Submitter: Ivan A. Kosarev (Unicals Group, RU)
Submission Date: 2005-04-19
Reference Document: N/A
Subject: Tokenization obscurities
126.96.36.199 #2 ("Translation phases") says:
[#2] ...2. Each instance of a backslash character (\) immediately followed by a new-line character is deleted, splicing physical source lines to form logical source lines. Only the last backslash on any physical source line shall be eligible for being part of such a splice. A source file that is not empty shall end in a new-line character, which shall not be immediately preceded by a backslash character before any such splicing takes place.
3. The source file is decomposed into preprocessing tokens6) and sequences of white-space characters (including comments). A source file shall not end in a partial preprocessing token or in a partial comment. Each comment is replaced by one space character. New-line characters are retained. Whether each nonempty sequence of white-space characters other than new-line is retained or replaced by one space character is implementation-defined.
Assuming there is a non-empty source file legally ending with a new-line character, what are examples of such partial preprocessing tokens that could end the file? And, generally, what the partial preprocessing tokens are?
188.8.131.52 ("Character constants") says:
[#3] The single-quote ', the double-quote ", the question-mark ?, the backslash \, and arbitrary integer values are representable according to the following table of escape sequences...
[#5] The octal digits that follow the backslash in an octal escape sequence are taken to be part of the construction of a single character for an integer character constant or of a single wide character for a wide character constant...
[#6] The hexadecimal digits that follow the backslash and the letter x in a hexadecimal escape sequence are taken to be part of the construction of a single character for an integer character constant or of a single wide character for a wide character constant...
[#8] In addition, characters not in the basic character set are representable by universal character names and certain nongraphic characters are representable by escape sequences consisting of the backslash \ followed by a lowercase letter: \a, \b, \f, \n, \r, \t, and \v.64)
64) The semantics of these characters were discussed in 5.2.2. If any other character follows a backslash, the result is not a token and a diagnostic is required. See ''future language directions'' (6.11.4).
6.4 #3 ("Lexical elements") says:
[#3] ...The categories of preprocessing tokens are: header names, identifiers, preprocessing numbers, character constants, string literals, punctuators, and single non-white-space characters that do not lexically match the other preprocessing token categories.58) If a ' or a " character matches the last category, the behavior is undefined.
What in the formal content of the standard says that if any other character follows a backslash, there should be a diagnostic? Does such a case causes undefined behaviour? Furthermore, if a character sequence that is coming just after a double quote " (that is not terminating a string literal) begins with, say, \l and the result of the tokenization is not a token, then what the result is?
6.10.9 #2 ("Pragma operator") gives the following example:
[#2] EXAMPLE A directive of the form:#pragma listing on "..\listing.dir"
can also be expressed as:_Pragma ( "listing on \"..\\listing.dir\"" )
The previous summary says that if there an unknown escape sequence encountered during tokenization of a character constant or string literal, then the result is not token. The question is whether the example above is well-defined.
Suggested Technical Corrigendum
Meyers & Plum Proposed Committee Response
Answers to Summary #1
"Partial preprocessing token" is not itself a technical term; it is merely the English Language word "partial" modifying the technical term "preprocessing token". A preprocessing token is defined by the grammar non-terminal preprocessing-token in Subclause 6.4. A partial preprocessing token is therefore just part of a preprocessing token that is not the entire preprocessing token.
The statement that "source files shall not end in a partial preprocessing token or in a partial comment" has two implications. First, a preprocessing token may not begin in one file and end in another file. Second, the last preprocessing token in a source file must be well-formed and complete. For example, the last token may not be a string literal missing the close quote.
Answers to Summary #2
Subclause 184.108.40.206 requires a diagnostic to be produced if there is a violation of any syntax rule or constraint. The syntax for character constants (Subclause 220.127.116.11) and string literals (Subclause 6.4.5) both require that any escape sequence be well formed according the nonterminal escape-sequence (Subclause 18.104.22.168). Thus a diagnostic is required if a character constant or string literal contains a \ not followed by the remainder of a valid escape sequence.
Note that a preprocessor token may be a header-name in addition to a string-literal. Although a header-name has the appearance of a string literal, it is parsed by a different grammar (Subclause 6.4.7). The syntax for header-name is not violated if a \ is not followed by the remainder of a valid escape sequence. Thus, no diagnostic is required. However, if a header name contains a \, the behavior is undefined (Subclause 6.4.7, paragraph 3).
Since the behavior is undefined, the implementation is free to do anything it wishes. Some possible examples:
The last alternative is useful if your operating system normally uses \ in the names of files.
Answers to Summary #3
Before your question can be answered, it is necessary to understand the precise steps taken in translating a _Pragma operator as controlled by the phases of translation (Subclause 22.214.171.124).
The _Pragma operator (Subclause 6.10.9) has an argument that is a string literal. If any escape sequence in that string literal is not grammatically well-formed, a diagnostic is required. In the example, the string literal is syntactically correct, so no diagnostic is required so far.
The _Pragma operator is executed during translation phase 4. Note that escape sequences in a string are not replaced by the characters that they represent until translation phase 5. However, the _Pragma operator itself does limited processing to handle \\ and \" escape sequences (Subclause 6.10.9 paragraph 1), and so when the value of the string literal is retokenized to produce preprocessor tokens, the result for the example matches the #pragma directive given.
The final answer to your question rests with whether the preprocessor tokens in the #pragma directive require a diagnostic. Subclause 6.10.6 states that a #pragma directive ends in just a list of preprocessor tokens without placing any requirements on which preprocessor tokens.
Your question then becomes whether the preprocessor token that looks like a string in the #pragma in the example is a string-literal or a header-name. If it is a string-literal, a diagnostic is required. If it is a header-name, no diagnostic is required, but there is undefined behavior. The choice of which preprocessor tokens are allowed in a #pragma directive is implementation defined.
Subclause 6.10.9 paragraph 1 states that a #pragma causes the implementation to behave in an implementation-defined manner including the possibility that the implementation might fail or otherwise behave in a non-conforming manner. The intent of the committee was that implementations could recognize header-name preprocessor tokens in #pragma directives, if the implementation chooses, but this seems to be contradicted by the two places requiring Technical Corrigenda below.
Suggested Technical Corrigendum
Change Subclause 6.4.7, paragraph 3, last sentence to:
Header name preprocessing tokens are recognized only within #include preprocessing directives or in implementation-defined locations within #pragma directives.
Change Subclause 6.4, paragraph 4, last sentence to:There is one exception to this rule: Header name preprocessing tokens are recognized only within #include preprocessing directives or in implementation-defined locations within #pragma directives. In such contexts, a sequence of characters that could be either a header name or a string literal is recognized as the former.
Previous Defect Report < - > Next Defect Report