> fabs(const complex<T>& x);
  } // namespace tr1
  } // namespace std

[...]

8.1.8 Function fabs                          [tr.c99.cmplx.fabs]

1 Effects: Behaves the same as C99 function cabs, defined in
  subclause 7.3.8.1.
----- SNIP -----

The current C++0X draft document (n2009.pdf) adopted this definition in chapter 26.3.1 (under the comment // 26.3.7 values) and 26.3.7/7.

But in C99 (ISO/IEC 9899:1999 as well as the 9899:TC2 draft document n1124), the referenced subclause reads

----- SNIP -----
7.3.8.1 The cabs functions

  Synopsis

1 #include <complex.h>
  double cabs(double complex z);
  float cabsf(float complex z);
  long double cabsl(long double z);

  Description

2 The cabs functions compute the complex absolute value (also called
  norm, modulus, or magnitude) of z.

  Returns

3 The cabs functions return the complex absolute value.
----- SNIP -----

Note that the return type of the cabs*() functions is not a complex type. Thus, they are equivalent to the already well established template<class T> T abs(const complex<T>& x); (26.2.7/2 in ISO/IEC 14882:1998, 26.3.7/2 in the current draft document n2009.pdf).

So either the return value of fabs() is specified wrongly, or fabs() does not behave the same as C99's cabs*().

Possible Resolutions

This depends on the intention behind the introduction of fabs().

If the intention was to provide a /complex/ valued function that calculates the magnitude of its argument, this should be explicitly specified. In TR1, the categorization under "C compatibility" is definitely wrong, since C99 does not provide such a complex valued function.

Also, it remains questionable if such a complex valued function is really needed, since complex<T> supports construction and assignment from real valued arguments. There is no difference in observable behaviour between

  complex<double> x, y;
  y = fabs(x);
  complex<double> z(fabs(x));

and

  complex<double> x, y;
  y = abs(x);
  complex<double> z(abs(x));

If on the other hand the intention was to provide the intended functionality of C99, fabs() should be either declared deprecated or (for C++0X) removed from the standard, since the functionality is already provided by the corresponding overloads of abs().

Proposed resolution:

Change the synopsis in 26.3.1 [complex.synopsis]:

template<class T> complex<T> fabs(const complex<T>&);

Change 26.3.7 [complex.value.ops], p7:

template<class T> complex<T> fabs(const complex<T>& x);

-7- Effects: Behaves the same as C99 function cabs, defined in subclause 7.3.8.1.

[ Kona (2007): Change the return type of fabs(complex) to T. Proposed Disposition: Ready ]


596. 27.8.1.3 Table 112 omits "a+" and "a+b" modes

Section: 27.8.1.4 [filebuf.members] Status: Review Submitter: Thomas Plum Date: 2006-09-26

View other active issues in [filebuf.members].

View all other issues in [filebuf.members].

View all issues with Review status.

Discussion:

In testing 27.8.1.4 [filebuf.members], Table 112 (in the latest N2009 draft), we invoke

   ostr.open("somename", ios_base::out | ios_base::in | ios_base::app)

and we expect the open to fail, because out|in|app is not listed in Table 92, and just before the table we see very specific words:

If mode is not some combination of flags shown in the table then the open fails.

But the corresponding table in the C standard, 7.19.5.3, provides two modes "a+" and "a+b", to which the C++ modes out|in|app and out|in|app|binary would presumably apply.

We would like to argue that the intent of Table 112 was to match the semantics of 7.19.5.3 and that the omission of "a+" and "a+b" was unintentional. (Otherwise there would be valid and useful behaviors available in C file I/O which are unavailable using C++, for no valid functional reason.)

We further request that the missing modes be explicitly restored to the WP, for inclusion in C++0x.

[ Martin adds: ]

...besides "a+" and "a+b" the C++ table is also missing a row for a lone app bit which in at least two current implementation as well as in Classic Iostreams corresponds to the C stdio "a" mode and has been traditionally documented as implying ios::out. Which means the table should also have a row for in|app meaning the same thing as "a+" already proposed in the issue.

Proposed resolution:

Add to the table "File open modes" in 27.8.1.4 [filebuf.members]:

File open modes
ios_base Flag combination stdio equivalent
binaryinouttruncapp 
    +     "w"
    +   + "a"
        + "a"
    + +   "w"
  +       "r"
  + +     "r+"
  + + +   "w+"
  + +   + "a+"
  +     + "a+"
+   +     "wb"
+   +   + "ab"
+       + "ab"
+   + +   "wb"
+ +       "rb"
+ + +     "r+b"
+ + + +   "w+b"
+ + +   + "a+b"
+ +     + "a+b"

[ Kona (2007) Added proposed wording and moved to Review. ]


597. Decimal: The notion of 'promotion' cannot be emulated by user-defined types.

Section: TRDecimal 3.2 [trdec.types.types] Status: Open Submitter: Daveed Vandevoorde Date: 2006-04-05

View other active issues in [trdec.types.types].

View all other issues in [trdec.types.types].

View all issues with Open status.

Discussion:

In a private email, Daveed writes:

I am not familiar with the C TR, but my guess is that the class type approach still won't match a built-in type approach because the notion of "promotion" cannot be emulated by user-defined types.

Here is an example:


		 struct S {
		   S(_Decimal32 const&);  // Converting constructor
		 };
		 void f(S);

		 void f(_Decimal64);

		 void g(_Decimal32 d) {
		   f(d);
		 }

If _Decimal32 is a built-in type, the call f(d) will likely resolve to f(_Decimal64) because that requires only a promotion, whereas f(S) requires a user-defined conversion.

If _Decimal32 is a class type, I think the call f(d) will be ambiguous because both the conversion to _Decimal64 and the conversion to S will be user-defined conversions with neither better than the other.

Robert comments:

In general, a library of arithmetic types cannot exactly emulate the behavior of the intrinsic numeric types. There are several ways to tell whether an implementation of the decimal types uses compiler intrinisics or a library. For example:

                 _Decimal32 d1;
                 d1.operator+=(5);  // If d1 is a builtin type, this won't compile.

In preparing the decimal TR, we have three options:

  1. require that the decimal types be class types
  2. require that the decimal types be builtin types, like float and double
  3. specify a library of class types, but allow enough implementor latitude that a conforming implementation could instead provide builtin types

We decided as a group to pursue option #3, but that approach implies that implementations may not agree on the semantics of certain use cases (first example, above), or on whether certain other cases are well-formed (second example). Another potentially important problem is that, under the present definition of POD, the decimal classes are not POD types, but builtins will be.

Note that neither example above implies any problems with respect to C-to-C++ compatibility, since neither example can be expressed in C.

Proposed resolution:


606. Decimal: allow narrowing conversions

Section: TRDecimal 3.2 [trdec.types.types] Status: Open Submitter: Martin Sebor Date: 2006-06-15

View other active issues in [trdec.types.types].

View all other issues in [trdec.types.types].

View all issues with Open status.

Discussion:

In c++std-lib-17205, Martin writes:

...was it a deliberate design choice to make narrowing assignments ill-formed while permitting narrowing compound assignments? For instance:

      decimal32 d32;
      decimal64 d64;

      d32 = 64;     // error
      d32 += 64;    // okay

In c++std-lib-17229, Robert responds:

It is a vestige of an old idea that I forgot to remove from the paper. Narrowing assignments should be permitted. The bug is that the converting constructors that cause narrowing should not be explicit. Thanks for pointing this out.

Proposed resolution:

1. In "3.2.2 Class decimal32" synopsis, remove the explicit specifier from the narrowing conversions:

                // 3.2.2.2 conversion from floating-point type:
                explicit decimal32(decimal64 d64);
                explicit decimal32(decimal128 d128);

2. Do the same thing in "3.2.2.2. Conversion from floating-point type."

3. In "3.2.3 Class decimal64" synopsis, remove the explicit specifier from the narrowing conversion:

                // 3.2.3.2 conversion from floating-point type:
                explicit decimal64(decimal128 d128);

4. Do the same thing in "3.2.3.2. Conversion from floating-point type."

[ Redmond: We prefer explicit conversions for narrowing and implicit for widening. ]


612. numeric_limits::is_modulo insufficently defined

Section: 18.2.1.2 [numeric.limits.members] Status: Open Submitter: Chris Jefferson Date: 2006-11-10

View all other issues in [numeric.limits.members].

View all issues with Open status.

Discussion:

18.2.1.2 55 states that "A type is modulo if it is possible to add two positive numbers together and have a result that wraps around to a third number that is less". This seems insufficent for the following reasons:

  1. Doesn't define what that value recieved is.
  2. Doesn't state the result is repeatable
  3. Doesn't require that doing addition, subtraction and other operations on all values is defined behaviour.

[ Batavia: Related to N2144. Pete: is there an ISO definition of modulo? Underflow on signed behavior is undefined. ]

Proposed resolution:

Suggest 18.2.1.2 [numeric.limits.members[numeric.limits.members], paragraph 57 is ammeded to:

A type is modulo if, it is possible to add two positive numbers and have a result that wraps around to a third number that is less. given any operation involving +,- or * on values of that type whose value would fall outside the range [min(), max()], then the value returned differs from the true value by an integer multiple of (max() - min() + 1).


614. std::string allocator requirements still inconsistent

Section: 21.3 [basic.string] Status: Open Submitter: Bo Persson Date: 2006-12-05

View other active issues in [basic.string].

View all other issues in [basic.string].

View all issues with Open status.

Discussion:

This is based on N2134, where 21.3.1/2 states: "... The Allocator object used shall be a copy of the Allocator object passed to the basic_string object's constructor or, if the constructor does not take an Allocator argument, a copy of a default-constructed Allocator object."

Section 21.3.2/1 lists two constructors:

basic_string(const basic_string<charT,traits,Allocator>& str );

basic_string(const basic_string<charT,traits,Allocator>& str ,
             size_type pos , size_type n = npos,
             const Allocator& a = Allocator());

and then says "In the first form, the Allocator value used is copied from str.get_allocator().", which isn't an option according to 21.3.1.

[ Batavia: We need blanket statement to the effect of: ]

  1. If an allocator is passed in, use it, or,
  2. If a string is passed in, use its allocator.

[ Review constructors and functions that return a string; make sure we follow these rules (substr, operator+, etc.). Howard to supply wording. ]

[ Bo adds: The new container constructor which takes only a size_type is not consistent with 23.1 [container.requirements], p9 which says in part:

All other constructors for these container types take an Allocator& argument (20.1.2), an allocator whose value type is the same as the container's value type. A copy of this argument is used for any memory allocation performed, by these constructors and by all member functions, during the lifetime of each container object.
]

Proposed resolution:


617. std::array is a sequence that doesn't satisfy the sequence requirements?

Section: 23.2.1 [array] Status: New Submitter: Bo Persson Date: 2006-12-30

View other active issues in [array].

View all other issues in [array].

View all issues with New status.

Discussion:

The <array> header is given under 23.2 [sequences]. 23.2.1 [array]/paragraph 3 says:

"Unless otherwise specified, all array operations are as described in 23.1 [container.requirements]".

However, array isn't mentioned at all in section 23.1 [container.requirements]. In particular, Table 82 "Sequence requirements" lists several operations (insert, erase, clear) that std::array does not have in 23.2.1 [array].

Also, Table 83 "Optional sequence operations" lists several operations that std::array does have, but array isn't mentioned.

Proposed resolution:


618. valarray::cshift() effects on empty array

Section: 26.5.2.7 [valarray.members] Status: Review Submitter: Gabriel Dos Reis Date: 2007-01-10

View all issues with Review status.

Discussion:

I would respectfully request an issue be opened with the intention to clarify the wording for size() == 0 for cshift.

Proposed resolution:

Change 26.5.2.7 [valarray.members], paragraph 10:

valarray<T> cshift(int n) const;

This function returns an object of class valarray<T>, of length size(), each of whose elements I is (*this)[(I + n ) % size()]. Thus, if element zero is taken as the leftmost element, a positive value of n shifts the elements circularly left n places. that is a circular shift of *this. If element zero is taken as the leftmost element, a non-negative value of n shifts the elements circularly left n places and a negative value of n shifts the elements circularly right -n places.

Rationale:

We do not believe that there is any real ambiguity about what happens when size() == 0, but we do believe that spelling this out as a C++ expression causes more trouble that it solves. The expression is certainly wrong when n < 0, since the sign of % with negative arguments is implementation defined.

[ Kona (2007) Changed proposed wording, added rationale and set to Review. ]


620. valid uses of empty valarrays

Section: 26.5.2.1 [valarray.cons] Status: Ready Submitter: Martin Sebor Date: 2007-01-20

View other active issues in [valarray.cons].

View all other issues in [valarray.cons].

View all issues with Ready status.

Discussion:

The Effects clause for the default valarray ctor suggests that it is possible to increase the size of an empty valarray object by calling other non-const member functions of the class besides resize(). However, such an interpretation would be contradicted by the requirement on the copy assignment operator (and apparently also that on the computed assignments) that the assigned arrays be the same size. See the reflector discussion starting with c++std-lib-17871.

In addition, Footnote 280 uses some questionable normative language.

Proposed resolution:

Reword the Effects clause and Footnote 280 as follows (26.5.2.1 [valarray.cons]):

valarray();

Effects: Constructs an object of class valarray<T>,279) which has zero length until it is passed into a library function as a modifiable lvalue or through a non-constant this pointer.280)

Postcondition: size() == 0.

Footnote 280: This default constructor is essential, since arrays of valarray are likely to prove useful. There shall also be a way to change the size of an array after initialization; this is supplied by the semantics may be useful. The length of an empty array can be increased after initialization by means of the resize() member function.


621. non-const copy assignment operators of helper arrays

Section: 26.5 [numarray] Status: Ready Submitter: Martin Sebor Date: 2007-01-20

View all other issues in [numarray].

View all issues with Ready status.

Discussion:

The computed and "fill" assignment operators of valarray helper array class templates (slice_array, gslice_array, mask_array, and indirect_array) are const member functions of each class template (the latter by the resolution of 123 since they have reference semantics and thus do not affect the state of the object on which they are called. However, the copy assignment operators of these class templates, which also have reference semantics, are non-const. The absence of constness opens the door to speculation about whether they really are intended to have reference semantics (existing implementations vary widely).

Pre-Kona, Martin adds:

I realized that adding the const qualifier to the functions as I suggested would break the const correctness of the classes. A few possible solutions come to mind:

  1. Add the const qualifier to the return types of these functions.
  2. Change the return type of all the functions to void to match the signatures of all the other assignment operators these classes define.
  3. Prohibit the copy assignment of these classes by declaring the copy assignment operators private (as is done and documented by some implementations).

Proposed resolution:

Declare the copy assignment operators of all four helper array class templates const.

Specifically, make the following edits:

Change the signature in 26.5.5 [template.slice.array] and 26.5.5.1 [slice.arr.assign] as follows:


const slice_array& operator= (const slice_array&) const;

        

Change the signature in 26.5.7 [template.gslice.array] and 26.5.7.1 [gslice.array.assign] as follows:


const gslice_array& operator= (const gslice_array&) const;

        

Change the signature in 26.5.8 [template.mask.array] and 26.5.8.1 [mask.array.assign] as follows:


const mask_array& operator= (const mask_array&) const;

        

Change the signature in 26.5.9 [template.indirect.array] and 26.5.9.1 [indirect.array.assign] as follows:


const indirect_array& operator= (const indirect_array&) const;

        

[ Kona (2007) Added const qualification to the return types and set to Ready. ]


622. behavior of filebuf dtor and close on error

Section: 27.8.1.17 [fstream.members] Status: Ready Submitter: Martin Sebor Date: 2007-01-20

View all issues with Ready status.

Discussion:

basic_filebuf dtor is specified to have the following straightforward effects:

Effects: Destroys an object of class basic_filebuf. Calls close().

close() does a lot of potentially complicated processing, including calling overflow() to write out the termination sequence (to bring the output sequence to its initial shift state). Since any of the functions called during the processing can throw an exception, what should the effects of an exception be on the dtor? Should the dtor catch and swallow it or should it propagate it to the caller? The text doesn't seem to provide any guidance in this regard other than the general restriction on throwing (but not propagating) exceptions from destructors of library classes in 17.4.4.8 [res.on.exception.handling].

Further, the last thing close() is specified to do is call fclose() to close the FILE pointer. The last sentence of the Effects clause reads:

... If any of the calls to overflow or std::fclose fails then close fails.

This suggests that close() might be required to call fclose() if and only if none of the calls to overflow() fails, and avoid closing the FILE otherwise. This way, if overflow() failed to flush out the data, the caller would have the opportunity to try to flush it again (perhaps after trying to deal with whatever problem may have caused the failure), rather than losing it outright.

On the other hand, the function's Postcondition specifies that is_open() == false, which suggests that it should call fclose() unconditionally. However, since Postcondition clauses are specified for many functions in the standard, including constructors where they obviously cannot apply after an exception, it's not clear whether this Postcondition clause is intended to apply even after an exception.

It might be worth noting that the traditional behavior (Classic Iostreams fstream::close() and C fclose()) is to close the FILE unconditionally, regardless of errors.

[ See 397 and 418 for related issues. ]

Proposed resolution:

After discussing this on the reflector (see the thread starting with c++std-lib-17650) we propose that close() be clarified to match the traditional behavior, that is to close the FILE unconditionally, even after errors or exceptions. In addition, we propose the dtor description be amended so as to explicitly require it to catch and swallow any exceptions thrown by close().

Specifically, we propose to make the following edits in 27.8.1.4 [filebuf.members]:


basic_filebuf<charT,traits>* close();

            

Effects: If is_open() == false, returns a null pointer. If a put area exists, calls overflow(traits::eof()) to flush characters. If the last virtual member function called on *this (between underflow, overflow, seekoff, and seekpos) was overflow then calls a_codecvt.unshift (possibly several times) to determine a termination sequence, inserts those characters and calls overflow(traits::eof()) again. Finally, regardless of whether any of the preceding calls fails or throws an exception, the function it closes the file ("as if" by calling std::fclose(file)).334) If any of the calls made by the functionto overflow or, including std::fclose, fails then close fails by returning a null pointer. If one of these calls throws an exception, the exception is caught and rethrown after closing the file.

And to make the following edits in 27.8.1.2 [filebuf.cons].


virtual ~basic_filebuf();

            

Effects: Destroys an object of class basic_filebuf<charT,traits>. Calls close(). If an exception occurs during the destruction of the object, including the call to close(), the exception is caught but not rethrown (see 17.4.4.8 [res.on.exception.handling]).


623. pubimbue forbidden to call imbue

Section: 27.1.1 [iostream.limits.imbue] Status: Ready Submitter: Martin Sebor Date: 2007-01-20

View all issues with Ready status.

Discussion:

27.1.1 [iostream.limits.imbue] specifies that "no function described in clause 27 except for ios_base::imbue causes any instance of basic_ios::imbue or basic_streambuf::imbue to be called."

That contradicts the Effects clause for basic_streambuf::pubimbue() which requires the function to do just that: call basic_streambuf::imbue().

Proposed resolution:

To fix this, rephrase the sentence above to allow pubimbue to do what it was designed to do. Specifically. change 27.1.1 [iostream.limits.imbue], p1 to read:

No function described in clause 27 except for ios_base::imbue and basic_filebuf::pubimbue causes any instance of basic_ios::imbue or basic_streambuf::imbue to be called. ...


624. valarray assignment and arrays of unequal length

Section: 26.5.2.2 [valarray.assign] Status: Ready Submitter: Martin Sebor Date: 2007-01-20

View all issues with Ready status.

Discussion:

The behavior of the valarray copy assignment operator is defined only when both sides have the same number of elements and the spec is explicit about assignments of arrays of unequal lengths having undefined behavior.

However, the generalized subscripting assignment operators overloaded on slice_array et al (26.5.2.2 [valarray.assign]) don't have any such restriction, leading the reader to believe that the behavior of these overloads is well defined regardless of the lengths of the arguments.

For example, based on the reading of the spec the behavior of the snippet below can be expected to be well-defined:

    const std::slice from_0_to_3 (0, 3, 1);   // refers to elements 0, 1, 2
    const std::valarray<int> a (1, 3);        // a = { 1, 1, 1 }
    std::valarray<int>       b (2, 4);        // b = { 2, 2, 2, 2 }

    b = a [from_0_to_3];
        

In practice, b may end up being { 1, 1, 1 }, { 1, 1, 1, 2 }, or anything else, indicating that existing implementations vary.

Quoting from Section 3.4, Assignment operators, of Al Vermeulen's Proposal for Standard C++ Array Classes (see c++std-lib-704; N0308):

...if the size of the array on the right hand side of the equal sign differs from the size of the array on the left, a run time error occurs. How this error is handled is implementation dependent; for compilers which support it, throwing an exception would be reasonable.

And see more history in N0280.

It has been argued in discussions on the committee's reflector that the semantics of all valarray assignment operators should be permitted to be undefined unless the length of the arrays being assigned is the same as the length of the one being assigned from. See the thread starting at c++std-lib-17786.

In order to reflect such views, the standard must specify that the size of the array referred to by the argument of the assignment must match the size of the array under assignment, for example by adding a Requires clause to 26.5.2.2 [valarray.assign] as follows:

Requires: The length of the array to which the argument refers equals size().

Note that it's far from clear that such leeway is necessary in order to implement valarray efficiently.

Proposed resolution:

Insert new paragraph into 26.5.2.2 [valarray.assign]:

valarray<T>& operator=(const slice_array<T>&); 
valarray<T>& operator=(const gslice_array<T>&); 
valarray<T>& operator=(const mask_array<T>&); 
valarray<T>& operator=(const indirect_array<T>&);

Requires: The length of the array to which the argument refers equals size().

These operators allow the results of a generalized subscripting operation to be assigned directly to a valarray.


625. mixed up Effects and Returns clauses

Section: 17 [library] Status: Open Submitter: Martin Sebor Date: 2007-01-20

View all other issues in [library].

View all issues with Open status.

Discussion:

Many member functions of basic_string are overloaded, with some of the overloads taking a string argument, others value_type*, others size_type, and others still iterators. Often, the requirements on one of the overloads are expressed in the form of Effects, Throws, and in the Working Paper (N2134) also Remark clauses, while those on the rest of the overloads via a reference to this overload and using a Returns clause.

The difference between the two forms of specification is that per 17.3.1.3 [structure.specifications], p3, an Effects clause specifies "actions performed by the functions," i.e., its observable effects, while a Returns clause is "a description of the return value(s) of a function" that does not impose any requirements on the function's observable effects.

Since only Notes are explicitly defined to be informative and all other paragraphs are explicitly defined to be normative, like Effects and Returns, the new Remark clauses also impose normative requirements.

So by this strict reading of the standard there are some member functions of basic_string that are required to throw an exception under some conditions or use specific traits members while many other otherwise equivalent overloads, while obliged to return the same values, aren't required to follow the exact same requirements with regards to the observable effects.

Here's an example of this problem that was precipitated by the change from informative Notes to normative Remarks (presumably made to address 424):

In the Working Paper, find(string, size_type) contains a Remark clause (which is just a Note in the current standard) requiring it to use traits::eq().

find(const charT *s, size_type pos) is specified to return find(string(s), pos) by a Returns clause and so it is not required to use traits::eq(). However, the Working Paper has replaced the original informative Note about the function using traits::length() with a normative requirement in the form of a Remark. Calling traits::length() may be suboptimal, for example when the argument is a very long array whose initial substring doesn't appear anywhere in *this.

Here's another similar example, one that existed even prior to the introduction of Remarks:

insert(size_type pos, string, size_type, size_type) is required to throw out_of_range if pos > size().

insert(size_type pos, string str) is specified to return insert(pos, str, 0, npos) by a Returns clause and so its effects when pos > size() are strictly speaking unspecified.

I believe a careful review of the current Effects and Returns clauses is needed in order to identify all such problematic cases. In addition, a review of the Working Paper should be done to make sure that the newly introduced normative Remark clauses do not impose any undesirable normative requirements in place of the original informative Notes.

[ Batavia: Alan and Pete to work. ]

Proposed resolution:


626. new Remark clauses not documented

Section: 17.3.1.3 [structure.specifications] Status: Open Submitter: Martin Sebor Date: 2007-01-20

View all other issues in [structure.specifications].

View all issues with Open status.

Discussion:

The Remark clauses newly introduced into the Working Paper (N2134) are not mentioned in 17.3.1.3 [structure.specifications] where we list the meaning of Effects, Requires, and other clauses (with the exception of Notes which are documented as informative in 17.3.1.1 [structure.summary], p2, and which they replace in many cases).

Propose add a bullet for Remarks along with a brief description.

[ Batavia: Alan and Pete to work. ]

Proposed resolution:


627. Low memory and exceptions

Section: 18.5.1.1 [new.delete.single] Status: Open Submitter: P.J. Plauger Date: 2007-01-23

View all other issues in [new.delete.single].

View all issues with Open status.

Discussion:

I recognize the need for nothrow guarantees in the exception reporting mechanism, but I strongly believe that implementors also need an escape hatch when memory gets really low. (Like, there's not enough heap to construct and copy exception objects, or not enough stack to process the throw.) I'd like to think we can put this escape hatch in 18.5.1.1 [new.delete.single], operator new, but I'm not sure how to do it. We need more than a footnote, but the wording has to be a bit vague. The idea is that if new can't allocate something sufficiently small, it has the right to abort/call terminate/call unexpected.

Proposed resolution:


629. complex insertion and locale dependence

Section: 26.3.6 [complex.ops] Status: Open Submitter: Gabriel Dos Reis Date: 2007-01-28

View all other issues in [complex.ops].

View all issues with Open status.

Discussion:

is there an issue opened for (0,3) as complex number with the French local? With the English local, the above parses as an imaginery complex number. With the French locale it parses as a real complex number.

Further notes/ideas from the lib-reflector, messages 17982-17984:

Add additional entries in num_punct to cover the complex separator (French would be ';').

Insert a space before the comma, which should eliminate the ambiguity.

Solve the problem for ordered sequences in general, perhaps with a dedicated facet. Then complex should use that solution.

Proposed resolution:


630. arrays of valarray

Section: 26.5.2.1 [valarray.cons] Status: Open Submitter: Martin Sebor Date: 2007-01-28

View other active issues in [valarray.cons].

View all other issues in [valarray.cons].

View all issues with Open status.

Discussion:

Section 26.1 [numeric.requirements], p1 suggests that a valarray specialization on a type T that satisfies the requirements enumerated in the paragraph is itself a valid type on which valarray may be instantiated (Footnote 269 makes this clear). I.e., valarray<valarray<T> > is valid as long as T is valid. However, since implementations of valarray are permitted to initialize storage allocated by the class by invoking the default ctor of T followed by the copy assignment operator, such implementations of valarray wouldn't work with (perhaps user-defined) specializations of valarray whose assignment operator had undefined behavior when the size of its argument didn't match the size of *this. By "wouldn't work" I mean that it would be impossible to resize such an array of arrays by calling the resize() member function on it if the function used the copy assignment operator after constructing all elements using the default ctor (e.g., by invoking new value_type[N]) to obtain default-initialized storage) as it's permitted to do.

Stated more generally, the problem is that valarray<valarray<T> >::resize(size_t) isn't required or guaranteed to have well-defined semantics for every type T that satisfies all requirements in 26.1 [numeric.requirements].

I believe this problem was introduced by the adoption of the resolution outlined in N0857, Assignment of valarrays, from 1996. The copy assignment operator of the original numerical array classes proposed in N0280, as well as the one proposed in N0308 (both from 1993), had well-defined semantics for arrays of unequal size (the latter explicitly only when *this was empty; assignment of non empty arrays of unequal size was a runtime error).

The justification for the change given in N0857 was the "loss of performance [deemed] only significant for very simple operations on small arrays or for architectures with very few registers."

Since tiny arrays on a limited subset of hardware architectures are likely to be an exceedingly rare case (despite the continued popularity of x86) I propose to revert the resolution and make the behavior of all valarray assignment operators well-defined even for non-conformal arrays (i.e., arrays of unequal size). I have implemented this change and measured no significant degradation in performance in the common case (non-empty arrays of equal size). I have measured a 50% (and in some cases even greater) speedup in the case of assignments to empty arrays versus calling resize() first followed by an invocation of the copy assignment operator.

Proposed resolution:

Change 26.5.2.2 [valarray.assign], p1 as follows:

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

-1- Each element of the *this array is assigned the value of the corresponding element of the argument array. The resulting behavior is undefined if When the length of the argument array is not equal to the length of the *this array. resizes *this to make the two arrays the same length, as if by calling resize(x.size()), before performing the assignment.

And add a new paragraph just below paragraph 1 with the following text:

-2- Postcondition: size() == x.size().

Also add the following paragraph to 26.5.2.2 [valarray.assign], immediately after p4:

-?- When the length, N of the array referred to by the argument is not equal to the length of *this, the operator resizes *this to make the two arrays the same length, as if by calling resize(N), before performing the assignment.

[ Kona (2007): Gaby to propose wording for an alternative resolution in which you can assign to a valarray of size 0, but not to any other valarray whose size is unequal to the right hand side of the assignment. ]


631. conflicting requirements for BinaryPredicate

Section: 25 [algorithms] Status: Open Submitter: James Kanze Date: 2007-01-31

View all other issues in [algorithms].

View all issues with Open status.

Discussion:

The general requirements for BinaryPredicate (in 25 [algorithms]/8) contradict the implied specific requirements for some functions. In particular, it says that:

[...] if an algorithm takes BinaryPredicate binary_pred as its argument and first1 and first2 as its iterator arguments, it should work correctly in the construct if (binary_pred (*first1 , *first2 )){...}. BinaryPredicate always takes the first iterator type as its first argument, that is, in those cases when T value is part of the signature, it should work correctly in the context of if (binary_pred (*first1 , value)){...}.

In the description of upper_bound (25.3.3.2 [upper.bound]/2), however, the use is described as "!comp(value, e)", where e is an element of the sequence (a result of dereferencing *first).

In the description of lexicographical_compare, we have both "*first1 < *first2" and "*first2 < *first1" (which presumably implies "comp( *first1, *first2 )" and "comp( *first2, *first1 )".

[ Toronto: Moved to Open. ConceptGCC seems to get lower_bound and upper_bound to work withoutt these changes. ]

Proposed resolution:

Logically, the BinaryPredicate is used as an ordering relationship, with the semantics of "less than". Depending on the function, it may be used to determine equality, or any of the inequality relationships; doing this requires being able to use it with either parameter first. I would thus suggest that the requirement be:

[...] BinaryPredicate always takes the first iterator value_type as one of its arguments, it is unspecified which. If an algorithm takes BinaryPredicate binary_pred as its argument and first1 and first2 as its iterator arguments, it should work correctly both in the construct if (binary_pred (*first1 , *first2 )){...} and if (binary_pred (*first2, *first1)){...}. In those cases when T value is part of the signature, it should work correctly in the context of if (binary_pred (*first1 , value)){...} and of if (binary_pred (value, *first1)){...}. [Note: if the two types are not identical, and neither is convertable to the other, this may require that the BinaryPredicate be a functional object with two overloaded operator()() functions. --end note]

Alternatively, one could specify an order for each function. IMHO, this would be more work for the committee, more work for the implementors, and of no real advantage for the user: some functions, such as lexicographical_compare or equal_range, will still require both functions, and it seems like a much easier rule to teach that both functions are always required, rather than to have a complicated list of when you only need one, and which one.


632. Time complexity of size() for std::set

Section: 23.1 [container.requirements] Status: Open Submitter: Lionel B Date: 2007-02-01

View other active issues in [container.requirements].

View all other issues in [container.requirements].

View all issues with Open status.

Discussion:

A recent news group discussion:

Anyone know if the Standard has anything to say about the time complexity of size() for std::set? I need to access a set's size (/not/ to know if it is empty!) heavily during an algorithm and was thus wondering whether I'd be better off tracking the size "manually" or whether that'd be pointless.

That would be pointless. size() is O(1).

Nit: the standard says "should" have constant time. Implementations may take license to do worse. I know that some do this for std::list<> as a part of some trade-off with other operation.

I was aware of that, hence my reluctance to use size() for std::set.

However, this reason would not apply to std::set<> as far as I can see.

Ok, I guess the only option is to try it and see...

If I have any recommendation to the C++ Standards Committee it is that implementations must (not "should"!) document clearly[1], where known, the time complexity of *all* container access operations.

[1] In my case (gcc 4.1.1) I can't swear that the time complexity of size() for std::set is not documented... but if it is it's certainly well hidden away.

Proposed resolution:

[ Kona (2007): This issue affects all the containers. We'd love to see a paper dealing with the broad issue. We think that the complexity of the size() member of every container -- except possibly list -- should be O(1). Alan has volunteered to provide wording. ]


635. domain of allocator::address

Section: 20.1.2 [allocator.requirements] Status: Open Submitter: Howard Hinnant Date: 2007-02-08

View other active issues in [allocator.requirements].

View all other issues in [allocator.requirements].

View all issues with Open status.

Discussion:

The table of allocator requirements in 20.1.2 [allocator.requirements] describes allocator::address as:

a.address(r)
a.address(s)

where r and s are described as:

a value of type X::reference obtained by the expression *p.

and p is

a value of type X::pointer, obtained by calling a1.allocate, where a1 == a

This all implies that to get the address of some value of type T that value must have been allocated by this allocator or a copy of it.

However sometimes container code needs to compare the address of an external value of type T with an internal value. For example list::remove(const T& t) may want to compare the address of the external value t with that of a value stored within the list. Similarly vector or deque insert may want to make similar comparisons (to check for self-referencing calls).

Mandating that allocator::address can only be called for values which the allocator allocated seems overly restrictive.

Proposed resolution:

Change 20.1.2 [allocator.requirements]:

r : a value of type X::reference obtained by the expression *p.

s : a value of type X::const_reference obtained by the expression *q or by conversion from a value r.

[ post Oxford: This would be rendered NAD Editorial by acceptance of N2257. ]

[ Kona (2007): This issue is section 8 of N2387. There was some discussion of it but no resolution to this issue was recorded. Moved to Open. ]


638. deque end invalidation during erase

Section: 23.2.2.3 [deque.modifiers] Status: Review Submitter: Steve LoBasso Date: 2007-02-17

View all issues with Review status.

Discussion:

The standard states at 23.2.2.3 [deque.modifiers]/4:

deque erase(...)

Effects: ... An erase at either end of the deque invalidates only the iterators and the references to the erased elements.

This does not state that iterators to end will be invalidated. It needs to be amended in such a way as to account for end invalidation.

Something like:

Any time the last element is erased, iterators to end are invalidated.

This would handle situations like:

erase(begin(), end())
erase(end() - 1)
pop_back()
resize(n, ...) where n < size()
pop_front() with size() == 1

[ Post Kona, Steve LoBasso notes: ]

My only issue with the proposed resolution is that it might not be clear that pop_front() [where size() == 1] can invalidate past-the-end iterators.

Proposed resolution:

Change 23.2.2.3 [deque.modifiers], p4:

iterator erase(const_iterator position); 
iterator erase(const_iterator first, const_iterator last);

-4- Effects: An erase in the middle of the deque invalidates all the iterators and references to elements of the deque and the past-the-end iterator. An erase at either end of the deque invalidates only the iterators and the references to the erased elements, except that erasing at the end also invalidates the past-the-end iterator.

[ Kona (2007): Proposed wording added and moved to Review. ]


645. Missing members in match_results

Section: 28.10 [re.results] Status: Review Submitter: Daniel Krügler Date: 2007-02-26

View other active issues in [re.results].

View all other issues in [re.results].

View all issues with Review status.

Discussion:

According to the description given in 28.10 [re.results]/2 the class template match_results "shall satisfy the requirements of a Sequence, [..], except that only operations defined for const-qualified Sequences are supported". Comparing the provided operations from 28.10 [re.results]/3 with the sequence/container tables 80 and 81 one recognizes the following missing operations:

1) The members

const_iterator rbegin() const;
const_iterator rend() const;

should exists because 23.1/10 demands these for containers (all sequences are containers) which support bidirectional iterators. Aren't these supported by match_result? This is not explicitely expressed, but it's somewhat implied by two arguments:

(a) Several typedefs delegate to iterator_traits<BidirectionalIterator>.

(b) The existence of const_reference operator[](size_type n) const implies even random-access iteration. I also suggest, that match_result should explicitly mention, which minimum iterator category is supported and if this does not include random-access the existence of operator[] is somewhat questionable.

2) The new "convenience" members

const_iterator cbegin() const;
const_iterator cend() const;
const_iterator crbegin() const;
const_iterator crend() const;

should be added according to tables 80/81.

Proposed resolution:

Add the following members to the match_results synopsis after end() in 28.10 [re.results] para 3:

const_iterator cbegin() const; 
const_iterator cend() const;

In section 28.10.3 [re.results.acc] change:

const_iterator begin() const;
const_iterator cbegin() const;

-7- Returns: A starting iterator that enumerates over all the sub-expressions stored in *this.

const_iterator end() const;
const_iterator cend() const;

-8- Returns: A terminating iterator that enumerates over all the sub-expressions stored in *this.

[ Kona (2007): Voted to adopt proposed wording in N2409 except removing the entry in the table container requirements. Moved to Review. ]


653. Library reserved names

Section: 1.2 [intro.refs] Status: Open Submitter: Alisdair Meredith Date: 2007-03-08

View all other issues in [intro.refs].

View all issues with Open status.

Discussion:

1.2 [intro.refs] Normative references

The following standards contain provisions which, through reference in this text, constitute provisions of this Interna- tional Standard. At the time of publication, the editions indicated were valid. All standards are subject to revision, and parties to agreements based on this International Standard are encouraged to investigate the possibility of applying the most recent editions of the standards indicated below. Members of IEC and ISO maintain registers of currently valid International Standards.

I'm not sure how many of those reserve naming patterns that might affect us, but I am equally sure I don't own a copy of any of these to check!

The point is to list the reserved naming patterns, rather than the individual names themselves - although we may want to list C keywords that are valid identifiers in C++ but likely to cause trouble in shared headers (e.g. restrict)

[ Kona (2007): Recommend NAD. No one has identified a specific defect, just the possibility of one. ]

[ Post-Kona: Alisdair request Open. A good example of the problem was a discussion of the system error proposal, where it was pointed out an all-caps identifier starting with a capital E conflicted with reserved macro names for both Posix and C. I had absolutely no idea of this rule, and suspect I was not the only one in the room.

Resolution will require someone with access to all the listed documents to research their respective name reservation rules, or people with access to specific documents add their rules to this issue until the list is complete. ]

Proposed resolution:


659. istreambuf_iterator should have an operator->()

Section: 24.5.3 [istreambuf.iterator] Status: Open Submitter: Niels Dekker Date: 2007-03-25

View all other issues in [istreambuf.iterator].

View all issues with Open status.

Discussion:

Greg Herlihy has clearly demonstrated that a user defined input iterator should have an operator->(), even if its value type is a built-in type (comp.std.c++, "Re: Should any iterator have an operator->() in C++0x?", March 2007).  And as Howard Hinnant remarked in the same thread that the input iterator istreambuf_iterator doesn't have one, this must be a defect!

Based on Greg's example, the following code demonstrates the issue:

 #include <iostream> 
 #include <fstream>
 #include <streambuf> 

 typedef char C;
 int main ()
 {
   std::ifstream s("filename", std::ios::in);
   std::istreambuf_iterator<char> i(s);

   (*i).~C();  // This is well-formed...
   i->~C();  // ... so this should be supported!
 }

Of course, operator-> is also needed when the value_type of istreambuf_iterator is a class.

The operator-> could be implemented in various ways.  For instance, by storing the current value inside the iterator, and returning its address.  Or by returning a proxy, like operator_arrow_proxy, from http://www.boost.org/boost/iterator/iterator_facade.hpp

I hope that the resolution of this issue will contribute to getting a clear and consistent definition of iterator concepts.

Proposed resolution:

Add to the synopsis in 24.5.3 [istreambuf.iterator]:

charT operator*() const;
pointer operator->() const;
istreambuf_iterator<charT,traits>& operator++();

Change 24.5.3 [istreambuf.iterator], p1:

The class template istreambuf_iterator reads successive characters from the streambuf for which it was constructed. operator* provides access to the current input character, if any. operator-> may return a proxy. Each time operator++ is evaluated, the iterator advances to the next input character. If the end of stream is reached (streambuf_type::sgetc() returns traits::eof()), the iterator becomes equal to the end of stream iterator value. The default constructor istreambuf_iterator() and the constructor istreambuf_iterator(0) both construct an end of stream iterator object suitable for use as an end-of-range.

[ Kona (2007): The proposed resolution is inconsistent because the return type of istreambuf_iterator::operator->() is specified to be pointer, but the proposed text also states that "operator-> may return a proxy." ]

[ Niels Dekker (mailed to Howard Hinnant): ]

The proposed resolution does not seem inconsistent to me. istreambuf_iterator::operator->() should have istreambuf_iterator::pointer as return type, and this return type may in fact be a proxy.

AFAIK, the resolution of 445 ("iterator_traits::reference unspecified for some iterator categories") implies that for any iterator class Iter, the return type of operator->() is Iter::pointer, by definition.  I don't think Iter::pointer needs to be a raw pointer.

Still I wouldn't mind if the text "operator-> may return a proxy" would be removed from the resolution. I think it's up to the library implementation, how to implement istreambuf_iterator::operator->().  As longs as it behaves as expected: i->m should have the same effect as (*i).m. Even for an explicit destructor call, i->~C().  The main issue is just: istreambuf_iterator should have an operator->()!


661. New 27.6.1.2.2 changes make special extractions useless

Section: 27.6.1.2.2 [istream.formatted.arithmetic] Status: Ready Submitter: Daniel Krügler Date: 2007-04-01

View other active issues in [istream.formatted.arithmetic].

View all other issues in [istream.formatted.arithmetic].

View all issues with Ready status.

Discussion:

To the more drastic changes of 27.6.1.2.2 [istream.formatted.arithmetic] in the current draft N2134 belong the explicit description of the extraction of the types short and int in terms of as-if code fragments.

  1. The corresponding as-if extractions in paragraph 2 and 3 will never result in a change of the operator>> argument val, because the contents of the local variable lval is in no case written into val. Furtheron both fragments need a currently missing parentheses in the beginning of the if-statement to be valid C++.
  2. I would like to ask whether the omission of a similar explicit extraction of unsigned short and unsigned int in terms of long - compared to their corresponding new insertions, as described in 27.6.2.6.2 [ostream.inserters.arithmetic], is a deliberate decision or an oversight.

Proposed resolution:

  1. In 27.6.1.2.2 [istream.formatted.arithmetic]/2 change the current as-if 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) {
      && if (lval < numeric_limits<short>::min() || numeric_limits<short>::max() < lval))
          err = ios_base::failbit;
      else
        val = static_cast<short>(lval);
    }
    setstate(err);
    

    Similarily in 27.6.1.2.2 [istream.formatted.arithmetic]/3 change the current as-if 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) {
      && if (lval < numeric_limits<int>::min() || numeric_limits<int>::max() < lval))
          err = ios_base::failbit;
      else
        val = static_cast<int>(lval);
    }
    setstate(err);
    
  2. ---

[ Kona (2007): Note to the editor: the name lval in the call to use_facet is incorrectly italicized in the code fragments corresponding to operator>>(short &) and operator >>(int &). Also, val -- which appears twice on the line with the static_cast in the proposed resolution -- should be italicized. Also, in response to part two of the issue: this is deliberate. ]


664. do_unshift for codecvt<char, char, mbstate_t>

Section: 22.2.1.4.2 [locale.codecvt.virtuals] Status: Ready Submitter: Thomas Plum Date: 2007-04-16

View other active issues in [locale.codecvt.virtuals].

View all other issues in [locale.codecvt.virtuals].

View all issues with Ready status.

Discussion:

22.2.1.4.2 [locale.codecvt.virtuals], para 7 says (regarding do_unshift):

Effects: Places characters starting at to that should be appended to terminate a sequence when the current stateT is given by state.237) 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.

The following objection has been raised:

Since the C++ Standard permits a nontrivial conversion for the required instantiations of codecvt, it is overly restrictive to say that do_unshift must store no characters and return noconv.

[Plum ref _222152Y50]

Proposed resolution:

Change 22.2.1.4.2 [locale.codecvt.virtuals], p7:

Effects: Places characters starting at to that should be appended to terminate a sequence when the current stateT is given by state.237) 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.


665. do_unshift return value

Section: 22.2.1.4.2 [locale.codecvt.virtuals] Status: Ready Submitter: Thomas Plum Date: 2007-04-16

View other active issues in [locale.codecvt.virtuals].

View all other issues in [locale.codecvt.virtuals].

View all issues with Ready status.

Discussion:

22.2.1.4.2 [locale.codecvt.virtuals], para 8 says:

codecvt<char,char,mbstate_t>, returns noconv.

The following objection has been raised:

Despite what the C++ Standard  says, unshift can't always return noconv for the default facets, since  they can be nontrivial. At least one implementation does whatever the  C functions do.

[Plum ref _222152Y62]

Proposed resolution:

Change 22.2.1.4.2 [locale.codecvt.virtuals], p8:

Returns: An enumeration value, as summarized in Table 76:

...

codecvt<char,char,mbstate_t>, returns noconv.


666. moneypunct::do_curr_symbol()

Section: 22.2.6.3.2 [locale.moneypunct.virtuals] Status: Ready Submitter: Thomas Plum Date: 2007-04-16

View all other issues in [locale.moneypunct.virtuals].

View all issues with Ready status.

Discussion:

22.2.6.3.2 [locale.moneypunct.virtuals], para 4 footnote 257 says

257) For international  specializations (second template parameter true) this is always four  characters long, usually three letters and a space.

The following objection has been raised:

The international currency  symbol is whatever the underlying locale says it is, not necessarily  four characters long.

[Plum ref _222632Y41]

Proposed resolution:

Change footnote 253 in 22.2.6.3.2 [locale.moneypunct.virtuals]:

253) For international specializations (second template parameter true) this is always typically four characters long, usually three letters and a space.


667. money_get's widened minus sign

Section: 22.2.6.1.2 [locale.money.get.virtuals] Status: Open Submitter: Thomas Plum Date: 2007-04-16

View other active issues in [locale.money.get.virtuals].

View all other issues in [locale.money.get.virtuals].

View all issues with Open status.

Discussion:

22.2.6.1.2 [locale.money.get.virtuals], para 1 says:

The result is returned as an integral value  stored in units or as a sequence of digits possibly preceded by a  minus sign (as produced by ct.widen(c) where c is '-' or in the range  from '0' through '9', inclusive) stored in digits.

The following objection has been raised:

Some implementations interpret this to mean that a facet derived from ctype<wchar_t> can provide its own member do_widen(char) which produces e.g. L'@' for the "widened" minus sign, and that the '@' symbol will appear in the resulting sequence of digits.  Other implementations have assumed that one or more places in the standard permit the implementation to "hard-wire" L'-' as the "widened" minus sign.  Are both interpretations permissible, or only  one?

[Plum ref _222612Y14]

Furthermore: if ct.widen('9') produces L'X' (a non-digit), does a parse fail if a '9' appears in the subject string? [Plum ref _22263Y33]

[ Kona (2007): Bill and Dietmar to provide proposed wording. ]

Proposed resolution:


668. money_get's empty minus sign

Section: 22.2.6.1.2 [locale.money.get.virtuals] Status: Open Submitter: Thomas Plum Date: 2007-04-16

View other active issues in [locale.money.get.virtuals].

View all other issues in [locale.money.get.virtuals].

View all issues with Open status.

Discussion:

22.2.6.1.2 [locale.money.get.virtuals], para 3 says:

If pos or neg is empty, the sign component is optional, and if no sign is detected, the result is given the sign  that corresponds to the source of the empty string.

The following objection has been raised:

A negative_sign of "" means "there is no  way to write a negative sign" not "any null sequence is a negative  sign, so it's always there when you look for it".

[Plum ref _222612Y32]

[ Kona (2007): Bill to provide proposed wording and interpretation of existing wording. ]

Proposed resolution:


669. Equivalent postive and negative signs in money_get

Section: 22.2.6.1.2 [locale.money.get.virtuals] Status: Open Submitter: Thomas Plum Date: 2007-04-16

View other active issues in [locale.money.get.virtuals].

View all other issues in [locale.money.get.virtuals].

View all issues with Open status.

Discussion:

22.2.6.1.2 [locale.money.get.virtuals], para 3 sentence 4 says:

If the first character of pos is equal to the first character of neg,  or if both strings are empty, the result is given a positive sign.

One interpretation is that an input sequence must match either the positive pattern or the negative pattern, and then in either event it is interpreted as positive.  The following objections has been raised:

The input can successfully match only a positive sign, so the negative pattern is an unsuccessful match.

[Plum ref _222612Y34, 222612Y51b]

[ Bill to provide proposed wording and interpretation of existing wording. ]

Proposed resolution:


670. money_base::pattern and space

Section: 22.2.6.3 [locale.moneypunct] Status: Open Submitter: Thomas Plum Date: 2007-04-16

View all issues with Open status.

Discussion:

22.2.6.3 [locale.moneypunct], para 2 says:

The value space indicates that at least one space is required at  that position.

The following objection has been raised:

Whitespace is optional when matching space. (See 22.2.6.1.2 [locale.money.get.virtuals], para 2.)

[Plum ref _22263Y22]

[ Kona (2007): Bill to provide proposed wording. We agree that C++03 is ambiguous, and that we want C++0X to say "space" means 0 or more whitespace characters on input. ]

Proposed resolution:


671. precision of hexfloat

Section: 22.2.2.2.2 [facet.num.put.virtuals] Status: Open Submitter: John Salmon Date: 2007-04-20

View all other issues in [facet.num.put.virtuals].

View all issues with Open status.

Discussion:

I am trying to understand how TR1 supports hex float (%a) output.

As far as I can tell, it does so via the following:

8.15 Additions to header <locale> [tr.c99.locale]

In subclause 22.2.2.2.2 [facet.num.put.virtuals], Table 58 Floating-point conversions, after the line: floatfield == ios_base::scientific %E

add the two lines:

floatfield == ios_base::fixed | ios_base::scientific && !uppercase %a
floatfield == ios_base::fixed | ios_base::scientific %A 2

[Note: The additional requirements on print and scan functions, later in this clause, ensure that the print functions generate hexadecimal floating-point fields with a %a or %A conversion specifier, and that the scan functions match hexadecimal floating-point fields with a %g conversion specifier.  end note]

Following the thread, in 22.2.2.2.2 [facet.num.put.virtuals], we find:

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 would seem to imply that when floatfield == fixed|scientific, the precision of the conversion specifier is to be taken from str.precision().  Is this really what's intended?  I sincerely hope that I'm either missing something or this is an oversight.  Please tell me that the committee did not intend to mandate that hex floats (and doubles) should by default be printed as if by %.6a.

[ Howard: I think the fundamental issue we overlooked was that with %f, %e, %g, the default precision was always 6.  With %a the default precision is not 6, it is infinity.  So for the first time, we need to distinguish between the default value of precision, and the precision value 6. ]

Proposed resolution:

[ Kona (2007): Robert volunteers to propose wording. ]


672. Swappable requirements need updating

Section: 20.1.1 [utility.arg.requirements] Status: Review Submitter: Howard Hinnant Date: 2007-05-04

View other active issues in [utility.arg.requirements].

View all other issues in [utility.arg.requirements].

View all issues with Review status.

Discussion:

The current Swappable is:

Table 37: Swappable requirements [swappable]
expressionreturn typepost-condition
swap(s,t)voidt has the value originally held by u, and u has the value originally held by t

The Swappable requirement is met by satisfying one or more of the following conditions:

  • T is Swappable if T satisfies the CopyConstructible requirements (Table 34) and the CopyAssignable requirements (Table 36);
  • T is Swappable if a namespace scope function named swap exists in the same namespace as the definition of T, such that the expression swap(t,u) is valid and has the semantics described in this table.

With the passage of rvalue reference into the language, Swappable needs to be updated to require only MoveConstructible and MoveAssignable. This is a minimum.

Additionally we may want to support proxy references such that the following code is acceptable:

namespace Mine {

template <class T>
struct proxy {...};

template <class T>
struct proxied_iterator
{
   typedef T value_type;
   typedef proxy<T> reference;
   reference operator*() const;
   ...
};

struct A
{
   // heavy type, has an optimized swap, maybe isn't even copyable or movable, just swappable
   void swap(A&);
   ...
};

void swap(A&, A&);
void swap(proxy<A>, A&);
void swap(A&, proxy<A>);
void swap(proxy<A>, proxy<A>);

}  // Mine

...

Mine::proxied_iterator<Mine::A> i(...)
Mine::A a;
swap(*i1, a);

I.e. here is a call to swap which the user enables swapping between a proxy to a class and the class itself. We do not need to anything in terms of implementation except not block their way with overly constrained concepts. That is, the Swappable concept should be expanded to allow swapping between two different types for the case that one is binding to a user-defined swap.

Proposed resolution:

Change 20.1.1 [utility.arg.requirements]:

-1- The template definitions in the C++ Standard Library refer to various named requirements whose details are set out in tables 31-38. In these tables, T is a type to be supplied by a C++ program instantiating a template; a, b, and c are values of type const T; s and t are modifiable lvalues of type T; u is a value of type (possibly const) T; and rv is a non-const rvalue of type T.

Table 37: Swappable requirements [swappable]
expressionreturn typepost-condition
swap(s,t)void t has the value originally held by u, and u has the value originally held by t

The Swappable requirement is met by satisfying one or more of the following conditions:

  • T is Swappable if T satisfies the CopyConstructible MoveConstructible requirements (Table 34 33) and the CopyAssignable MoveAssignable requirements (Table 36 35);
  • T is Swappable if a namespace scope function named swap exists in the same namespace as the definition of T, such that the expression swap(t,u) is valid and has the semantics described in this table.

[ Kona (2007): We like the change to the Swappable requirements to use move semantics. The issue relating to the support of proxies is separable from the one relating to move semantics, and it's bigger than just swap. We'd like to address only the move semantics changes under this issue, and open a separated issue (742) to handle proxies. Also, there may be a third issue, in that the current definition of Swappable does not permit rvalues to be operands to a swap operation, and Howard's proposed resolution would allow the right-most operand to be an rvalue, but it would not allow the left-most operand to be an rvalue (some swap functions in the library have been overloaded to permit left operands to swap to be rvalues). ]


673. unique_ptr update

Section: 20.6.5 [unique.ptr] Status: Open Submitter: Howard Hinnant Date: 2007-05-04

View other active issues in [unique.ptr].

View all other issues in [unique.ptr].

View all issues with Open status.

Discussion:

Since the publication of N1856 there have been a few small but significant advances which should be included into unique_ptr. There exists a reference implmenation for all of these changes.

[ Kona (2007): We don't like the solution given to the first bullet in light of concepts. The second bullet solves the problem of supporting fancy pointers for one library component only. The full LWG needs to decide whether to solve the problem of supporting fancy pointers piecemeal, or whether a paper addressing the whole library is needed. We think that the third bullet is correct. ]

[ Post Kona: Howard adds example user code related to the first bullet: ]

void legacy_code(void*, std::size_t);

void foo(std::size_t N)
{
    std::unique_ptr<void, void(*)(void*)> ptr(std::malloc(N), std::free);
    legacy_code(ptr.get(), N);
}   // unique_ptr used for exception safety purposes

I.e. unique_ptr<void> is a useful tool that we don't want to disable with concepts. The only part of unique_ptr<void> we want to disable (with concepts or by other means) are the two member functions:

T& operator*() const;
T* operator->() const;

Proposed resolution:

[ I am grateful for the generous aid of Peter Dimov and Ion Gaztañaga in helping formulate and review the proposed resolutions below. ]


674. shared_ptr interface changes for consistency with N1856

Section: 20.6.6.2 [util.smartptr.shared] Status: Ready Submitter: Peter Dimov Date: 2007-05-05

View other active issues in [util.smartptr.shared].

View all other issues in [util.smartptr.shared].

View all issues with Ready status.

Discussion:

N1856 does not propose any changes to shared_ptr. It needs to be updated to use a rvalue reference where appropriate and to interoperate with unique_ptr as it does with auto_ptr.

Proposed resolution:

Change 20.6.6.2 [util.smartptr.shared] as follows:

template<class Y> explicit shared_ptr(auto_ptr<Y>&&& r);
template<class Y, class D> explicit shared_ptr(const unique_ptr<Y,D>& r) = delete;
template<class Y, class D> explicit shared_ptr(unique_ptr<Y,D>&& r);
...
template<class Y> shared_ptr& operator=(auto_ptr<Y>&&& r);
template<class Y, class D> shared_ptr& operator=(const unique_ptr<Y,D>& r) = delete;
template<class Y, class D> shared_ptr& operator=(unique_ptr<Y,D>&& r);

Change 20.6.6.2.1 [util.smartptr.shared.const] as follows:

template<class Y> shared_ptr(auto_ptr<Y>&&& r);

Add to 20.6.6.2.1 [util.smartptr.shared.const]:

template<class Y, class D> shared_ptr(unique_ptr<Y, D>&& r);

Effects: Equivalent to shared_ptr( r.release(), r.get_deleter() ) when D is not a reference type, shared_ptr( r.release(), ref( r.get_deleter() ) ) otherwise.

Exception safety: If an exception is thrown, the constructor has no effect.

Change 20.6.6.2.3 [util.smartptr.shared.assign] as follows:

template<class Y> shared_ptr& operator=(auto_ptr<Y>&&& r);

Add to 20.6.6.2.3 [util.smartptr.shared.assign]:

template<class Y, class D> shared_ptr& operator=(unique_ptr<Y,D>&& r);

-4- Effects: Equivalent to shared_ptr(std::move(r)).swap(*this).

-5- Returns: *this.

[ Kona (2007): We may need to open an issue (743) to deal with the question of whether shared_ptr needs an rvalue swap. ]


675. Move assignment of containers

Section: 23.1 [container.requirements] Status: Ready Submitter: Howard Hinnant Date: 2007-05-05

View other active issues in [container.requirements].

View all other issues in [container.requirements].

View all issues with Ready status.

Discussion:

James Hopkin pointed out to me that if vector<T> move assignment is O(1) (just a swap) then containers such as vector<shared_ptr<ostream>> might have the wrong semantics under move assignment when the source is not truly an rvalue, but a moved-from lvalue (destructors could run late).

vector<shared_ptr<ostream>> v1;
vector<shared_ptr<ostream>> v2;
...
v1 = v2;               // #1
v1 = std::move(v2);    // #2

Move semantics means not caring what happens to the source (v2 in this example). It doesn't mean not caring what happens to the target (v1). In the above example both assignments should have the same effect on v1. Any non-shared ostream's v1 owns before the assignment should be closed, whether v1 is undergoing copy assignment or move assignment.

This implies that the semantics of move assignment of a generic container should be clear, swap instead of just swap. An alternative which could achieve the same effect would be to move assign each element. In either case, the complexity of move assignment needs to be relaxed to O(v1.size()).

The performance hit of this change is not nearly as drastic as it sounds. In practice, the target of a move assignment has always just been move constructed or move assigned from. Therefore under clear, swap semantics (in this common use case) we are still achieving O(1) complexity.

Proposed resolution:

Change 23.1 [container.requirements]:

Table 86: Container requirements
expressionreturn typeoperational semantics assertion/note pre/post-conditioncomplexity
a = rv;X& All existing elements of a are either move assigned or destructed a shall be equal to the value that rv had before this construction constant linear in a.size()

676. Moving the unordered containers

Section: 23.4 [unord] Status: Ready Submitter: Howard Hinnant Date: 2007-05-05

View other active issues in [unord].

View all other issues in [unord].

View all issues with Ready status.

Discussion:

Move semantics are missing from the unordered containers. The proposed resolution below adds move-support consistent with N1858 and the current working draft.

The current proposed resolution simply lists the requirements for each function. These might better be hoisted into the requirements table for unordered associative containers. Futhermore a mild reorganization of the container requirements could well be in order. This defect report is purposefully ignoring these larger issues and just focusing on getting the unordered containers "moved".

Proposed resolution:

Add to 23.4 [unord]:

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_map<Key, T, Hash, Pred, Alloc>& x, 
            unordered_map<Key, T, Hash, Pred, Alloc>& y); 

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_map<Key, T, Hash, Pred, Alloc>& x, 
            unordered_map<Key, T, Hash, Pred, Alloc>&& y);

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_map<Key, T, Hash, Pred, Alloc>&& x, 
            unordered_map<Key, T, Hash, Pred, Alloc>& y);

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multimap<Key, T, Hash, Pred, Alloc>& x, 
            unordered_multimap<Key, T, Hash, Pred, Alloc>& y);

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multimap<Key, T, Hash, Pred, Alloc>& x, 
            unordered_multimap<Key, T, Hash, Pred, Alloc>&& y);

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multimap<Key, T, Hash, Pred, Alloc>&& x, 
            unordered_multimap<Key, T, Hash, Pred, Alloc>& y);

...

template <class Value, class Hash, class Pred, class Alloc> 
  void swap(unordered_set<Value, Hash, Pred, Alloc>& x, 
            unordered_set<Value, Hash, Pred, Alloc>& y); 

template <class Value, class Hash, class Pred, class Alloc> 
  void swap(unordered_set<Value, Hash, Pred, Alloc>& x, 
            unordered_set<Value, Hash, Pred, Alloc>&& y);

template <class Value, class Hash, class Pred, class Alloc> 
  void swap(unordered_set<Value, Hash, Pred, Alloc>&& x, 
            unordered_set<Value, Hash, Pred, Alloc>& y);

template <class Value, class Hash, class Pred, class Alloc> 
  void swap(unordered_multiset<Value, Hash, Pred, Alloc>& x, 
            unordered_multiset<Value, Hash, Pred, Alloc>& y);

template <class Value, class Hash, class Pred, class Alloc> 
  void swap(unordered_multiset<Value, Hash, Pred, Alloc>& x, 
            unordered_multiset<Value, Hash, Pred, Alloc>&& y);

template <class Value, class Hash, class Pred, class Alloc> 
  void swap(unordered_multiset<Value, Hash, Pred, Alloc>&& x, 
            unordered_multiset<Value, Hash, Pred, Alloc>& y);

unordered_map

Change 23.4.1 [unord.map]:

class unordered_map
{
    ...
    unordered_map(const unordered_map&);
    unordered_map(unordered_map&&);
    ~unordered_map();
    unordered_map& operator=(const unordered_map&);
    unordered_map& operator=(unordered_map&&);
    ...
    // modifiers 
    std::pair<iterator, bool> insert(const value_type& obj); 
    template <class P> pair<iterator, bool> insert(P&& obj);
    iterator       insert(iterator hint, const value_type& obj);
    template <class P> iterator       insert(iterator hint, P&& obj);
    const_iterator insert(const_iterator hint, const value_type& obj);
    template <class P> const_iterator insert(const_iterator hint, P&& obj);
    ...
    void swap(unordered_map&&);
    ...
    mapped_type& operator[](const key_type& k);
    mapped_type& operator[](key_type&& k);
    ...
};

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_map<Key, T, Hash, Pred, Alloc>& x, 
            unordered_map<Key, T, Hash, Pred, Alloc>& y);

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_map<Key, T, Hash, Pred, Alloc>& x, 
            unordered_map<Key, T, Hash, Pred, Alloc>&& y);

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_map<Key, T, Hash, Pred, Alloc>&& x, 
            unordered_map<Key, T, Hash, Pred, Alloc>& y);

Add to 23.4.1.1 [unord.map.cnstr]:

template <class InputIterator>
  unordered_map(InputIterator f, InputIterator l, 
                size_type n = implementation-defined, 
                const hasher& hf = hasher(), 
                const key_equal& eql = key_equal(), 
                const allocator_type& a = allocator_type());

Requires: If the iterator's dereference operator returns an lvalue or a const rvalue pair<key_type, mapped_type>, then both key_type and mapped_type shall be CopyConstructible.

Add to 23.4.1.2 [unord.map.elem]:

mapped_type& operator[](const key_type& k);

...

Requires: key_type shall be CopyConstructible and mapped_type shall be DefaultConstructible.

mapped_type& operator[](key_type&& k);

Effects: If the unordered_map does not already contain an element whose key is equivalent to k , inserts the value std::pair<const key_type, mapped_type>(std::move(k), mapped_type()).

Requires: mapped_type shall be DefaultConstructible.

Returns: A reference to x.second, where x is the (unique) element whose key is equivalent to k.

Add new section [unord.map.modifiers]:

pair<iterator, bool> insert(const value_type& x);
template <class P> pair<iterator, bool> insert(P&& x);
iterator       insert(iterator hint, const value_type& x);
template <class P> iterator       insert(iterator hint, P&& x);
const_iterator insert(const_iterator hint, const value_type& x);
template <class P> const_iterator insert(const_iterator hint, P&& x);
template <class InputIterator>
  void insert(InputIterator first, InputIterator last);

Requires: Those signatures taking a const value_type& parameter requires both the key_type and the mapped_type to be CopyConstructible.

P shall be convertible to value_type. If P is instantiated as a reference type, then the argument x is copied from. Otherwise x is considered to be an rvalue as it is converted to value_type and inserted into the unordered_map. Specifically, in such cases CopyConstructible is not required of key_type or mapped_type unless the conversion from P specifically requires it (e.g. if P is a tuple<const key_type, mapped_type>, then key_type must be CopyConstructible).

The signature taking InputIterator parameters requires CopyConstructible of both key_type and mapped_type if the dereferenced InputIterator returns an lvalue or const rvalue value_type.

Add to 23.4.1.3 [unord.map.swap]:

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_map<Key, T, Hash, Pred, Alloc>& x, 
            unordered_map<Key, T, Hash, Pred, Alloc>& y);
template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_map<Key, T, Hash, Pred, Alloc>& x, 
            unordered_map<Key, T, Hash, Pred, Alloc>&& y);
template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_map<Key, T, Hash, Pred, Alloc>&& x, 
            unordered_map<Key, T, Hash, Pred, Alloc>& y);

unordered_multimap

Change 23.4.2 [unord.multimap]:

class unordered_multimap
{
    ...
    unordered_multimap(const unordered_multimap&);
    unordered_multimap(unordered_multimap&&);
    ~unordered_multimap();
    unordered_multimap& operator=(const unordered_multimap&);
    unordered_multimap& operator=(unordered_multimap&&);
    ...
    // modifiers 
    iterator insert(const value_type& obj); 
    template <class P> iterator insert(P&& obj);
    iterator       insert(iterator hint, const value_type& obj);
    template <class P> iterator       insert(iterator hint, P&& obj);
    const_iterator insert(const_iterator hint, const value_type& obj);
    template <class P> const_iterator insert(const_iterator hint, P&& obj);
    ...
    void swap(unordered_multimap&&);
    ...
};

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multimap<Key, T, Hash, Pred, Alloc>& x, 
            unordered_multimap<Key, T, Hash, Pred, Alloc>& y);

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multimap<Key, T, Hash, Pred, Alloc>& x, 
            unordered_multimap<Key, T, Hash, Pred, Alloc>&& y);

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multimap<Key, T, Hash, Pred, Alloc>&& x, 
            unordered_multimap<Key, T, Hash, Pred, Alloc>& y);

Add to 23.4.2.1 [unord.multimap.cnstr]:

template <class InputIterator>
  unordered_multimap(InputIterator f, InputIterator l, 
                size_type n = implementation-defined, 
                const hasher& hf = hasher(), 
                const key_equal& eql = key_equal(), 
                const allocator_type& a = allocator_type());

Requires: If the iterator's dereference operator returns an lvalue or a const rvalue pair<key_type, mapped_type>, then both key_type and mapped_type shall be CopyConstructible.

Add new section [unord.multimap.modifiers]:

iterator insert(const value_type& x);
template <class P> iterator       insert(P&& x);
iterator       insert(iterator hint, const value_type& x);
template <class P> iterator       insert(iterator hint, P&& x);
const_iterator insert(const_iterator hint, const value_type& x);
template <class P> const_iterator insert(const_iterator hint, P&& x);
template <class InputIterator>
  void insert(InputIterator first, InputIterator last);

Requires: Those signatures taking a const value_type& parameter requires both the key_type and the mapped_type to be CopyConstructible.

P shall be convertible to value_type. If P is instantiated as a reference type, then the argument x is copied from. Otherwise x is considered to be an rvalue as it is converted to value_type and inserted into the unordered_multimap. Specifically, in such cases CopyConstructible is not required of key_type or mapped_type unless the conversion from P specifically requires it (e.g. if P is a tuple<const key_type, mapped_type>, then key_type must be CopyConstructible).

The signature taking InputIterator parameters requires CopyConstructible of both key_type and mapped_type if the dereferenced InputIterator returns an lvalue or const rvalue value_type.

Add to 23.4.2.2 [unord.multimap.swap]:

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multimap<Key, T, Hash, Pred, Alloc>& x, 
            unordered_multimap<Key, T, Hash, Pred, Alloc>& y);
template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multimap<Key, T, Hash, Pred, Alloc>& x, 
            unordered_multimap<Key, T, Hash, Pred, Alloc>&& y);
template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multimap<Key, T, Hash, Pred, Alloc>&& x, 
            unordered_multimap<Key, T, Hash, Pred, Alloc>& y);

unordered_set

Change 23.4.3 [unord.set]:

class unordered_set
{
    ...
    unordered_set(const unordered_set&);
    unordered_set(unordered_set&&);
    ~unordered_set();
    unordered_set& operator=(const unordered_set&);
    unordered_set& operator=(unordered_set&&);
    ...
    // modifiers 
    std::pair<iterator, bool> insert(const value_type& obj); 
    pair<iterator, bool> insert(value_type&& obj);
    iterator       insert(iterator hint, const value_type& obj);
    iterator       insert(iterator hint, value_type&& obj);
    const_iterator insert(const_iterator hint, const value_type& obj);
    const_iterator insert(const_iterator hint, value_type&& obj);
    ...
    void swap(unordered_set&&);
    ...
};

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_set<Key, T, Hash, Pred, Alloc>& x, 
            unordered_set<Key, T, Hash, Pred, Alloc>& y);

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_set<Key, T, Hash, Pred, Alloc>& x, 
            unordered_set<Key, T, Hash, Pred, Alloc>&& y);

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_set<Key, T, Hash, Pred, Alloc>&& x, 
            unordered_set<Key, T, Hash, Pred, Alloc>& y);

Add to 23.4.3.1 [unord.set.cnstr]:

template <class InputIterator>
  unordered_set(InputIterator f, InputIterator l, 
                size_type n = implementation-defined, 
                const hasher& hf = hasher(), 
                const key_equal& eql = key_equal(), 
                const allocator_type& a = allocator_type());

Requires: If the iterator's dereference operator returns an lvalue or a const rvalue value_type, then the value_type shall be CopyConstructible.

Add new section [unord.set.modifiers]:

pair<iterator, bool> insert(const value_type& x);
pair<iterator, bool> insert(value_type&& x);
iterator       insert(iterator hint, const value_type& x);
iterator       insert(iterator hint, value_type&& x);
const_iterator insert(const_iterator hint, const value_type& x);
const_iterator insert(const_iterator hint, value_type&& x);
template <class InputIterator>
  void insert(InputIterator first, InputIterator last);

Requires: Those signatures taking a const value_type& parameter requires the value_type to be CopyConstructible.

The signature taking InputIterator parameters requires CopyConstructible of value_type if the dereferenced InputIterator returns an lvalue or const rvalue value_type.

Add to 23.4.3.2 [unord.set.swap]:

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_set<Key, T, Hash, Pred, Alloc>& x, 
            unordered_set<Key, T, Hash, Pred, Alloc>& y);
template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_set<Key, T, Hash, Pred, Alloc>& x, 
            unordered_set<Key, T, Hash, Pred, Alloc>&& y);
template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_set<Key, T, Hash, Pred, Alloc>&& x, 
            unordered_set<Key, T, Hash, Pred, Alloc>& y);

unordered_multiset

Change 23.4.4 [unord.multiset]:

class unordered_multiset
{
    ...
    unordered_multiset(const unordered_multiset&);
    unordered_multiset(unordered_multiset&&);
    ~unordered_multiset();
    unordered_multiset& operator=(const unordered_multiset&);
    unordered_multiset& operator=(unordered_multiset&&);
    ...
    // modifiers 
    iterator insert(const value_type& obj); 
    iterator insert(value_type&& obj);
    iterator       insert(iterator hint, const value_type& obj);
    iterator       insert(iterator hint, value_type&& obj);
    const_iterator insert(const_iterator hint, const value_type& obj);
    const_iterator insert(const_iterator hint, value_type&& obj);
    ...
    void swap(unordered_multiset&&);
    ...
};

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multiset<Key, T, Hash, Pred, Alloc>& x, 
            unordered_multiset<Key, T, Hash, Pred, Alloc>& y);

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multiset<Key, T, Hash, Pred, Alloc>& x, 
            unordered_multiset<Key, T, Hash, Pred, Alloc>&& y);

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multiset<Key, T, Hash, Pred, Alloc>&& x, 
            unordered_multiset<Key, T, Hash, Pred, Alloc>& y);

Add to 23.4.4.1 [unord.multiset.cnstr]:

template <class InputIterator>
  unordered_multiset(InputIterator f, InputIterator l, 
                size_type n = implementation-defined, 
                const hasher& hf = hasher(), 
                const key_equal& eql = key_equal(), 
                const allocator_type& a = allocator_type());

Requires: If the iterator's dereference operator returns an lvalue or a const rvalue value_type, then the value_type shall be CopyConstructible.

Add new section [unord.multiset.modifiers]:

iterator insert(const value_type& x);
iterator insert(value_type&& x);
iterator       insert(iterator hint, const value_type& x);
iterator       insert(iterator hint, value_type&& x);
const_iterator insert(const_iterator hint, const value_type& x);
const_iterator insert(const_iterator hint, value_type&& x);
template <class InputIterator>
  void insert(InputIterator first, InputIterator last);

Requires: Those signatures taking a const value_type& parameter requires the value_type to be CopyConstructible.

The signature taking InputIterator parameters requires CopyConstructible of value_type if the dereferenced InputIterator returns an lvalue or const rvalue value_type.

Add to 23.4.4.2 [unord.multiset.swap]:

template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multiset<Key, T, Hash, Pred, Alloc>& x, 
            unordered_multiset<Key, T, Hash, Pred, Alloc>& y);
template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multiset<Key, T, Hash, Pred, Alloc>& x, 
            unordered_multiset<Key, T, Hash, Pred, Alloc>&& y);
template <class Key, class T, class Hash, class Pred, class Alloc> 
  void swap(unordered_multiset<Key, T, Hash, Pred, Alloc>&& x, 
            unordered_multiset<Key, T, Hash, Pred, Alloc>& y);

679. resize parameter by value

Section: 23.2 [sequences] Status: Ready Submitter: Howard Hinnant Date: 2007-06-11

View all issues with Ready status.

Discussion:

The C++98 standard specifies that one member function alone of the containers passes its parameter (T) by value instead of by const reference:

void resize(size_type sz, T c = T());

This fact has been discussed / debated repeatedly over the years, the first time being even before C++98 was ratified. The rationale for passing this parameter by value has been:

So that self referencing statements are guaranteed to work, for example:

v.resize(v.size() + 1, v[0]);

However this rationale is not convincing as the signature for push_back is:

void push_back(const T& x);

And push_back has similar semantics to resize (append). And push_back must also work in the self referencing case:

v.push_back(v[0]);  // must work

The problem with passing T by value is that it can be significantly more expensive than passing by reference. The converse is also true, however when it is true it is usually far less dramatic (e.g. for scalar types).

Even with move semantics available, passing this parameter by value can be expensive. Consider for example vector<vector<int>>:

std::vector<int> x(1000);
std::vector<std::vector<int>> v;
...
v.resize(v.size()+1, x);

In the pass-by-value case, x is copied once to the parameter of resize. And then internally, since the code can not know at compile time by how much resize is growing the vector, x is usually copied (not moved) a second time from resize's parameter into its proper place within the vector.

With pass-by-const-reference, the x in the above example need be copied only once. In this case, x has an expensive copy constructor and so any copies that can be saved represents a significant savings.

If we can be efficient for push_back, we should be efficient for resize as well. The resize taking a reference parameter has been coded and shipped in the CodeWarrior library with no reports of problems which I am aware of.

Proposed resolution:

Change 23.2.2 [deque], p2:

class deque {
   ...
   void resize(size_type sz, const T& c);

Change 23.2.2.2 [deque.capacity], p3:

void resize(size_type sz, const T& c);

Change 23.2.3 [list], p2:

class list {
   ...
   void resize(size_type sz, const T& c);

Change 23.2.3.2 [list.capacity], p3:

void resize(size_type sz, const T& c);

Change 23.2.5 [vector], p2:

class vector {
   ...
   void resize(size_type sz, const T& c);

Change 23.2.5.2 [vector.capacity], p11:

void resize(size_type sz, const T& c);

680. move_iterator operator-> return

Section: 24.4.3.1 [move.iterator] Status: Ready Submitter: Howard Hinnant Date: 2007-06-11

View all issues with Ready status.

Discussion:

move_iterator's operator-> return type pointer does not consistently match the type which is returned in the description in 24.4.3.3.5 [move.iter.op.ref].

template <class Iterator>
class move_iterator {
public:
    ...
    typedef typename iterator_traits<Iterator>::pointer pointer;
    ...
    pointer operator->() const {return current;}
    ...
private: 
    Iterator current; // exposition only
};

There are two possible fixes.

  1. pointer operator->() const {return &*current;}
  2. typedef Iterator pointer;

The first solution is the one chosen by reverse_iterator. A potential disadvantage of this is it may not work well with iterators which return a proxy on dereference and that proxy has overloaded operator&(). Proxy references often need to overloaad operator&() to return a proxy pointer. That proxy pointer may or may not be the same type as the iterator's pointer type.

By simply returning the Iterator and taking advantage of the fact that the language forwards calls to operator-> automatically until it finds a non-class type, the second solution avoids the issue of an overloaded operator&() entirely.

Proposed resolution:

Change the synopsis in 24.4.3.1 [move.iterator]:

typedef typename iterator_traits<Iterator>::pointer pointer;

684. Unclear which members of match_results should be used in comparison

Section: 28.10 [re.results] Status: Review Submitter: Nozomu Katoo Date: 2007-05-27

View other active issues in [re.results].

View all other issues in [re.results].

View all issues with Review status.

Discussion:

In 28.4 [re.syn] of N2284, two template functions are declared here:

// 28.10, class template match_results: 
  <snip>
// match_results comparisons 
  template <class BidirectionalIterator, class Allocator> 
    bool operator== (const match_results<BidirectionalIterator, Allocator>& m1, 
                     const match_results<BidirectionalIterator, Allocator>& m2); 
  template <class BidirectionalIterator, class Allocator> 
    bool operator!= (const match_results<BidirectionalIterator, Allocator>& m1, 
                     const match_results<BidirectionalIterator, Allocator>& m2); 

// 28.10.6, match_results swap:

But the details of these two bool operator functions (i.e., which members of match_results should be used in comparison) are not described in any following sections.

[ John adds: ]

That looks like a bug: operator== should return true only if the two objects refer to the same match - ie if one object was constructed as a copy of the other.

[ Kona (2007): Bill and Pete to add minor wording to that proposed in N2409. ]

Proposed resolution:

Add a new section after 28.10.6 [re.results.swap], which reads:

28.10.7 match_results non-member functions.

template<class BidirectionalIterator, class Allocator> 
  bool operator==(const match_results<BidirectionalIterator, Allocator>& m1, 
                  const match_results<BidirectionalIterator, Allocator>& m2);

Returns: true only if the two objects refer to the same match.

template<class BidirectionalIterator, class Allocator> 
  bool operator!=(const match_results<BidirectionalIterator, Allocator>& m1, 
                  const match_results<BidirectionalIterator, Allocator>& m2);

Returns: !(m1 == m2).

template<class BidirectionalIterator, class Allocator> 
  void swap(match_results<BidirectionalIterator, Allocator>& m1, 
            match_results<BidirectionalIterator, Allocator>& m2);

Returns: m1.swap(m2).


685. reverse_iterator/move_iterator difference has invalid signatures

Section: 24.4.1.3.19 [reverse.iter.opdiff], 24.4.3.3.14 [move.iter.nonmember] Status: Review Submitter: Bo Persson Date: 2007-06-10

View all issues with Review status.

Discussion:

In C++03 the difference between two reverse_iterators

ri1 - ri2

is possible to compute only if both iterators have the same base iterator. The result type is the difference_type of the base iterator.

In the current draft, the operator is defined as 24.4.1.3.19 [reverse.iter.opdiff]

template<class Iterator1, class Iterator2> 
typename reverse_iterator<Iterator>::difference_type 
   operator-(const reverse_iterator<Iterator1>& x, 
                    const reverse_iterator<Iterator2>& y);

The return type is the same as the C++03 one, based on the no longer present Iterator template parameter.

Besides being slightly invalid, should this operator work only when Iterator1 and Iterator2 has the same difference_type? Or should the implementation choose one of them? Which one?

The same problem now also appears in operator-() for move_iterator 24.4.3.3.14 [move.iter.nonmember].

Proposed resolution:

Change the synopsis in 24.4.1.1 [reverse.iterator]:

template <class Iterator1, class Iterator2> 
  typename reverse_iterator<Iterator>::difference_type auto operator-( 
    const reverse_iterator<Iterator1>& x, 
    const reverse_iterator<Iterator2>& y) -> decltype(operator-(Iterator1, Iterator2));

Returns: y.current - x.current.

Change 24.4.1.3.19 [reverse.iter.opdiff]:

template <class Iterator1, class Iterator2> 
  typename reverse_iterator<Iterator>::difference_type auto operator-( 
    const reverse_iterator<Iterator1>& x, 
    const reverse_iterator<Iterator2>& y) -> decltype(operator-(Iterator1, Iterator2));

Returns: y.current - x.current.

Change the synopsis in 24.4.3.1 [move.iterator]:

template <class Iterator1, class Iterator2> 
  typename move_iterator<Iterator>::difference_type auto operator-( 
    const move_iterator<Iterator1>& x, 
    const move_iterator<Iterator2>& y) -> decltype(operator-(Iterator1, Iterator2));

Returns: y.current - x.current.

Change 24.4.3.3.14 [move.iter.nonmember]:

template <class Iterator1, class Iterator2> 
  typename move_iterator<Iterator>::difference_type auto operator-( 
    const move_iterator<Iterator1>& x, 
    const move_iterator<Iterator2>& y) -> decltype(operator-(Iterator1, Iterator2));

Returns: x.base() - y.base().


686. Unique_ptr and shared_ptr fail to specify non-convertibility to int for unspecified-bool-type

Section: 20.6.5.2.4 [unique.ptr.single.observers], 20.6.6.2.5 [util.smartptr.shared.obs] Status: Open Submitter: Beman Dawes Date: 2007-06-14

View all issues with Open status.

Discussion:

The standard library uses the operator unspecified-bool-type() const idiom in five places. In three of those places (20.5.15.2.3 [func.wrap.func.cap], function capacity for example) the returned value is constrained to disallow unintended conversions to int. The standardese is

The return type shall not be convertible to int.

This constraint is omitted for unique_ptr and shared_ptr. It should be added for those.

Proposed resolution:

To the Returns paragraph for operator unspecified-bool-type() const of 20.6.5.2.4 [unique.ptr.single.observers] paragraph 11 and 20.6.6.2.5 [util.smartptr.shared.obs] paragraph 16, add the sentence:

The return type shall not be convertible to int.

[ Kona (2007): Uncertain if nullptr will address this issue. ]


687. shared_ptr conversion constructor not constrained

Section: 20.6.6.2.1 [util.smartptr.shared.const], 20.6.6.3.1 [util.smartptr.weak.const] Status: Ready Submitter: Peter Dimov Date: 2007-05-10

View all issues with Ready status.

Discussion:

Since all conversions from shared_ptr<T> to shared_ptr<U> have the same rank regardless of the relationship between T and U, reasonable user code that works with raw pointers fails with shared_ptr:

void f( shared_ptr<void> );
void f( shared_ptr<int> );

int main()
{
  f( shared_ptr<double>() ); // ambiguous
}

Now that we officially have enable_if, we can constrain the constructor and the corresponding assignment operator to only participate in the overload resolution when the pointer types are compatible.

Proposed resolution:

In 20.6.6.2.1 [util.smartptr.shared.const], change:

-14- Requires: For the second constructor The second constructor shall not participate in the overload resolution unless Y* shall be is implicitly convertible to T*.

In 20.6.6.3.1 [util.smartptr.weak.const], change:

template<class Y> weak_ptr(shared_ptr<Y> const& r);
weak_ptr(weak_ptr const& r);
template<class Y> weak_ptr(weak_ptr<Y> const& r);
weak_ptr(weak_ptr const& r);
template<class Y> weak_ptr(weak_ptr<Y> const& r);
template<class Y> weak_ptr(shared_ptr<Y> const& r);

-4- Requires: For tThe second and third constructors, shall not participate in the overload resolution unless Y* shall be is implicitly convertible to T*.


688. reference_wrapper, cref unsafe, allow binding to rvalues

Section: 20.5.5.1 [refwrap.const] Status: Ready Submitter: Peter Dimov Date: 2007-05-10

View other active issues in [refwrap.const].

View all other issues in [refwrap.const].

View all issues with Ready status.

Discussion:

A reference_wrapper can be constructed from an rvalue, either by using the constructor, or via cref (and ref in some corner cases). This leads to a dangling reference being stored into the reference_wrapper object. Now that we have a mechanism to detect an rvalue, we can fix them to disallow this source of undefined behavior.

Also please see the thread starting at c++std-lib-17398 for some good discussion on this subject.

Proposed resolution:

In 20.5.5 [refwrap], add:

private:
  explicit reference_wrapper(T&&);

In 20.5.5.1 [refwrap.const], add:

explicit reference_wrapper(T&&);

-?- Not defined to disallow creating a reference_wrapper from an rvalue.

In the synopsis of <functional> (20.5.5 [refwrap]), change the declarations of ref and cref to:

template<class T> reference_wrapper<T> ref(T&&);
template<class T> reference_wrapper<const T> cref(const T&&);

In 20.5.5.5 [refwrap.helpers], change:

template<class T> reference_wrapper<T> ref(T&& t);

-1- Requires: t shall be an lvalue.

and change:

template<class T> reference_wrapper<const T> cref(const T&& t);

-6- Requires: t shall be an lvalue.

[ N2292 addresses the first part of the resolution but not the second. ]


689. reference_wrapper constructor overly constrained

Section: 20.5.5.1 [refwrap.const] Status: Ready Submitter: Peter Dimov Date: 2007-05-10

View other active issues in [refwrap.const].

View all other issues in [refwrap.const].

View all issues with Ready status.

Discussion:

The constructor of reference_wrapper is currently explicit. The primary motivation behind this is the safety problem with respect to rvalues, which is addressed by the proposed resolution of the previous issue. Therefore we should consider relaxing the requirements on the constructor since requests for the implicit conversion keep resurfacing.

Also please see the thread starting at c++std-lib-17398 for some good discussion on this subject.

Proposed resolution:

Remove the explicit from the constructor of reference_wrapper. If the proposed resolution of the previous issue is accepted, remove the explicit from the T&& constructor as well to keep them in sync.


691. const_local_iterator cbegin, cend missing from TR1

Section: 23.4 [unord], TR1 6.3 [tr.hash] Status: Review Submitter: Joaquín M López Muñoz Date: 2007-06-14

View other active issues in [unord].

View all other issues in [unord].

View all issues with Review status.

Discussion:

The last version of TR1 does not include the following member functions for unordered containers:

const_local_iterator cbegin(size_type n) const;
const_local_iterator cend(size_type n) const;

which looks like an oversight to me. I've checked th TR1 issues lists and the latest working draft of the C++0x std (N2284) and haven't found any mention to these menfuns or to their absence.

Is this really an oversight, or am I missing something?

Proposed resolution:

Add the following two rows to table 93 (unordered associative container requirements) in section 23.1.3 [unord.req]:

Unordered associative container requirements (in addition to container)
expression return type assertion/note pre/post-condition complexity
b.cbegin(n) const_local_iterator n shall be in the range [0, bucket_count()). Note: [b.cbegin(n), b.cend(n)) is a valid range containing all of the elements in the nth bucket. Constant
b.cend(n) const_local_iterator n shall be in the range [0, bucket_count()). Constant

Add to the synopsis in 23.4.1 [unord.map]:

const_local_iterator cbegin(size_type n) const;
const_local_iterator cend(size_type n) const;

Add to the synopsis in 23.4.2 [unord.multimap]:

const_local_iterator cbegin(size_type n) const;
const_local_iterator cend(size_type n) const;

Add to the synopsis in 23.4.3 [unord.set]:

const_local_iterator cbegin(size_type n) const;
const_local_iterator cend(size_type n) const;

Add to the synopsis in 23.4.4 [unord.multiset]:

const_local_iterator cbegin(size_type n) const;
const_local_iterator cend(size_type n) const;

692. get_money and put_money should be formatted I/O functions

Section: 27.6.4 [ext.manip] Status: New Submitter: Martin Sebor Date: 2007-06-22

View all other issues in [ext.manip].

View all issues with New status.

Discussion:

In a private email Bill Plauger notes:

I  believe that  the function  that  implements get_money [from N2072] should behave  as a  formatted input function,  and the  function that implements put_money should  behave as a formatted output function. This  has implications regarding the  skipping of whitespace and the handling of errors, among other things.

The words  don't say that  right now and  I'm far from  convinced that such a change is editorial.

Martin's response:

I agree that the manipulators should handle exceptions the same way as formatted I/O functions do. The text in N2072 assumes so but the Returns clause explicitly omits exception handling for the sake of brevity. The spec should be clarified to that effect.

As for dealing  with whitespace, I also agree it  would make sense for the extractors  and inserters involving the new  manipulators to treat it the same way as formatted I/O.

Proposed resolution:

Add  a new  paragraph immediately  above  p4 of 27.6.4 [ext.manip] with  the following text:

Effects:  The   expression  in >> get_money(mon, intl) described below behaves as a formatted input function (as described in 27.6.1.2.1 [istream.formatted.reqmts]).

Also change p4 of 27.6.4 [ext.manip] as follows:

Returns: An object s of unspecified type such that if in is  an object of type basic_istream<charT, traits>    then    the    expression   in >> get_money(mon, intl) behaves as a formatted input function that    calls    f(in, mon, intl)    were called. The function f can be defined as...


693. std::bitset::all() missing

Section: 23.3.5 [template.bitset] Status: Ready Submitter: Martin Sebor Date: 2007-06-22

View other active issues in [template.bitset].

View all other issues in [template.bitset].

View all issues with Ready status.

Discussion:

The bitset class template provides the member function any() to determine whether an object of the type has any bits set, and the member function none() to determine whether all of an object's bits are clear. However, the template does not provide a corresponding function to discover whether a bitset object has all its bits set. While it is possible, even easy, to obtain this information by comparing the result of count() with the result of size() for equality (i.e., via b.count() == b.size()) the operation is less efficient than a member function designed specifically for that purpose could be. (count() must count all non-zero bits in a bitset a word at a time while all() could stop counting as soon as it encountered the first word with a zero bit).

Proposed resolution:

Add a declaration of the new member function all() to the defintion of the bitset template in 23.3.5 [template.bitset], p1, right above the declaration of any() as shown below:

bool operator!=(const bitset<N>& rhs) const;
bool test(size_t pos) const;
bool all() const;
bool any() const;
bool none() const;

Add a description of the new member function to the end of 23.3.5.2 [bitset.members] with the following text:

bool all() const;

Returns: count() == size().

In addition, change the description of any() and none() for consistency with all() as follows:

bool any() const;

Returns: true if any bit in *this is onecount() != 0.

bool none() const;

Returns: true if no bit in *this is onecount() == 0.


694. std::bitset and long long

Section: 23.3.5 [template.bitset] Status: Ready Submitter: Martin Sebor Date: 2007-06-22

View other active issues in [template.bitset].

View all other issues in [template.bitset].

View all issues with Ready status.

Discussion:

Objects of the bitset class template specializations can be constructed from and explicitly converted to values of the widest C++ integer type, unsigned long. With the introduction of long long into the language the template should be enhanced to make it possible to interoperate with values of this type as well, or perhaps uintmax_t. See c++std-lib-18274 for a brief discussion in support of this change.

Proposed resolution:

For simplicity, instead of adding overloads for unsigned long long and dealing with possible ambiguities in the spec, replace the bitset ctor that takes an unsigned long argument with one taking unsigned long long in the definition of the template as shown below. (The standard permits implementations to add overloads on other integer types or employ template tricks to achieve the same effect provided they don't cause ambiguities or changes in behavior.)

// [bitset.cons] constructors:
bitset();
bitset(unsigned long long val);
template<class charT, class traits, class Allocator>
explicit bitset(
                const basic_string<charT,traits,Allocator>& str,
                typename basic_string<charT,traits,Allocator>::size_type pos = 0,
                typename basic_string<charT,traits,Allocator>::size_type n =
                    basic_string<charT,traits,Allocator>::npos);

Make a corresponding change in 23.3.5.1 [bitset.cons], p2:

bitset(unsigned long long val);

Effects: Constructs an object of class bitset<N>, initializing the first M bit positions to the corresponding bit values in val. M is the smaller of N and the number of bits in the value representation (section [basic.types]) of unsigned long long. If M < N is true, the remaining bit positions are initialized to zero.

Additionally, introduce a new member function to_ullong() to make it possible to convert bitset to values of the new type. Add the following declaration to the definition of the template, immediate after the declaration of to_ulong() in 23.3.5 [template.bitset], p1, as shown below:

// element access:
bool operator[](size_t pos) const; // for b[i];
reference operator[](size_t pos); // for b[i];
unsigned long to_ulong() const;
unsigned long long to_ullong() const;
template <class charT, class traits, class Allocator>
basic_string<charT, traits, Allocator> to_string() const;

And add a description of the new member function to 23.3.5.2 [bitset.members], below the description of the existing to_ulong() (if possible), with the following text:

unsigned long long to_ullong() const;

Throws: overflow_error if the integral value x corresponding to the bits in *this cannot be represented as type unsigned long long.
Returns: x.

695. ctype<char>::classic_table() not accessible

Section: 22.2.1.3 [facet.ctype.special] Status: Ready Submitter: Martin Sebor Date: 2007-06-22

View all issues with Ready status.

Discussion:

The ctype<char>::classic_table() static member function returns a pointer to an array of const ctype_base::mask objects (enums) that contains ctype<char>::table_size elements. The table describes the properties of the character set in the "C" locale (i.e., whether a character at an index given by its value is alpha, digit, punct, etc.), and is typically used to initialize the ctype<char> facet in the classic "C" locale (the protected ctype<char> member function table() then returns the same value as classic_table()).

However, while ctype<char>::table_size (the size of the table) is a public static const member of the ctype<char> specialization, the classic_table() static member function is protected. That makes getting at the classic data less than convenient (i.e., one has to create a whole derived class just to get at the masks array). It makes little sense to expose the size of the table in the public interface while making the table itself protected, especially when the table is a constant object.

The same argument can be made for the non-static protected member function table().

Proposed resolution:

Make the ctype<char>::classic_table() and ctype<char>::table() member functions public by moving their declarations into the public section of the definition of specialization in 22.2.1.3 [facet.ctype.special] as shown below:

  static locale::id id;
  static const size_t table_size = IMPLEMENTATION_DEFINED;
protected:
  const mask* table() const throw();
  static const mask* classic_table() throw();
protected:

~ctype(); // virtual
virtual char do_toupper(char c) const;

696. istream::operator>>(int&) broken

Section: 27.6.1.2.2 [istream.formatted.arithmetic] Status: New Submitter: Martin Sebor Date: 2007-06-23

View other active issues in [istream.formatted.arithmetic].

View all other issues in [istream.formatted.arithmetic].

View all issues with New status.

Discussion:

From message c++std-lib-17897:

The code shown in 27.6.1.2.2 [istream.formatted.arithmetic] as the "as if" implementation of the two arithmetic extractors that don't have a corresponding num_get interface (i.e., the short and int overloads) is subtly buggy in how it deals with EOF, overflow, and other similar conditions (in addition to containing a few typos).

One problem is that if num_get::get() reaches the EOF after reading in an otherwise valid value that exceeds the limits of the narrower type (but not LONG_MIN or LONG_MAX), it will set err to eofbit. Because of the if condition testing for (err == 0), the extractor won't set failbit (and presumably, return a bogus value to the caller).

Another problem with the code is that it never actually sets the argument to the extracted value. It can't happen after the call to setstate() since the function may throw, so we need to show when and how it's done (we can't just punt as say: "it happens afterwards"). However, it turns out that showing how it's done isn't quite so easy since the argument is normally left unchanged by the facet on error except when the error is due to a misplaced thousands separator, which causes failbit to be set but doesn't prevent the facet from storing the value.

Proposed resolution:


697. New <system_error> header leads to name clashes

Section: 19.4 [syserr] Status: New Submitter: Daniel Krügler Date: 2007-06-24

View all issues with New status.

Discussion:

The most recent state of N2241 as well as the current draft N2284 (section 19.4 [syserr], p.2) proposes a new enumeration type posix_errno immediatly in the namespace std. One of the enumerators has the name invalid_argument, or fully qualified: std::invalid_argument. This name clashes with the exception type std::invalid_argument, see 19.1 [std.exceptions]/p.3. This clash makes e.g. the following snippet invalid:

#include <system_error>
#include <stdexcept>

void foo() { throw std::invalid_argument("Don't call us - we call you!"); }

I propose that this enumeration type (and probably the remaining parts of <system_error> as well) should be moved into one additional inner namespace, e.g. sys or system to reduce foreseeable future clashes due to the great number of members that std::posix_errno already contains (Btw.: Why has the already proposed std::sys sub-namespace from N2066 been rejected?). A further clash candidate seems to be std::protocol_error (a reasonable name for an exception related to a std network library, I guess).

Another possible resolution would rely on the proposed strongly typed enums, as described in N2213. But maybe the forbidden implicit conversion to integral types would make these enumerators less attractive in this special case?

Proposed resolution:


698. Some system_error issues

Section: 19.4.5.1 [syserr.syserr.overview] Status: New Submitter: Daniel Krügler Date: 2007-06-24

View all issues with New status.

Discussion:

In 19.4.5.1 [syserr.syserr.overview] we have the class definition of std::system_error. In contrast to all exception classes, which are constructible with a what_arg string (see 19.1 [std.exceptions], or ios_base::failure in 27.4.2.1.1 [ios::failure]), only overloads with with const string& are possible. For consistency with the re-designed remaining exception classes this class should also provide c'tors which accept a const char* what_arg string.

Please note that this proposed addition makes sense even considering the given implementation hint for what(), because what_arg is required to be set as what_arg of the base class runtime_error, which now has the additional c'tor overload accepting a const char*.

Proposed resolution:


700. N1856 defines struct identity

Section: 20.2.2 [forward] Status: Ready Submitter: P.J. Plauger Date: 2007-07-01

View all issues with Ready status.

Discussion:

N1856 defines struct identity in <utility> which clashes with the traditional definition of struct identity in <functional> (not standard, but a common extension from old STL). Be nice if we could avoid this name clash for backward compatibility.

Proposed resolution:

Change 20.2.2 [forward]:

template <class T> struct identity
{
    typedef T type;
    const T& operator()(const T& x) const;
};
const T& operator()(const T& x) const;

Returns: x.


701. assoc laguerre poly's

Section: TR1 5.2.1.1 [tr.num.sf.Lnm] Status: New Submitter: Christopher Crawford Date: 2007-06-30

View all issues with New status.

Discussion:

I see that the definition the associated Laguerre polynomials TR1 5.2.1.1 [tr.num.sf.Lnm] has been corrected since N1687. However, the draft standard only specifies ranks of integer value m, while the associated Laguerre polynomials are actually valid for real values of m > -1.  In the case of non-integer values of m, the definition  Ln(m) = (1/n!)exx-m (d/dx)n (e-xxm+n) must be used, which also holds for integer values of m.  See Abramowitz & Stegun, 22.11.6 for the general case, and 22.5.16-17 for the integer case.  In fact fractional values are most commonly used in physics, for example to m = +/- 1/2 to describe the harmonic oscillator in 1 dimension, and 1/2, 3/2, 5/2, ... in 3 dimensions.

If I am correct, the calculation of the more general case is no more difficult, and is in fact the function implemented in the GNU Scientific Library.  I would urge you to consider upgrading the standard, either adding extra functions for real m or switching the current ones to double.

Proposed resolution:


702. Restriction in associated Legendre functions

Section: TR1 5.2.1.2 [tr.num.sf.Plm] Status: New Submitter: Christopher Crawford Date: 2007-06-30

View all issues with New status.

Discussion:

One other small thing, in TR1 5.2.1.2 [tr.num.sf.Plm], the restriction should  be |x| <= 1, not x >= 0.

Proposed resolution:


703. map::at() need a complexity specification

Section: 23.3.1.2 [map.access] Status: Ready Submitter: Joe Gottman Date: 2007-07-03

View all other issues in [map.access].

View all issues with Ready status.

Discussion:

map::at() need a complexity specification.

Proposed resolution:

Add the following to the specification of map::at(), 23.3.1.2 [map.access]:

Complexity: logarithmic.


704. MoveAssignable requirement for container value type overly strict

Section: 23.1 [container.requirements] Status: Open Submitter: Howard Hinnant Date: 2007-05-20

View other active issues in [container.requirements].

View all other issues in [container.requirements].

View all issues with Open status.

Discussion:

The move-related changes inadvertently overwrote the intent of 276. Issue 276 removed the requirement of CopyAssignable from most of the member functions of node-based containers. But the move-related changes unnecessarily introduced the MoveAssignable requirement for those members which used to require CopyAssignable.

We also discussed (c++std-lib-18722) the possibility of dropping MoveAssignable from some of the sequence requirements. Additionally the in-place construction work may further reduce requirements. For purposes of an easy reference, here are the minimum sequence requirements as I currently understand them. Those items in requirements table in the working draft which do not appear below have been purposefully omitted for brevity as they do not have any requirements of this nature. Some items which do not have any requirements of this nature are included below just to confirm that they were not omitted by mistake.

Container Requirements
X u(a)value_type must be CopyConstructible
X u(rv)array requires value_type to be MoveConstructible
a = uSequences require value_type to be CopyConstructible and CopyAssignable. Associative containers require value_type to be CopyConstructible.
a = rvarray requires value_type to be MoveAssignable. Sequences with non-Swappable allocators require value_type to be MoveConstructible and MoveAssignable. Associative containers with non-Swappable allocators require value_type to be MoveConstructible.
swap(a,u)array requires value_type to be Swappable. Sequences with non-Swappable allocators require value_type to be Swappable, MoveConstructible and MoveAssignable. Associative containers with non-Swappable allocators require value_type to be MoveConstructible.

Sequence Requirements
X(n)value_type must be DefaultConstructible
X(n, t)value_type must be CopyConstructible
X(i, j)If the iterators return an lvalue the value_type must be CopyConstructible. If the iterators return an rvalue the value_type must be MoveConstructible.
a.insert(p, t)The value_type must be CopyConstructible. The sequences vector and deque also require the value_type to be CopyAssignable.
a.insert(p, rv)The value_type must be MoveConstructible. The sequences vector and deque also require the value_type to be MoveAssignable.
a.insert(p, n, t)The value_type must be CopyConstructible. The sequences vector and deque also require the value_type to be CopyAssignable.
a.insert(p, i, j)If the iterators return an lvalue the value_type must be CopyConstructible. The sequences vector and deque also require the value_type to be CopyAssignable when the iterators return an lvalue. If the iterators return an rvalue the value_type must be MoveConstructible. The sequences vector and deque also require the value_type to be MoveAssignable when the iterators return an rvalue.
a.erase(p)The sequences vector and deque require the value_type to be MoveAssignable.
a.erase(q1, q2)The sequences vector and deque require the value_type to be MoveAssignable.
a.clear()
a.assign(i, j)If the iterators return an lvalue the value_type must be CopyConstructible and CopyAssignable. If the iterators return an rvalue the value_type must be MoveConstructible and MoveAssignable.
a.assign(n, t)The value_type must be CopyConstructible and CopyAssignable.
a.resize(n)The value_type must be DefaultConstructible. The sequences vector and deque also require the value_type to be MoveConstructible.
a.resize(n, t)The value_type must be CopyConstructible.

Optional Sequence Requirements
a.front()
a.back()
a.push_front(t)The value_type must be CopyConstructible.
a.push_front(rv)The value_type must be MoveConstructible.
a.push_back(t)The value_type must be CopyConstructible.
a.push_back(rv)The value_type must be MoveConstructible.
a.pop_front()
a.pop_back()
a[n]
a.at[n]

Associative Container Requirements
X(i, j)If the iterators return an lvalue the value_type must be CopyConstructible. If the iterators return an rvalue the value_type must be MoveConstructible.
a_uniq.insert(t)The value_type must be CopyConstructible.
a_uniq.insert(rv)The key_type and the mapped_type (if it exists) must be MoveConstructible.
a_eq.insert(t)The value_type must be CopyConstructible.
a_eq.insert(rv)The key_type and the mapped_type (if it exists) must be MoveConstructible.
a.insert(p, t)The value_type must be CopyConstructible.
a.insert(p, rv)The key_type and the mapped_type (if it exists) must be MoveConstructible.
a.insert(i, j)If the iterators return an lvalue the value_type must be CopyConstructible. If the iterators return an rvalue the key_type and the mapped_type (if it exists) must be MoveConstructible..

Unordered Associative Container Requirements
X(i, j, n, hf, eq)If the iterators return an lvalue the value_type must be CopyConstructible. If the iterators return an rvalue the value_type must be MoveConstructible.
a_uniq.insert(t)The value_type must be CopyConstructible.
a_uniq.insert(rv)The key_type and the mapped_type (if it exists) must be MoveConstructible.
a_eq.insert(t)The value_type must be CopyConstructible.
a_eq.insert(rv)The key_type and the mapped_type (if it exists) must be MoveConstructible.
a.insert(p, t)The value_type must be CopyConstructible.
a.insert(p, rv)The key_type and the mapped_type (if it exists) must be MoveConstructible.
a.insert(i, j)If the iterators return an lvalue the value_type must be CopyConstructible. If the iterators return an rvalue the key_type and the mapped_type (if it exists) must be MoveConstructible..

Miscellaneous Requirements
map[lvalue-key]The key_type must be CopyConstructible. The mapped_type must be DefaultConstructible and MoveConstructible.
map[rvalue-key]The key_type must be MoveConstructible. The mapped_type must be DefaultConstructible and MoveConstructible.

[ Kona (2007): Howard and Alan to update requirements table in issue with emplace signatures. ]

Proposed resolution:


705. type-trait decay incompletely specified

Section: 20.4.7 [meta.trans.other] Status: Ready Submitter: Thorsten Ottosen Date: 2007-07-08

View all issues with Ready status.

Discussion:

The current working draft has a type-trait decay in 20.4.7 [meta.trans.other].

Its use is to turn C++03 pass-by-value parameters into efficient C++0x pass-by-rvalue-reference parameters. However, the current definition introduces an incompatible change where the cv-qualification of the parameter type is retained. The deduced type should loose such cv-qualification, as pass-by-value does.

Proposed resolution:

In 20.4.7 [meta.trans.other] change the last sentence:

Otherwise the member typedef type equals remove_cv<U>::type.

In 20.3.1.2 [tuple.creation]/1 change:

where each Vi in VTypes is X& if, for the corresponding type Ti in Types, remove_cv<remove_reference<Ti>::type>::type equals reference_wrapper<X>, otherwise Vi is decay<Ti>::type. Let Ui be decay<Ti>::type for each Ti in Types. Then each Vi in VTypes is X& if Ui equals reference_wrapper<X>, otherwise Vi is Ui.


706. make_pair() should behave as make_tuple() wrt. reference_wrapper()

Section: 20.2.3 [pairs] Status: Ready Submitter: Thorsten Ottosen Date: 2007-07-08

View all other issues in [pairs].

View all issues with Ready status.

Discussion:

The current draft has make_pair() in 20.2.3 [pairs]/16 and make_tuple() in 20.3.1.2 [tuple.creation]. make_tuple() detects the presence of reference_wrapper<X> arguments and "unwraps" the reference in such cases. make_pair() would OTOH create a reference_wrapper<X> member. I suggest that the two functions are made to behave similar in this respect to minimize confusion.

Proposed resolution:

In 20.2 [utility] change the synopsis for make_pair() to read

template <class T1, class T2>
  pair<typename decay<T1>::type V1, typename decay<T2>::type V2> make_pair(T1&&, T2&&);

In 20.2.3 [pairs]/16 change the declaration to match the above synopsis. Then change the 20.2.3 [pairs]/17 to:

Returns: pair<typename decay<T1>::type V1,typename decay<T2>::type V2>(forward<T1>(x),forward<T2>(y)) where V1 and V2 are determined as follows: Let Ui be decay<Ti>::type for each Ti. Then each Vi is X& if Ui equals reference_wrapper<X>, otherwise Vi is Ui.


707. null pointer constant for exception_ptr

Section: 18.7.1 [exception] Status: Open Submitter: Jens Maurer Date: 2007-07-20

View all issues with Open status.

Discussion:

From the Toronto Core wiki:

What do you mean by "null pointer constant"? How do you guarantee that exception_ptr() == 1 doesn't work?  Do you even want to prevent that? What's the semantics?  What about void *p = 0; exception_ptr() == p? Maybe disallow those in the interface, but how do you do that with portable C++? Could specify just "make it work".

Peter's response:

null pointer constant as defined in 4.10 [conv.ptr]. Intent is "just make it work", can be implemented as assignment operator taking a unique pointer to member, as in the unspecified bool type idiom.

Proposed resolution:


708. Locales need to be per thread and updated for POSIX changes

Section: 22 [localization] Status: Open Submitter: Peter Dimov Date: 2007-07-28

View all other issues in [localization].

View all issues with Open status.

Discussion:

The POSIX "Extended API Set Part 4,"

http://www.opengroup.org/sib/details.tpl?id=C065

introduces extensions to the C locale mechanism that allow multiple concurrent locales to be used in the same application by introducing a type locale_t that is very similar to std::locale, and a number of _l functions that make use of it.

The global locale (set by setlocale) is now specified to be per- process. If a thread does not call uselocale, the global locale is in effect for that thread. It can install a per-thread locale by using uselocale.

There is also a nice querylocale mechanism by which one can obtain the name (such as "de_DE") for a specific facet, even for combined locales, with no std::locale equivalent.

std::locale should be harmonized with the new POSIX locale_t mechanism and provide equivalents for uselocale and querylocale.

[ Kona (2007): Bill and Nick to provide wording. ]

Proposed resolution:


709. char_traits::not_eof has wrong signature

Section: 21.1.3 [char.traits.specializations] Status: New Submitter: Bo Persson Date: 2007-08-13

View all issues with New status.

Discussion:

The changes made for constexpr in 21.1.3 [char.traits.specializations] have not only changed the not_eof function from pass by const reference to pass by value, it has also changed the parameter type from int_type to char_type.

This doesn't work for type char, and is inconsistent with the requirements in Table 56, Traits requirements, 21.1.1 [char.traits.require].

Pete adds:

For what it's worth, that may not have been an intentional change. N2349, which detailed the changes for adding constant expressions to the library, has strikeout bars through the const and the & that surround the char_type argument, but none through char_type itself. So the intention may have been just to change to pass by value, with text incorrectly copied from the standard.

Proposed resolution:

Change the signature in 21.1.3.1 [char.traits.specializations.char], 21.1.3.2 [char.traits.specializations.char16_t], 21.1.3.3 [char.traits.specializations.char32_t], and 21.1.3.4 [char.traits.specializations.wchar.t] to

static constexpr int_type not_eof(char_type int_type c);

710. Missing postconditions

Section: 20.6.6.2 [util.smartptr.shared] Status: New Submitter: Peter Dimov Date: 2007-08-24

View other active issues in [util.smartptr.shared].

View all other issues in [util.smartptr.shared].

View all issues with New status.

Discussion:

A discussion on comp.std.c++ has identified a contradiction in the shared_ptr specification. The shared_ptr move constructor and the cast functions are missing postconditions for the get() accessor.

Proposed resolution:

Add to 20.6.6.2.1 [util.smartptr.shared.const]:

shared_ptr(shared_ptr&& r);
template<class Y> shared_ptr(shared_ptr<Y>&& r);

Postconditions: *this shall contain the old value of r. r shall be empty. r.get() == 0.

Add to 20.6.6.2.10 [util.smartptr.shared.cast]:

template<class T, class U> shared_ptr<T> static_pointer_cast(shared_ptr<U> const& r);

Postconditions: If w is the return value, w.get() == static_cast<T*>(r.get()) && w.use_count() == r.use_count().

template<class T, class U> shared_ptr<T> dynamic_pointer_cast(shared_ptr<U> const& r);

Postconditions: If w is the return value, w.get() == dynamic_cast<T*>(r.get()).

template<class T, class U> shared_ptr<T> const_pointer_cast(shared_ptr<U> const& r);

Postconditions: If w is the return value, w.get() == const_cast<T*>(r.get()) && w.use_count() == r.use_count().

Alberto Ganesh Barbati has written an alternative proposal where he suggests (among other things) that the casts be respecified in terms of the aliasing constructor as follows:

Change 20.6.6.2.10 [util.smartptr.shared.cast]:

-2- Returns: If r is empty, an empty shared_ptr<T>; otherwise, a shared_ptr<T> object that stores static_cast<T*>(r.get()) and shares ownership with r. shared_ptr<T>(r, static_cast<T*>(r.get()).

-6- Returns:

-10- Returns: If r is empty, an empty shared_ptr<T>; otherwise, a shared_ptr<T> object that stores const_cast<T*>(r.get()) and shares ownership with r. shared_ptr<T>(r, const_cast<T*>(r.get()).

This takes care of the missing postconditions for the casts by bringing in the aliasing constructor postcondition "by reference".


711. Contradiction in empty shared_ptr

Section: 20.6.6.2.5 [util.smartptr.shared.obs] Status: New Submitter: Peter Dimov Date: 2007-08-24

View all other issues in [util.smartptr.shared.obs].

View all issues with New status.

Discussion:

A discussion on comp.std.c++ has identified a contradiction in the shared_ptr specification. The note:

[ Note: this constructor allows creation of an empty shared_ptr instance with a non-NULL stored pointer. -end note ]

after the aliasing constructor

template<class Y> shared_ptr(shared_ptr<Y> const& r, T *p);

reflects the intent of N2351 to, well, allow the creation of an empty shared_ptr with a non-NULL stored pointer.

This is contradicted by the second sentence in the Returns clause of 20.6.6.2.5 [util.smartptr.shared.obs]:

T* get() const;

Returns: the stored pointer. Returns a null pointer if *this is empty.

Proposed resolution:

In keeping the N2351 spirit and obviously my preference, change 20.6.6.2.5 [util.smartptr.shared.obs]:

T* get() const;

Returns: the stored pointer. Returns a null pointer if *this is empty.

Alternative proposed resolution: (I won't be happy if we do this, but it's possible):

Change 20.6.6.2.1 [util.smartptr.shared.const]:

template<class Y> shared_ptr(shared_ptr<Y> const& r, T *p);

Requires: If r is empty, p shall be 0.

[ Note: this constructor allows creation of an empty shared_ptr instance with a non-NULL stored pointer. -- end note ]


713. sort() complexity is too lax

Section: 25.3.1.1 [sort] Status: New Submitter: Matt Austern Date: 2007-08-30

View all issues with New status.

Discussion:

The complexity of sort() is specified as "Approximately N log(N) (where N == last - first ) comparisons on the average", with no worst case complicity specified. The intention was to allow a median-of-three quicksort implementation, which is usually O(N log N) but can be quadratic for pathological inputs. However, there is no longer any reason to allow implementers the freedom to have a worst-cast-quadratic sort algorithm. Implementers who want to use quicksort can use a variant like David Musser's "Introsort" (Software Practice and Experience 27:983-993, 1997), which is guaranteed to be O(N log N) in the worst case without incurring additional overhead in the average case. Most C++ library implementers already do this, and there is no reason not to guarantee it in the standard.

Proposed resolution:

In 25.3.1.1 [sort], change the complexity to "O(N log N)", and remove footnote 266:

Complexity: Approximately O(N log(N)) (where N == last - first ) comparisons on the average.266)

266) If the worst case behavior is important stable_sort() (25.3.1.2) or partial_sort() (25.3.1.3) should be used.


714. search_n complexity is too lax

Section: 25.1.9 [alg.search] Status: New Submitter: Matt Austern Date: 2007-08-30

View all other issues in [alg.search].

View all issues with New status.

Discussion:

The complexity for search_n (25.1.9 [alg.search] par 7) is specified as "At most (last - first ) * count applications of the corresponding predicate if count is positive, or 0 otherwise." This is unnecessarily pessimistic. Regardless of the value of count, there is no reason to examine any element in the range more than once.

Proposed resolution:

Change the complexity to "At most (last - first) applications of the corresponding predicate".

template<class ForwardIterator, class Size, class T> 
  ForwardIterator 
    search_n(ForwardIterator first , ForwardIterator last , Size count , 
             const T& value ); 

template<class ForwardIterator, class Size, class T, 
         class BinaryPredicate> 
  ForwardIterator 
    search_n(ForwardIterator first , ForwardIterator last , Size count , 
             const T& value , BinaryPredicate pred );

Complexity: At most (last - first ) * count applications of the corresponding predicate if count is positive, or 0 otherwise.


715. minmax_element complexity is too lax

Section: 25.3.7 [alg.min.max] Status: New Submitter: Matt Austern Date: 2007-08-30

View all other issues in [alg.min.max].

View all issues with New status.

Discussion:

The complexity for minmax_element (25.3.7 [alg.min.max] par 16) says "At most max(2 * (last - first ) - 2, 0) applications of the corresponding comparisons", i.e. the worst case complexity is no better than calling min_element and max_element separately. This is gratuitously inefficient. There is a well known technique that does better: see section 9.1 of CLRS (Introduction to Algorithms, by Cormen, Leiserson, Rivest, and Stein).

Proposed resolution:

Change 25.3.7 [alg.min.max] to:

template<class ForwardIterator> 
  pair<ForwardIterator, ForwardIterator> 
    minmax_element(ForwardIterator first , ForwardIterator last); 
template<class ForwardIterator, class Compare> 
  pair<ForwardIterator, ForwardIterator> 
    minmax_element(ForwardIterator first , ForwardIterator last , Compare comp);

Returns: make_pair(m, M), where m is min_element(first, last) or min_element(first, last, comp) the first iterator i in [first, last) such that no other element in the range is smaller, and where M is max_element(first, last) or max_element(first, last, comp) the last iterator i in [first, last) such that no other element in the range is larger.

Complexity: At most max(2 * (last - first ) - 2, 0) max(⌊(3/2) (N-1)⌋, 0) applications of the corresponding comparisons predicate, where N is distance(first, last).


716. Production in [re.grammar] not actually modified

Section: 28.13 [re.grammar] Status: New Submitter: Stephan T. Lavavej Date: 2007-08-31

View all issues with New status.

Discussion:

TR1 7.13 [tr.re.grammar]/3 and C++0x WP 28.13 [re.grammar]/3 say:

The following productions within the ECMAScript grammar are modified as follows:

CharacterClass ::
[ [lookahead ∉ {^}] ClassRanges ]
[ ^ ClassRanges ]

This definition for CharacterClass appears to be exactly identical to that in ECMA-262.

Was an actual modification intended here and accidentally omitted, or was this production accidentally included?

Proposed resolution:

Remove this mention of the CharacterClass production.

CharacterClass ::
[ [lookahead ∉ {^}] ClassRanges ]
[ ^ ClassRanges ]

717. Incomplete valarray::operator[] specification in [valarray.access]

Section: 26.5.2.3 [valarray.access] Status: New Submitter: Daniel Krügler Date: 2007-08-27

View all other issues in [valarray.access].

View all issues with New status.

Discussion:

Since the return type of valarray's operator[] const overload has been changed to const T& as described in 389 several paragraphs of the section 26.5.2.3 [valarray.access] are now incompletely specified, because many requirements and guarantees should now also apply to the const overload. Most notably, the address and reference guarantees should be extended to the const overload case.

Proposed resolution:

Change 26.5.2.3 [valarray.access]:

-1- When applied to a constant array, the subscript operator returns a reference to the corresponding element of the array. When applied to a non-constant array, tThe subscript operator returns a reference to the corresponding element of the array.

-3- The expression &a[i+j] == &a[i] + j evaluates as true for all size_t i and size_t j such that i+j is less than the length of the non-constant array a.

-4- Likewise, the expression &a[i] != &b[j] evaluates as true for any two non-constant arrays a and b and for any size_t i and size_t j such that i is less than the length of a and j is less than the length of b. This property indicates an absence of aliasing and may be used to advantage by optimizing compilers.281)

-5- The reference returned by the subscript operator for an non-constant array is guaranteed to be valid until the member function resize(size_t, T) (26.5.2.7) is called for that array or until the lifetime of that array ends, whichever happens first.


718. basic_string is not a sequence

Section: 21.3 [basic.string] Status: New Submitter: Bo Persson Date: 2007-08-18

View other active issues in [basic.string].

View all other issues in [basic.string].

View all issues with New status.

Discussion:

Paragraph 21.3 [basic.string]/3 states:

The class template basic_string conforms to the requirements for a Sequence (23.1.1) and for a Reversible Container (23.1).

First of all, 23.1.1 [sequence.reqmts] is no longer "Sequence" but "Sequence container". Secondly, after the resent changes to containers (emplace, push_back, const_iterator parameters to insert and erase), basic_string is not even close to conform to the current requirements.

Proposed resolution:

Remove this sentence, in recognition of the fact that basic_string is not just a vector-light for literal types, but something quite different, a string abstraction in its own right.


719. std::is_literal type traits should be provided

Section: 20.4 [meta] Status: New Submitter: Daniel Krügler Date: 2007-08-25

View all other issues in [meta].

View all issues with New status.

Discussion:

Since the inclusion of constexpr in the standard draft N2369 we have a new type category "literal", which is defined in 3.9 [basic.types]/p.11:

-11- A type is a literal type if it is:

I strongly suggest that the standard provides a type traits for literal types in 20.4.4.3 [meta.unary.prop] for several reasons:

  1. To keep the traits in sync with existing types.
  2. I see many reasons for programmers to use this trait in template code to provide optimized template definitions for these types, see below.
  3. A user-provided definition of this trait is practically impossible to write portably.

The special problem of reason (c) is that I don't see currently a way to portably test the condition for literal class types:

Here follows a simply example to demonstrate it's usefulness:

template <typename T>
constexpr typename std::enable_if<std::is_literal<T>::value, T>::type
abs(T x) {
  return x < T() ? -x : x;
}

template <typename T>
typename std::enable_if<!std::is_literal<T>::value, T>::type
abs(const T& x) {
  return x < T() ? -x : x;
}

Here we have the possibility to provide a general abs function template that can be used in ICE's if it's argument is a literal type which's value is a constant expression, otherwise we have an optimized version for arguments which are expensive to copy and therefore need the usage of arguments of reference type (instead of const T& we could decide to use T&&, but that is another issue).

Proposed resolution:

In 20.4.2 [meta.type.synop] in the group "type properties", just below the line

template <class T> struct is_pod;

add a new one:

template <class T> struct is_literal;

In 20.4.4.3 [meta.unary.prop], table Type Property Predicates, just below the line for the is_pod property add a new line:

TemplateConditionPreconditions
template <class T> struct is_literal; T is a literal type (3.9) T shall be a complete type, an array of unknown bound, or (possibly cv-qualified) void.

720. Omissions in constexpr usages

Section: 23.2.1 [array], 23.3.5 [template.bitset] Status: New Submitter: Daniel Krügler Date: 2007-08-25

View other active issues in [array].

View all other issues in [array].

View all issues with New status.

Discussion:

  1. The member function bool array<T,N>::empty() const should be a constexpr because this is easily to proof and to implement following it's operational semantics defined by Table 87 (Container requirements) which says: a.size() == 0.
  2. The member function bool bitset<N>::test() const must be a constexpr (otherwise it would violate the specification of constexpr bitset<N>::operator[](size_t) const, because it's return clause delegates to test()).
  3. I wonder how the constructor bitset<N>::bitset(unsigned long) can be declared as a constexpr. Current implementations usually have no such bitset c'tor which would fulfill the requirements of a constexpr c'tor because they have a non-empty c'tor body that typically contains for-loops or memcpy to compute the initialisation. What have I overlooked here?

Proposed resolution:

  1. In the class template definition of 23.2.1 [array]/p. 3 change

    constexpr bool empty() const;
    
  2. In the class template definition of 23.3.5 [template.bitset]/p. 1 change

    constexpr bool test(size_t pos ) const;
    

    and in 23.3.5.2 [bitset.members] change

    constexpr bool test(size_t pos ) const;
    

721. wstring_convert inconsistensies

Section: 22.1.3.2.2 [conversions.string] Status: New Submitter: Bo Persson Date: 2007-08-27

View all issues with New status.

Discussion:

Paragraph 3 says that the Codecvt template parameter shall meet the requirements of std::codecvt, even though std::codecvt itself cannot be used (because of a protected destructor).

How are we going to explain this code to beginning programmers?

template<class I, class E, class S>
struct codecvt : std::codecvt<I, E, S>
{
    ~codecvt()
    { }
};

void main()
{
    std::wstring_convert<codecvt<wchar_t, char, std::mbstate_t> > compiles_ok;
    
    std::wstring_convert<std::codecvt<wchar_t, char, std::mbstate_t> >   not_ok;
}

Proposed resolution:


722. Missing [c.math] functions nanf and nanl

Section: 26.7 [c.math] Status: New Submitter: Daniel Krügler Date: 2007-08-27

View other active issues in [c.math].

View all other issues in [c.math].

View all issues with New status.

Discussion:

In the listing of 26.7 [c.math], table 108: Header <cmath> synopsis I miss the following C99 functions (from 7.12.11.2):

float nanf(const char *tagp);
long double nanl(const char *tagp);

(Note: These functions cannot be overloaded and they are also not listed anywhere else)

Proposed resolution:

In 26.7 [c.math], table 108, section "Functions", add nanf and nanl just after the existing entry nan.


723. basic_regex should be moveable

Section: 28.8 [re.regex] Status: New Submitter: Daniel Krügler Date: 2007-08-29

View all other issues in [re.regex].

View all issues with New status.

Discussion:

According to the current state of the standard draft, the class template basic_regex, as described in 28.8 [re.regex]/3, is neither MoveConstructible nor MoveAssignable. IMO it should be, because typical regex state machines tend to have a rather large data quantum and I have seen several use cases, where a factory function returns regex values, which would take advantage of moveabilities.

Proposed resolution:

  1. In the header <regex> synopsis 28.4 [re.syn], just below the function template swap add two further overloads:

    template <class charT, class traits> 
      void swap(basic_regex<charT, traits>& e1,  basic_regex<charT, traits>& e2);
    template <class charT, class traits>
      void swap(basic_regex<charT, traits>&& e1, basic_regex<charT, traits>& e2);
    template <class charT, class traits>
      void swap(basic_regex<charT, traits>& e1,  basic_regex<charT, traits>&& e2);
    

    In the class definition of basic_regex, just below 28.8 [re.regex]/3, perform the following changes:

  2. Just after the copy c'tor:

    basic_regex(basic_regex&&);
    
  3. Just after the copy-assignment op.:

    basic_regex& operator=(basic_regex&&);
    
  4. Just after the first assign overload insert:

    basic_regex& assign(basic_regex&& that);
    
  5. Change the current swap function to read:

    void swap(basic_regex&&);
    
  6. In 28.8.2 [re.regex.construct], just below the copy c'tor add a corresponding member definition of:

    basic_regex(basic_regex&&);
    
  7. Also in 28.8.2 [re.regex.construct], just below the copy assignment c'tor add a corresponding member definition of:

    basic_regex& operator=(basic_regex&&);
    
  8. In 28.8.3 [re.regex.assign], just below the first assign overload add a corresponding member definition of:

    basic_regex& assign(basic_regex&& that);
    
  9. In 28.8.6 [re.regex.swap], change the signature of swap to say:

    void swap(basic_regex&& e);
    
  10. In 28.8.7.1 [re.regex.nmswap], just below the single binary swap function, add the two missing overloads:

    template <class charT, class traits>
      void swap(basic_regex<charT, traits>&& e1, basic_regex<charT, traits>& e2);
    template <class charT, class traits>
      void swap(basic_regex<charT, traits>& e1, basic_regex<charT, traits>&& e2);
    

Of course there would be need of corresponding proper standardese to describe these additions.


724. DefaultConstructible is not defined

Section: 20.1.1 [utility.arg.requirements] Status: New Submitter: Pablo Halpern Date: 2007-09-12

View other active issues in [utility.arg.requirements].

View all other issues in [utility.arg.requirements].

View all issues with New status.

Discussion:

The DefaultConstructible requirement is referenced in several places in the August 2007 working draft N2369, but is not defined anywhere.

Proposed resolution:

In section 20.1.1 [utility.arg.requirements], before table 33, add the following table:

Table 33: DefaultConstructible requirements

expression

post-condition

T t;
T()

T is default constructed.


725. Optional sequence container requirements column label

Section: 23.1.1 [sequence.reqmts] Status: New Submitter: David Abrahams Date: 2007-09-16

View all other issues in [sequence.reqmts].

View all issues with New status.

Discussion:

Table 90: (Optional sequence container operations) states the "assertion note pre/post-condition" of operator[] to be

*(a.begin() + n)

Surely that's meant to be "operational semantics?"

Proposed resolution:

Table 90: Optional sequence container operations
expression return type assertion/note
pre/post-condition

operational semantics
container

726. Missing regex_replace() overloads

Section: 28.11.4 [re.alg.replace] Status: New Submitter: Stephan T. Lavavej Date: 2007-09-22

View other active issues in [re.alg.replace].

View all other issues in [re.alg.replace].

View all issues with New status.

Discussion:

Two overloads of regex_replace() are currently provided:

template <class OutputIterator, class BidirectionalIterator, 
    class traits, class charT> 
  OutputIterator 
  regex_replace(OutputIterator out, 
                BidirectionalIterator first, BidirectionalIterator last, 
                const basic_regex<charT, traits>& e, 
                const basic_string<charT>& fmt, 
                regex_constants::match_flag_type flags = 
                  regex_constants::match_default);
 
template <class traits, class charT> 
  basic_string<charT> 
  regex_replace(const basic_string<charT>& s, 
                const basic_regex<charT, traits>& e, 
                const basic_string<charT>& fmt, 
                regex_constants::match_flag_type flags = 
                  regex_constants::match_default);
  1. Overloads taking const charT * are provided for regex_match() and regex_search(), but not regex_replace().  This is inconsistent.
  2. The absence of const charT * overloads prevents ordinary-looking code from compiling, such as:

    const string s("kitten");
    const regex r("en");
    cout << regex_replace(s, r, "y") << endl;
    

    The compiler error message will be something like "could not deduce template argument for 'const std::basic_string<_Elem> &' from 'const char[1]'".

    Users expect that anything taking a basic_string<charT> can also take a const charT *.  In their own code, when they write a function taking std::string (or std::wstring), they can pass a const char * (or const wchar_t *), thanks to basic_string's implicit constructor.  Because the regex algorithms are templated on charT, they can't rely on basic_string's implicit constructor (as the compiler error message indicates, template argument deduction fails first).

    If a user figures out what the compiler error message means, workarounds are available - but they are all verbose.  Explicit template arguments could be given to regex_replace(), allowing basic_string's implicit constructor to be invoked - but charT is the last template argument, not the first, so this would be extremely verbose.  Therefore, constructing a basic_string from each C string is the simplest workaround.

  3. There is an efficiency consideration: constructing basic_strings can impose performance costs that could be avoided by a library implementation taking C strings and dealing with them directly.  (Currently, for replacement sources, C strings can be converted into iterator pairs at the cost of verbosity, but for format strings, there is no way to avoid constructing a basic_string.)

Proposed resolution:

Provide additional overloads for regex_replace(): one additional overload of the iterator-based form (taking const charT* fmt), and three additional overloads of the convenience form (one taking const charT* str, another taking const charT* fmt, and the third taking both const charT* str and const charT* fmt). 28.11.4 [re.alg.replace]:

template <class OutputIterator, class BidirectionalIterator, 
    class traits, class charT> 
  OutputIterator 
  regex_replace(OutputIterator out, 
                BidirectionalIterator first, BidirectionalIterator last, 
                const basic_regex<charT, traits>& e, 
                const basic_string<charT>& fmt, 
                regex_constants::match_flag_type flags = 
                  regex_constants::match_default);

template <class OutputIterator, class BidirectionalIterator, 
    class traits, class charT> 
  OutputIterator 
  regex_replace(OutputIterator out, 
                BidirectionalIterator first, BidirectionalIterator last, 
                const basic_regex<charT, traits>& e, 
                const charT* fmt, 
                regex_constants::match_flag_type flags = 
                  regex_constants::match_default);

...

template <class traits, class charT> 
  basic_string<charT> 
  regex_replace(const basic_string<charT>& s, 
                const basic_regex<charT, traits>& e, 
                const basic_string<charT>& fmt, 
                regex_constants::match_flag_type flags = 
                  regex_constants::match_default);

template <class traits, class charT> 
  basic_string<charT> 
  regex_replace(const basic_string<charT>& s, 
                const basic_regex<charT, traits>& e, 
                const charT* fmt, 
                regex_constants::match_flag_type flags = 
                  regex_constants::match_default);

template <class traits, class charT> 
  basic_string<charT> 
  regex_replace(const charT* s, 
                const basic_regex<charT, traits>& e, 
                const basic_string<charT>& fmt, 
                regex_constants::match_flag_type flags = 
                  regex_constants::match_default);

template <class traits, class charT> 
  basic_string<charT> 
  regex_replace(const charT* s, 
                const basic_regex<charT, traits>& e, 
                const charT* fmt, 
                regex_constants::match_flag_type flags = 
                  regex_constants::match_default);

727. regex_replace() doesn't accept basic_strings with custom traits and allocators

Section: 28.11.4 [re.alg.replace] Status: New Submitter: Stephan T. Lavavej Date: 2007-09-22

View other active issues in [re.alg.replace].

View all other issues in [re.alg.replace].

View all issues with New status.

Discussion:

regex_match() and regex_search() take const basic_string<charT, ST, SA>&regex_replace() takes const basic_string<charT>&.  This prevents regex_replace() from accepting basic_strings with custom traits and allocators.

Proposed resolution:

Overloads of regex_replace() taking basic_string should be additionally templated on class ST, class SA and take const basic_string<charT, ST, SA>&.  Consistency with regex_match() and regex_search() would place class ST, class SA as the first template arguments; compatibility with existing code using TR1 and giving explicit template arguments to regex_replace() would place class ST, class SA as the last template arguments.


728. Problem in [rand.eng.mers]/6

Section: 26.4.3.2 [rand.eng.mers] Status: New Submitter: Stephan Tolksdorf Date: 2007-09-21

View all issues with New status.

Discussion:

The mersenne_twister_engine is required to use a seeding method that is given as an algorithm parameterized over the number of bits W. I doubt whether the given generalization of an algorithm that was originally developed only for unsigned 32-bit integers is appropriate for other bit widths. For instance, W could be theoretically 16 and UIntType a 16-bit integer, in which case the given multiplier would not fit into the UIntType. Moreover, T. Nishimura and M. Matsumoto have chosen a dif ferent multiplier for their 64 bit Mersenne Twister [reference].

I see two possible resolutions:

  1. Restrict the parameter W of the mersenne_twister_template to values of 32 or 64 and use the multiplier from [the above reference] for the 64-bit case (my preference)
  2. Interpret the state array for any W as a 32-bit array of appropriate length (and a specified byte order) and always employ the 32-bit algorithm for seeding

See N2423 for further discussion.

Proposed resolution:

See N2423 for the proposed resolution.


729. Problem in [rand.req.eng]/3

Section: 26.4.1.3 [rand.req.eng] Status: New Submitter: Stephan Tolksdorf Date: 2007-09-21

View all other issues in [rand.req.eng].

View all issues with New status.

Discussion:

The 3rd table row in 26.4.1.3 [rand.req.eng]/3 requires random number engines to accept any arithmetic type as a seed, which is then casted to the engine's result_type and subsequently used for seeding the state of the engine. The requirement stated as "Creates an engine with initial state determined by static_cast<X::result_type>(s)" forces random number engines to either use a seeding method that completely depends on the result_type (see the discussion of seeding for the mersenne_twister_engine in point T2 above) or at least to throw away "bits of randomness" in the seed value if the result_type is smaller than the seed type. This seems to be inappropriate for many modern random number generators, in particular F2-linear or cryptographic ones, which operate on an internal bit array that in principle is independent of the type of numbers returned.

Posible resolution: I propose to change the wording to a version similar to "Creates an engine with initial state determined by static_cast<UintType>(s), where UintType is an implementation specific unsigned integer type."

Additionally, the definition of s in 26.4.1.3 [rand.req.eng]/1 c) could be restricted to unsigned integer types.

Similarly, the type of the seed in 26.4.1.4 [rand.req.adapt]/3 e) could be left unspecified.

See N2423 for further discussion.

Proposed resolution:

See N2423 for further discussion.


730. Comment on [rand.req.adapt]/3 e)

Section: 26.4.1.4 [rand.req.adapt] Status: New Submitter: Stephan Tolksdorf Date: 2007-09-21

View all issues with New status.

Discussion:

If an engine adaptor is invoked with an argument of type seed_seq, then all base engines are specified to be seeded with this seed_seq. As seed_seq's randomization method is qualified as constant, this procedure will ef fectively initialize all base engines with the same seed (though the resulting state might still dif fer to a certain degree if the engines are of different types). It is not clear whether this mode of operation is in general appropriate, hence -- as far as the stated requirements are of general nature and not just specific to the engine adaptors provided by the library -- it might be better to leave the behaviour unspecified, since the current definition of seed_seq does not allow for a generally satisfying specification.

Posssible resolution: [As above]

See N2423 for further discussion.

Proposed resolution:

See N2423 for the proposed resolution.


731. proposal for a customizable seed_seq

Section: 26.4.7.1 [rand.util.seedseq] Status: New Submitter: Stephan Tolksdorf Date: 2007-09-21

View all other issues in [rand.util.seedseq].

View all issues with New status.

Discussion:

The proper way to seed random number engines seems to be the most frequently discussed issue of the 26.4 [rand] proposal. While the new seed_seq approach is already rather general and probably sufficient for most situations, it is unlikely to be optimal in every case (one problem was pointed out in point T5 above). In some situations it might, for instance, be better to seed the state with a cryptographic generator.

In my opinion this is a pretty strong argument for extending the standard with a simple facility to customize the seeding procedure. This could, for example, be done with the following minimal changes:

Possible resolution:

  1. Turn the interface specification of 26.4.7.1 [rand.util.seedseq]/2 into a "SeedSeq" requirement, where the exact behaviour of the constructors and the randomize method are left unspecified and where the const qualification for randomize is removed. Classes implementing this interface are additionally required to specialize the traits class in c).
  2. Provide the class seed_seq as a default implementation of the SeedSeq interface.
  3. Supplement the seed_seq with a traits class

    template <typename T> 
    struct is_seed_seq { static const bool value = false; }
    

    and the specialization

    template <> 
    struct is_seed_seq<seed_seq> { static const bool value = true; }
    

    which users can supplement with further specializations.

  4. Change 26.4.1.3 [rand.req.eng]/1 d) to "q is an lvalue of a type that fulfils the SeedSeq requirements", and modify the constructors and seed methods in 26.4.3 [rand.eng] appropriately (the actual implementation could be done using the SFINAE technique).

Proposed resolution:

See N2423 for the proposed resolution.


732. Defect in [rand.dist.samp.genpdf]

Section: 26.4.8.5.3 [rand.dist.samp.genpdf] Status: New Submitter: Stephan Tolksdorf Date: 2007-09-21

View all issues with New status.

Discussion:

26.4.8.5.3 [rand.dist.samp.genpdf] describes the interface for a distribution template that is meant to simulate random numbers from any general distribution given only the density and the support of the distribution. I'm not aware of any general purpose algorithm that would be capable of correctly and efficiently implementing the described functionality. From what I know, this is essentially an unsolved research problem. Existing algorithms either require more knowledge about the distribution and the problem domain or work only under very limited circumstances. Even the state of the art special purpose library UNU.RAN does not solve the problem in full generality, and in any case, testing and customer support for such a library feature would be a nightmare.

Possible resolution: For these reasons, I propose to delete section 26.4.8.5.3 [rand.dist.samp.genpdf].

Proposed resolution:

See N2423 for the proposed resolution.


733. Comment on [rand.req.dist]/9

Section: 26.4.1.5 [rand.req.dist] Status: New Submitter: Stephan Tolksdorf Date: 2007-09-21

View all issues with New status.

Discussion:

The requirement "P shall have a declaration of the form typedef X distribution_- type" effectively makes the use of inheritance for implementing distributions very inconvenient, because the child of a distribution class in general will not satisfy this requirement. In my opinion the benefits of having a typedef in the parameter class pointing back to the distribution class are not worth the hassle this requirement causes. [In my code base I never made use of the nested typedef but on several occasions could have profited from being able to use simple inheritance for the implementation of a distribution class.]

Proposed resolution: I propose to drop this requirement.

Proposed resolution:

See N2423 for the proposed resolution.


734. Unnecessary restriction in [rand.dist.norm.chisq]

Section: 26.4.8.4.3 [rand.dist.norm.chisq] Status: New Submitter: Stephan Tolksdorf Date: 2007-09-21

View all issues with New status.

Discussion:

chi_squared_distribution, fisher_f_distribution and student_t_distribution have parameters for the "degrees of freedom" n and m that are specified as integers. For the following two reasons this is an unnecessary restriction: First, in many applications such as Bayesian inference or Monte Carlo simulations it is more convenient to treat the respective param- eters as continuous variables. Second, the standard non-naive algorithms (i.e. O(1) algorithms) for simulating from these distributions work with floating-point parameters anyway (all three distributions could be easily implemented using the Gamma distribution, for instance).

Similar arguments could in principle be made for the parameters t and k of the discrete binomial_distribution and negative_binomial_distribution, though in both cases continuous parameters are less frequently used in practice and in case of the binomial_distribution the implementation would be significantly complicated by a non-discrete parameter (in most implementations one would need an approximation of the log-gamma function instead of just the log-factorial function).

Possible resolution: For these reasons, I propose to change the type of the respective parameters to double.

Proposed resolution:

See N2423 for the proposed resolution.


735. Unfortunate naming

Section: 26.4.8.2.2 [rand.dist.bern.bin], 26.4.8.2.4 [rand.dist.bern.negbin] Status: New Submitter: Stephan Tolksdorf Date: 2007-09-21

View all issues with New status.

Discussion:

In my opinion the choice of name for the t parameter of the binomial_distribution is very unfortunate. In virtually every internet reference, book and software implementation this parameter is called n instead, see for example Wikipedia, Mathworld, Evans et al. (1993) Statistical Distributions, 2nd E., Wiley, p. 38, the R statistical computing language, p. 926, Mathematica and Matlab.

Similarly, the choice of k for the parameter of the negative binomial distributions is rather unusual. The most common choice for the negative binomial distribution seems to be r instead.

Choosing unusual names for the parameters causes confusion among users and makes the interface unnecessarily inconvenient to use.

Possible resolution: For these reasons, I propose to change the name of the respective parameters to n and r.

Proposed resolution:

See N2423 for the proposed resolution.


736. Comment on [rand.dist.samp.discrete]

Section: 26.4.8.5.1 [rand.dist.samp.discrete] Status: New Submitter: Stephan Tolksdorf Date: 2007-09-21

View all issues with New status.

Discussion:

  1. The specification for discrete_distribution requires the member probabilities() to return a vector of standardized probabilities, which forces the implementation every time to divide each probability by the sum of all probabilities, as the sum will in practice almost never be exactly 1.0. This is unnecessarily inef ficient as the implementation would otherwise not need to compute the standardized probabilities at all and could instead work with the non-standardized probabilities and the sum. If there was no standardization the user would just get back the probabilities that were previously supplied to the distribution object, which to me seems to be the more obvious solution.
  2. The behaviour of discrete_distribution is not specified in case the number of given probabilities is larger than the maximum number representable by the IntType.

Possible resolution: I propose to change the specification such that the non-standardized probabilities need to be returned and that an additional requirement is included for the number of probabilities to be smaller than the maximum of IntType.

Proposed resolution:

See N2423 for the proposed resolution.


737. Comment on [rand.dist.samp.pconst]

Section: 26.4.8.5.2 [rand.dist.samp.pconst] Status: New Submitter: Stephan Tolksdorf Date: 2007-09-21

View all issues with New status.

Discussion:

  1. The discussion in point T11 above regarding probabilities() similarly applies to the method densities() of piecewise_constant_distribution.
  2. The design of the constructor

    template <class InputIteratorB, class InputIteratorW> 
    piecewise_constant_distribution( InputIteratorB firstB, InputIteratorB lastB, 
                                     InputIteratorW firstW);
    

    is unnecessarily unsafe, as there is no separate end-iterator given for the weights. I can't see any performance or convenience reasons that would justify the risks inherent in such a function interface, in particular the risk that input error might go unnoticed.

Possible resolution: I propose to add an InputIteratorW lastW argument to the interface.

Proposed resolution:

See N2423 for the proposed resolution.


738. Editorial issue in [rand.adapt.disc]/3

Section: 26.4.4.1 [rand.adapt.disc] Status: New Submitter: Stephan Tolksdorf Date: 2007-09-21

View all issues with New status.

Discussion:

Since the template parameter p and r are of type size_t, the member n in the class exposition should have type size_t, too.

Proposed resolution:

See N2423 for the proposed resolution.


739. Defect in [rand.util.canonical]/3

Section: 26.4.7.2 [rand.util.canonical] Status: New Submitter: Stephan Tolksdorf Date: 2007-09-21

View all other issues in [rand.util.canonical].

View all issues with New status.

Discussion:

The complexity of generate_canonical is specified to be "exactly k=max(1, ceil(b/log2 R)) invocations of g". This terms involves a logarithm that is not rounded and hence can not (in general) be computed at compile time. As this function template is performance critical, I propose to replace ceil(b/log2 R) with ceil(b/floor(log2 R)).

See N2423 for further discussion.

Proposed resolution:

See N2423 for the proposed resolution.


740. Please remove *_ptr<T[N]>

Section: 20.6.5.4 [unique.ptr.compiletime] Status: New Submitter: Herb Sutter Date: 2007-10-04

View all issues with New status.

Discussion:

Please don't provide *_ptr<T[N]>. It doesn't enable any useful bounds-checking (e.g., you could imagine that doing op++ on a shared_ptr<T[N]> yields a shared_ptr<T[N-1]>, but that promising path immediately falters on op-- which can't reliably dereference because we don't know the lower bound). Also, most buffers you'd want to point to don't have a compile-time known size.

To enable any bounds-checking would require run-time information, with the usual triplet: base (lower bound), current offset, and max offset (upper bound). And I can sympathize with the point of view that you wouldn't want to require this on *_ptr itself. But please let's not follow the <T[N]> path, especially not with additional functions to query the bounds etc., because this sets wrong user expectations by embarking on a path that doesn't go all the way to bounds checking as it seems to imply.

If bounds checking is desired, consider a checked_*_ptr instead (e.g., checked_shared_ptr). And make the interfaces otherwise identical so that user code could easily #define/typedef between prepending checked_ on debug builds and not doing so on release builds (for example).

Note that some may object that checked_*_ptr may seem to make the smart pointer more like vector, and we don't want two ways to spell vector. I don't agree, but if that were true that would be another reason to remove *_ptr<T[N]> which equally makes the smart pointer more like std::array. :-)

Proposed resolution:

Change the synopsis under 20.6.5 [unique.ptr] p2:

...
template<class T> struct default_delete; 
template<class T> struct default_delete<T[]>; 
template<class T, size_t N> struct default_delete<T[N]>;

template<class T, class D = default_delete<T>> class unique_ptr; 
template<class T, class D> class unique_ptr<T[], D>; 
template<class T, class D, size_t N> class unique_ptr<T[N], D>;
...

Remove the entire section 20.6.5.1.3 [unique.ptr.dltr.dflt2] default_delete<T[N]>.

Remove the entire section 20.6.5.4 [unique.ptr.compiletime] unique_ptr for array objects with a compile time length and its subsections: 20.6.5.4.1 [unique.ptr.compiletime.dtor], 20.6.5.4.2 [unique.ptr.compiletime.observers], 20.6.5.4.3 [unique.ptr.compiletime.modifiers].


741. Const-incorrect get_deleter function for shared_ptr

Section: 20.6.6.2.11 [util.smartptr.getdeleter] Status: New Submitter: Daniel Krügler Date: 2007-09-27

View all other issues in [util.smartptr.getdeleter].

View all issues with New status.

Discussion:

The following issue was raised by Alf P. Steinbach in c.l.c++.mod:

According to the recent draft N2369, both the header memory synopsis of 20.6 [memory] and 20.6.6.2.11 [util.smartptr.getdeleter] declare:

template<class D, class T> D* get_deleter(shared_ptr<T> const& p);

This allows to retrieve the pointer to a mutable deleter of a const shared_ptr (if that owns one) and therefore contradicts the usual philosophy that associated functors are either read-only (e.g. key_comp or value_comp of std::map) or do at least reflect the mutability of the owner (as seen for the both overloads of unique_ptr::get_deleter). Even the next similar counter-part of get_deleter - the two overloads of function::target in the class template function synopsis 20.5.15.2 [func.wrap.func] or in 20.5.15.2.5 [func.wrap.func.targ] - do properly mirror the const-state of the owner.

Possible proposed resolutions:

Replace the declarations of get_deleter in the header <memory> synopsis of 20.6 [memory] and in 20.6.6.2.11 [util.smartptr.getdeleter] by one of the following alternatives (A) or (B):

  1. Provide only the immutable variant. This would reflect the current praxis of container::get_allocator(), map::key_comp(), or map::value_comp.
    template<class D, class T> const D* get_deleter(shared_ptr<T> const& p);
    
  2. Just remove the function.

Alberto Ganesh Barbati adds:

  1. Replace it with two functions:

    template <class D, class T> D get_deleter(shared_ptr<T> const&);
    template <class D, class T> bool has_deleter(shared_ptr<T> const&);
    

    The first one would throw if D is the wrong type, while the latter would never throw. This approach would reflect the current praxis of use_facet/has_facet, with the twist of returning the deleter by value as container::get_allocator() do.

Peter Dimov adds:

My favorite option is "not a defect". A, B and C break useful code.

Proposed resolution:


742. Enabling swap for proxy iterators

Section: 20.1.1 [utility.arg.requirements] Status: New Submitter: Howard Hinnant Date: 2007-10-10

View other active issues in [utility.arg.requirements].

View all other issues in [utility.arg.requirements].

View all issues with New status.

Discussion:

This issue was split from 672. 672 now just deals with changing the requirements of T in the Swappable requirement from CopyConstructible and CopyAssignable to MoveConstructible and MoveAssignable.

This issue seeks to widen the Swappable requirement to support proxy iterators. Here is example code:

namespace Mine {

template <class T>
struct proxy {...};

template <class T>
struct proxied_iterator
{
   typedef T value_type;
   typedef proxy<T> reference;
   reference operator*() const;
   ...
};

struct A
{
   // heavy type, has an optimized swap, maybe isn't even copyable or movable, just swappable
   void swap(A&);
   ...
};

void swap(A&, A&);
void swap(proxy<A>, A&);
void swap(A&, proxy<A>);
void swap(proxy<A>, proxy<A>);

}  // Mine

...

Mine::proxied_iterator<Mine::A> i(...)
Mine::A a;
swap(*i1, a);

The key point to note in the above code is that in the call to swap, *i1 and a are different types (currently types can only be Swappable with the same type). A secondary point is that to support proxies, one must be able to pass rvalues to swap. But note that I am not stating that the general purpose std::swap should accept rvalues! Only that overloaded swaps, as in the example above, be allowed to take rvalues.

That is, no standard library code needs to change. We simply need to have a more flexible definition of Swappable.

Proposed resolution:

Change 20.1.1 [utility.arg.requirements]:

-1- The template definitions in the C++ Standard Library refer to various named requirements whose details are set out in tables 31-38. In these tables, T and V are is a types to be supplied by a C++ program instantiating a template; a, b, and c are values of type const T; s and t are modifiable lvalues of type T; u is a value of type (possibly const) T; and rv is a non-const rvalue of type T; w is a value of type T; and v is a value of type V.

Table 37: Swappable requirements [swappable]
expressionreturn typepost-condition
swap(sw,tv)void tw has the value originally held by uv, and uv has the value originally held by tw

The Swappable requirement is met by satisfying one or more of the following conditions:

  • T is Swappable if T and V are the same type and T satisfies the CopyConstructible MoveConstructible requirements (Table 34 33) and the CopyAssignable MoveAssignable requirements (Table 36 35);
  • T is Swappable with V if a namespace scope function named swap exists in the same namespace as the definition of T or V, such that the expression swap(tw,u v) is valid and has the semantics described in this table.

743. rvalue swap for shared_ptr

Section: 20.6.6.2.9 [util.smartptr.shared.spec] Status: New Submitter: Howard Hinnant Date: 2007-10-10

View all issues with New status.

Discussion:

When the LWG looked at 674 in Kona the following note was made:

We may need to open an issue to deal with the question of whether shared_ptr needs an rvalue swap.

This issue was opened in response to that note.

I believe allowing rvalue shared_ptrs to swap is both appropriate, and consistent with how other library components are currently specified.

Proposed resolution:

Change the synopsis in 20.6.6.2 [util.smartptr.shared]:

void swap(shared_ptr&& r);
...
template<class T> void swap(shared_ptr<T>& a, shared_ptr<T>& b);
template<class T> void swap(shared_ptr<T>&& a, shared_ptr<T>& b);
template<class T> void swap(shared_ptr<T>& a, shared_ptr<T>&& b);

Change 20.6.6.2.4 [util.smartptr.shared.mod]:

void swap(shared_ptr&& r);

Change 20.6.6.2.9 [util.smartptr.shared.spec]:

template<class T> void swap(shared_ptr<T>& a, shared_ptr<T>& b);
template<class T> void swap(shared_ptr<T>&& a, shared_ptr<T>& b);
template<class T> void swap(shared_ptr<T>& a, shared_ptr<T>&& b);

744. What is the lifetime of an exception pointed to by an exception_ptr?

Section: 18.7.5 [propagation] Status: New Submitter: Alisdair Meredith Date: 2007-10-10

View other active issues in [propagation].

View all other issues in [propagation].

View all issues with New status.

Discussion:

Without some lifetime guarantee, it is hard to know how this type can be used.  Very specifically, I don't see how the current wording would guarantee and exception_ptr caught at the end of one thread could be safely stored and rethrown in another thread - the original motivation for this API.

(Peter Dimov agreed it should be clearer, maybe a non-normative note to explain?)

Proposed resolution:


745. copy_exception API slices.

Section: 18.7.5 [propagation] Status: New Submitter: Alisdair Meredith Date: 2007-10-10

View other active issues in [propagation].

View all other issues in [propagation].

View all issues with New status.

Discussion:

It could be I did not understand the design rationale, but I thought copy_exception would produce an exception_ptr to the most-derived (dynamic) type of the passed exception.  Instead it slices, which appears to be less useful, and a likely source of FAQ questions in the future.

(Peter Dimov suggests NAD)

Proposed resolution:


746. current_exception may fail with bad_alloc

Section: 18.7.5 [propagation] Status: New Submitter: Alisdair Meredith Date: 2007-10-10

View other active issues in [propagation].

View all other issues in [propagation].

View all issues with New status.

Discussion:

I understand that the attempt to copy an exception may run out of memory, but I believe this is the only part of the standard that mandates failure with specifically bad_alloc, as opposed to allowing an implementation-defined type derived from bad_alloc.  For instance, the Core language for a failed new expression is:

Any other allocation function that fails to allocate storage shall indicate failure only by throwing an exception of a type that would match a handler (15.3) of type std::bad_alloc (18.5.2.1).

I think we should allow similar freedom here (or add a blanket compatible-exception freedom paragraph in 17)

I prefer the clause 17 approach myself, and maybe clean up any outstanding wording that could also rely on it.

Proposed resolution:


747. We have 3 separate type traits to identify classes supporting no-throw operations

Section: 20.4.4.3 [meta.unary.prop] Status: New Submitter: Alisdair Meredith Date: 2007-10-10

View other active issues in [meta.unary.prop].

View all other issues in [meta.unary.prop].

View all issues with New status.

Discussion:

We have 3 separate type traits to identify classes supporting no-throw operations, which are very useful when trying to provide exception safety guarantees.  However, I'm not entirely clear on what the current wording requires of a conforming implementation.  To quote from has_nothrow_default_constructor:

or T is a class type with a default constructor that is known not to throw any exceptions

What level of magic do we expect to deduce if this is known?

E.g.

struct test{
 int x;
 test() : x() {}
};

Should I expect a conforming compiler to  assert( has_nothrow_constructor<test>::value )

Is this a QoI issue?

Should I expect to 'know' only if-and-only-if there is an inline definition available?

Should I never expect that to be true, and insist that the user supplies an empty throw spec if they want to assert the no-throw guarantee?

It would be helpful to maybe have a footnote explaining what is required, but right now I don't know what to suggest putting in the footnote.

(agreement since is that trivial ops and explicit no-throws are required. Open if QoI should be allowed to detect further)

Proposed resolution:


748. The is_abstract type trait is defined by reference to 10.4.

Section: 20.4.4.3 [meta.unary.prop] Status: New Submitter: Alisdair Meredith Date: 2007-10-10

View other active issues in [meta.unary.prop].

View all other issues in [meta.unary.prop].

View all issues with New status.

Discussion:

I am trying to decide is a pure virtual function is a necessary as well as sufficient requirement to be classified as abstract?

For instance, is the following (non-polymorphic) type considered abstract?

struct abstract {
protected:
 abstract(){}
 abstract( abstract const & ) {}
 ~abstract() {}
};

(Suggested that this may be NAD, with an editorial fix-up from Pete on the core wording to make clear that abstract requires a pure virtual function)

Proposed resolution:


749. Currently has_nothrow_copy_constructor<T>::value is true if T has 'a' nothrow copy constructor.

Section: 20.4.4.3 [meta.unary.prop] Status: New Submitter: Alisdair Meredith Date: 2007-10-10

View other active issues in [meta.unary.prop].

View all other issues in [meta.unary.prop].

View all issues with New status.

Discussion:

Unfortunately a class can have multiple copy constructors, and I believe to be useful this trait should only return true is ALL copy constructors are no-throw.

For instance:

struct awkward {
 awkward( const awkward & ) throw() {}
 awkward( awkward & ) { throw "oops"; } };

Proposed resolution:


750. The current definition for is_convertible requires that the type be implicitly convertible, so explicit constructors are ignored.

Section: 20.4.5 [meta.rel] Status: New Submitter: Alisdair Meredith Date: 2007-10-10

View all issues with New status.

Discussion:

With the pending arrival of explicit conversion functions though, I'm wondering if we want an additional trait, is_explictly_convertible?

Proposed resolution:


751. change pass-by-reference members of vector<bool> to pass-by-value?

Section: 23.2.6 [vector.bool] Status: New Submitter: Alisdair Meredith Date: 2007-10-10

View all issues with New status.

Discussion:

A number of vector<bool> members take const bool& as arguments. Is there any chance we could change them to pass-by-value or would I be wasting everyone's time if wrote up an issue?

Proposed resolution:


752. Allocator complexity requirement

Section: 20.1.2 [allocator.requirements] Status: New Submitter: Hans Boehm Date: 2007-10-11

View other active issues in [allocator.requirements].

View all other issues in [allocator.requirements].

View all issues with New status.

Discussion:

Did LWG recently discuss 20.1.2 [allocator.requirements]-2, which states that "All the operations on the allocators are expected to be amortized constant time."?

As I think I pointed out earlier, this is currently fiction for allocate() if it has to obtain memory from the OS, and it's unclear to me how to interpret this for construct() and destroy() if they deal with large objects.  Would it be controversial to officially let these take time linear in the size of the object, as they already do in real life?

Allocate() more blatantly takes time proportional to the size of the object if you mix in GC.  But it's not really a new problem, and I think we'd be confusing things by leaving the bogus requirements there.  The current requirement on allocate() is generally not important anyway, since it takes O(size) to construct objects in the resulting space. There are real performance issues here, but they're all concerned with the constants, not the asymptotic complexity.

Proposed resolution:

Change 20.1.2 [allocator.requirements]/2:

-2- Table 39 describes the requirements on types manipulated through allocators. All the operations on the allocators are expected to be amortized constant time, except that allocate and construct may require time proportional to the size of the object allocated or constructed. Table 40 describes the requirements on allocator types.


753. Move constructor in draft

Section: 20.1.1 [utility.arg.requirements] Status: New Submitter: Yechezkel Mett Date: 2007-10-14

View other active issues in [utility.arg.requirements].

View all other issues in [utility.arg.requirements].

View all issues with New status.

Discussion:

The draft standard n2369 uses the term move constructor in a few places, but doesn't seem to define it.

MoveConstructible requirements are defined in Table 33 in 20.1.1 [utility.arg.requirements] as follows:

MoveConstructible requirements
expression post-condition
T t = rv t is equivalent to the value of rv before the construction
[Note: There is no requirement on the value of rv after the construction. -- end note]

(where rv is a non-const rvalue of type T).

So I assume the move constructor is the constructor that would be used in filling the above requirement.

For vector::reserve, vector::resize and the vector modifiers given in 23.2.5.4 [vector.modifiers] we have

Requires: If value_type has a move constructor, that constructor shall not throw any exceptions.

Firstly "If value_type has a move constructor" is superfluous; every type which can be put into a vector has a move constructor (a copy constructor is also a move constructor). Secondly it means that for any value_type which has a throwing copy constructor and no other move constructor these functions cannot be used -- which I think will come as a shock to people who have been using such types in vector until now!

I can see two ways to correct this. The simpler, which is presumably what was intended, is to say "If value_type has a move constructor and no copy constructor, the move constructor shall not throw any exceptions" or "If value_type has a move constructor which changes the value of its parameter,".

The other alternative is add to MoveConstructible the requirement that the expression does not throw. This would mean that not every type that satisfies the CopyConstructible requirements also satisfies the MoveConstructible requirements. It would mean changing requirements in various places in the draft to allow either MoveConstructible or CopyConstructible, but I think the result would be clearer and possibly more concise too.

Proposed resolution:


754. Ambiguous return clause for std::uninitialized_copy

Section: 20.6.4.1 [uninitialized.copy] Status: New Submitter: Daniel Krügler Date: 2007-10-15

View other active issues in [uninitialized.copy].

View all other issues in [uninitialized.copy].

View all issues with New status.

Discussion:

14882-2003, [lib.uninitialized.copy] is currently written as follows:

template <class InputIterator, class ForwardIterator>
  ForwardIterator uninitialized_copy(InputIterator first, InputIterator last,
                                     ForwardIterator result);

-1- Effects:

for (; first != last; ++result, ++first)
  new (static_cast<void*>(&*result))
    typename iterator_traits<ForwardIterator>::value_type(*first);

-2- Returns: result

similarily for N2369, and its corresponding section 20.6.4.1 [uninitialized.copy].

It's not clear to me what the return clause is supposed to mean, I see two possible interpretations:

  1. The notion of result is supposed to mean the value given by the function parameter result [Note to the issue editor: Please use italics for result]. This seems somewhat implied by recognizing that both the function parameter and the name used in the clause do have the same italic font.
  2. The notion of "result" is supposed to mean the value of result after the preceding effects clause. This is in fact what all implementations I checked do (and which is probably it's intend, because it matches the specification of std::copy).

The problem is: I see nothing in the standard which grants that this interpretation is correct, specifically [lib.structure.specifications] or 17.3.1.3 [structure.specifications] resp. do not clarify which "look-up" rules apply for names found in the elements of the detailed specifications - Do they relate to the corresponding synopsis or to the effects clause (or possibly other elements)? Fortunately most detailed descriptions are unambigious in this regard, e.g. this problem does not apply for std::copy.

Proposed resolution:

Change the wording of the return clause to say (20.6.4.1 [uninitialized.copy]):

-2- Returns: The value of result after effects have taken place.


755. std::vector and std:string lack explicit shrink-to-fit operations

Section: 23.2.5.2 [vector.capacity], 21.3.4 [string.capacity] Status: New Submitter: Beman Dawes Date: 2007-10-31

View all other issues in [vector.capacity].

View all issues with New status.

Discussion:

A std::vector can be shrunk-to-fit via the swap idiom:

vector<int> v;
...
v.swap(vector<int>(v));  // shrink to fit

or:

vector<int>(v).swap(v);  // shrink to fit

or:

swap(v, vector<int>(v));  // shrink to fit

A non-binding request for shrink-to-fit can be made to a std::string via:

string s;
...
s.reserve(0);

Neither of these is at all obvious to beginners, and even some experienced C++ programmers are not aware that shrink-to-fit is trivially available.

Lack of explicit functions to perform these commonly requested operations makes vector and string less usable for non-experts. Because the idioms are somewhat obscure, code readability is impaired. It is also unfortunate that two similar vector-like containers use different syntax for the same operation.

The proposed resolution addresses these concerns. The proposed function takes no arguments to keep the solution simple and focused.

Proposed resolution:

To Class template basic_string 21.3 [basic.string] synopsis, Class template vector 23.2.5 [vector] synopsis, and Class vector<bool> 23.2.6 [vector.bool] synopsis, add:

    
void shrink_to_fit();

To basic_string capacity 21.3.4 [string.capacity] and vector capacity 23.2.5.2 [vector.capacity], add:

void shrink_to_fit();
Remarks: shrink_to_fit is a non-binding request to reduce capacity() to size(). [Note: The request is non-binding to allow latitude for implementation-specific optimizations. -- end note]

756. Container adaptors push

Section: 23.2.4 [container.adaptors] Status: New Submitter: Paolo Carlini Date: 2007-10-31

View all issues with New status.

Discussion:

After n2369 we have a single push_back overload in the sequence containers, of the "emplace" type. At variance with that, still in n2461, we have two separate overloads, the C++03 one + one taking an rvalue reference in the container adaptors. Therefore, simply from a consistency point of view, I was wondering whether the container adaptors should be aligned with the specifications of the sequence container themselves: thus have a single push along the lines:

template<typename... _Args>
void
push(_Args&&... __args)
  { c.push_back(std::forward<_Args>(__args)...); }

Proposed resolution:

Change 23.2.4.1.1 [queue.defn]:

void push(const value_type& x) { c.push_back(x); }
void push(value_type&& x) { c.push_back(std::move(x)); }
template<class... Args> void push(Args&&... args) { c.push_back(std::forward<Args>(args)...); }

Change 23.2.4.2 [priority.queue]:

void push(const value_type& x) { c.push_back(x); }
void push(value_type&& x) { c.push_back(std::move(x)); }
template<class... Args> void push(Args&&... args) { c.push_back(std::forward<Args>(args)...); }

Change 23.2.4.2.2 [priqueue.members]:

void push(const value_type& x);

Effects:

c.push_back(x);
push_heap(c.begin(), c.end(), comp);
template<class... Args> void push(value_type Args&&... x args);

Effects:

c.push_back(std::moveforward<Args>(x args)...);
push_heap(c.begin(), c.end(), comp);

Change 23.2.4.3.1 [stack.defn]:

void push(const value_type& x) { c.push_back(x); }
void push(value_type&& x) { c.push_back(std::move(x)); }
template<class... Args> void push(Args&&... args) { c.push_back(std::forward<Args>(args)...); }

757. Typo in the synopsis of vector

Section: 23.2.5 [vector] Status: New Submitter: Paolo Carlini Date: 2007-11-04

View other active issues in [vector].

View all other issues in [vector].

View all issues with New status.

Discussion:

In the synopsis 23.2.5 [vector], there is the signature:

void insert(const_iterator position, size_type n, T&& x);

instead of:

iterator insert(const_iterator position, T&& x);

23.2.5.4 [vector.modifiers] is fine.

Proposed resolution:

Change the synopsis in 23.2.5 [vector]:

iterator insert(const_iterator position, const T& x); 
iterator insert(const_iterator position, T&& x);
void     insert(const_iterator position, size_type n, const T& x); 
void     insert(const_iterator position, size_type n, T&& x);

758. shared_ptr and nullptr

Section: 20.6.6.2 [util.smartptr.shared] Status: New Submitter: Joe Gottman Date: 2007-10-31

View other active issues in [util.smartptr.shared].

View all other issues in [util.smartptr.shared].

View all issues with New status.

Discussion:

Consider the following program:

int main() {
   shared_ptr<int> p(nullptr); 
   return 0;
}

This program will fail to compile because shared_ptr uses the following template constructor to construct itself from pointers:

template <class Y> shared_ptr(Y *);

According to N2431, the conversion from nullptr_t to Y * is not deducible, so the above constructor will not be found. There are similar problems with the constructors that take a pointer and a deleter or a pointer, a deleter and an allocator, as well as the corresponding forms of reset(). Note that N2435 will solve this problem for constructing from just nullptr, but not for constructors that use deleters or allocators or for reset().

In the case of the functions that take deleters, there is the additional question of what argument should be passed to the deleter when it is eventually called. There are two reasonable possibilities: nullptr or static_cast<T *>(0), where T is the template argument of the shared_ptr. It is not immediately clear which of these is better. If D::operator() is a template function similar to shared_ptr's constructor, then d(static_cast<T*>(0)) will compile and d(nullptr) will not. On the other hand, if D::operator()() takes a parameter that is a pointer to some type other that T (for instance U* where U derives from T) then d(nullptr) will compile and d(static_cast<T *>(0)) may not.

Proposed resolution:

Add the following constructors to 20.6.6.2 [util.smartptr.shared]:

shared_ptr(nullptr_t);
template <class D> shared_ptr(nullptr_t, D d);
template <class D, class A> shared_ptr(nullptr_t, D d, A a);

Add the following methods to 20.6.6.2 [util.smartptr.shared]:

void reset(nullptr_t);
template <class D> void reset(nullptr_t, D d);
template <class D, class A> void reset(nullptr_t, D d, A a);

Add the following constructor definitions to 20.6.6.2.1 [util.smartptr.shared.const]:

 explicit shared_ptr(nullptr_t);

Effects: Constructs an empty shared_ptr object.

Postconditions: use_count() == 0 && get() == 0.

Throws: nothing.

template <class D> shared_ptr(nullptr_t, D d);
template <class D, class A> shared_ptr<nullptr_t, D d, A a);

Requires: D shall be CopyConstructible. The copy constructor and destructor of D shall not throw exceptions. The expression d(static_cast<T *>(0)) shall be well-formed, shall have well defined behavior, and shall not throw exceptions. A shall be an allocator (20.1.2 [allocator.requirements]). The copy constructor and destructor of A shall not throw exceptions.

Effects: Constructs a shared_ptr object that owns a null pointer of type T * and deleter d. The second constructor shall use a copy of a to allocate memory for internal use.

Postconditions: use_count() == 1 and get() == 0.

Throws: bad_alloc, or an implementation-defined exception when a resource other than memory could not be obtained.

Exception safety: If an exception is thrown, d(static_cast<Y *>(nullptr)) is called.

Add the following method definitions to 20.6.6.2.4 [util.smartptr.shared.mod]:

void reset(nullptr_t);

Effects: Equivalent to shared_ptr(nullptr).swap(*this).

template <class D> void reset(nullptr_t, const D d)

Effects: Equivalent to shared_ptr(nullptr, d).swap(*this).

template <class D, class A> void reset(nullptr_t, D d, A a);

Effects: Equivalent to shared_ptr(nullptr, d, a).swap(*this).


759. A reference is not an object

Section: 23.1 [container.requirements] Status: New Submitter: Jens Maurer Date: 2007-11-06

View other active issues in [container.requirements].

View all other issues in [container.requirements].

View all issues with New status.

Discussion:

23.1 [container.requirements] says:

-12- Objects passed to member functions of a container as rvalue references shall not be elements of that container. No diagnostic required.

A reference is not an object, but this sentence appears to claim so.

What is probably meant here:

An object bound to an rvalue reference parameter of a member function of a container shall not be an element of that container; no diagnostic required.

Proposed resolution:

Change 23.1 [container.requirements]:

-12- Objects passed to member functions of a container as rvalue references shall not be elements An object bound to an rvalue reference parameter of a member function of a container shall not be an element of that container.; Nno diagnostic required.

760. The emplace issue

Section: 23.1 [container.requirements] Status: New Submitter: Paolo Carlini Date: 2007-11-11

View other active issues in [container.requirements].

View all other issues in [container.requirements].

View all issues with New status.

Discussion:

In an emplace member function the function parameter pack may be bound to a priori unlimited number of objects: some or all of them can be elements of the container itself. Apparently, in order to conform to the blanket statement 23.1 [container.requirements]/11, the implementation must check all of them for that possibility. A possible solution can involve extending the exception in 23.1 [container.requirements]/12 also to the emplace member. As a side note, the push_back and push_front member functions are luckily not affected by this problem, can be efficiently implemented anyway

Proposed resolution:

Add after 23.1 [container.requirements]/12:

-12- Objects passed to member functions of a container as rvalue references shall not be elements of that container. No diagnostic required.

-13- Objects bound to the function parameter pack of the emplace member function shall not be elements or sub-objects of elements of the container. No diagnostic required.


761. unordered_map needs an at() member function

Section: 23.4.1.2 [unord.map.elem] Status: New Submitter: Joe Gottman Date: 2007-11-15

View all issues with New status.

Discussion:

The new member function at() was recently added to std::map(). It acts like operator[](), except it throws an exception when the input key is not found. It is useful when the map is const, the value_type of the key doesn't have a default constructor, it is an error if the key is not found, or the user wants to avoid accidentally adding an element to the map. For exactly these same reasons, at() would be equally useful in std::unordered_map.

Proposed resolution:

Add the following functions to the definition of unordered_map under "lookup" (23.4.1 [unord.map]):

mapped_type& at(const key_type& k);
const mapped_type &at(const key_type &k) const;

Add the following definitions to 23.4.1.2 [unord.map.elem]:

mapped_type& at(const key_type& k);
const mapped_type &at(const key_type &k) const;

Returns: A reference to x.second, where x is the (unique) element whose key is equivalent to k.

Throws: An exception object of type out_of_range if no such element is present.


762. std::unique_ptr requires complete type?

Section: 20.6.5 [unique.ptr] Status: New Submitter: Daniel Krügler Date: 2007-11-30

View other active issues in [unique.ptr].

View all other issues in [unique.ptr].

View all issues with New status.

Discussion:

In contrast to the proposed std::shared_ptr, std::unique_ptr does currently not support incomplete types, because it gives no explicit grant - thus instantiating unique_ptr with an incomplete pointee type T automatically belongs to undefined behaviour according to 17.4.3.6 [res.on.functions]/2, last bullet. This is an unnecessary restriction and prevents many well-established patterns - like the bridge pattern - for std::unique_ptr.

Proposed resolution:

In 20.6.5 [unique.ptr]/2 add as the last sentence to the existing para:

The unique_ptr provides a semantics of strict ownership. A unique_ptr owns the object it holds a pointer to. A unique_ptr is not CopyConstructible, nor CopyAssignable, however it is MoveConstructible and MoveAssignable. The template parameter T of unique_ptr may be an incomplete type. [ Note: The uses of unique_ptr include providing exception safety for dynamically allcoated memory, passing ownership of dynamically allocated memory to a function, and returning dynamically allocated memory from a function. -- end note ]

Change the 2nd sentence of 20.6.5.2 [unique.ptr.single]/1

The default type for the template parameter D is default_delete. A client-supplied template argument D must shall be a function pointer or functor for which, given a value d of type D and a pointer ptr of type T*, the expression d(ptr) is valid well-formed and has well-defined behavior when its evaluation is part of the operational semantic, and has the effect of deallocating the pointer as appropriate for that deleter. D may also be an lvalue-reference to a deleter.

Add the following sentence at the end of 20.6.5.2 [unique.ptr.single]/2:

If the deleter D maintains state, it is intended that this state stay with the associated pointer as ownership is transferred from unique_ptr to unique_ptr. The deleter state need never be copied, only moved or swapped as pointer ownership is moved around. That is, the deleter need only be MoveConstructible, MoveAssignable, and Swappable, and need not be CopyConstructible (unless copied into the unique_ptr) nor CopyAssignable. If any of these operations is part of the operational semantic, the corresponding expression shall be well-formed and shall have well-defined behavior.

763. Renaming emplace() overloads

Section: 23.1.2 [associative.reqmts] Status: New Submitter: Sylvain Pion Date: 2007-12-04

View all other issues in [associative.reqmts].

View all issues with New status.

Discussion:

The associative containers provide 2 overloads of emplace():

template <class... Args> pair<iterator, bool> emplace(Args&&... args);
template <class... Args> iterator emplace(const_iterator position, Args&&... args);

This is a problem if you mean the first overload while passing a const_iterator as first argument.

Proposed resolution:

Rename one of the two overloads. For example to emplace_here, hint_emplace...


764. equal_range on unordered containers should return a pair of local_iterators

Section: 23.1.3 [unord.req] Status: New Submitter: Joe Gottman Date: 2007-11-29

View other active issues in [unord.req].

View all other issues in [unord.req].

View all issues with New status.

Discussion:

A major attribute of the unordered containers is that iterating though them inside a bucket is very fast while iterating between buckets can be much slower. If an unordered container has a low load factor, iterating between the last iterator in one bucket and the next iterator, which is in another bucket, is O(bucket_count()) which may be much larger than O(size()).

If b is an non-const unordered container of type B and k is an object of it's key_type, then b.equal_range(k) currently returns pair<B::iterator, B::iterator>. Consider the following code:

B::iterator lb, ub;
tie(lb, ub) = b.equal_range(k);
for (B::iterator it = lb; it != ub; ++it) {
        // Do something with *it
}

If b.equal_range(k) returns a non-empty range (i.e. b contains at least on element whose key is equivalent to k), then every iterator in the half-open range [lb, ub) will be in the same bucket, but ub will likely either be in a different bucket or be equal to b.end(). In either case, iterating between ub - 1 and ub could take a much longer time than iterating through the rest of the range.

If instead of returning pair<iterator, iterator>, equal_range were to return pair<local_iterator, local_iterator>, then ub (which, like lb, would now be a local_iterator) could be guaranteed to always be in the same bucket as lb. In the cases where currently ub is equal to b.end() or is in a different bucket, ub would be equal to b.end(b.bucket(key)). This would make iterating between lb and ub much faster, as every iteration would be constant time.

Proposed resolution:

Change the entry for equal_range in Table 93 (23.1.3 [unord.req]) as follows:

expression return type assertion/note pre/post-condition complexity
b.equal_range(k) pair<local_iterator,local_iterator>; pair<const_local_iterator,const_local_iterator> for const b. Returns a range containing all elements with keys equivalent to k. Returns make_pair(b.end(b.bucket(key)),b.end(b.bucket(key))) if no such elements exist. Average case Θ(b.count(k)). Worst case Θ(b.size()).