karl@haddock.ima.isc.com (Karl Heuer) (06/07/90)
In article <371@necssd.NEC.COM> harrison@necssd.NEC.COM (Mark Harrison) writes: > #if VMS > extern char * x$something; /* however it's done */ > #endif As noted in the previous article, the quote is shielded by the comment. That `$' is a bit trickier, though. It lexes into a single pp-token, which is excluded along with its neighbors on a non-VMS system (assuming the symbol VMS has the obvious meaning); I presume that if there's ever an ANSI compiler on VMS, it will issue a trivial diagnostic, and, since the presence of the invalid token throws us into the realm of Undefined Behavior, the compiler will conveniently choose to define this situation to mean `glue these together into a single token'. Or more likely, such a compiler will support something like `#pragma enable extended_identifiers', since that would allow the expected behavior for macros, too. (The Standard requires that the strictly conforming program #include <stdio.h> #define foo$bar /* define foo to be dollar-bar */ #ifdef foo int main(void) { printf("yes\n"); return 0; } #else int main(void) { printf("no\n"); return 0; } #endif must print "yes"; normal VMS behavior would be to treat `foo$bar' as a single token and print "no".) Personally, I think the treatment of unexpected characters should have been implementation-defined. But hey, DEC and Apollo both had representatives on the Committee, and if they're satisfied, who am I to complain? Karl W. Z. Heuer (karl@ima.ima.isc.com or harvard!ima!karl), The Walking Lint
walter@hpcllca.HP.COM (Walter Murray) (06/09/90)
Karl Heuer writes: > (The Standard requires that the strictly conforming program > #include <stdio.h> > #define foo$bar /* define foo to be dollar-bar */ > #ifdef foo > int main(void) { printf("yes\n"); return 0; } > #else > int main(void) { printf("no\n"); return 0; } > #endif > must print "yes"; normal VMS behavior would be to treat `foo$bar' as a single > token and print "no".) I don't see how this can be a strictly conforming program. The source character set is implementation-defined (page 11, lines 3-4), and there is nothing in the standard that requires $ to be a member. (See also page 11, lines 29-32.) In fact, I would have trouble convincing myself that either of the following is a strictly conforming program: /* Program 1 */ int main () { /* $ */ } /* Program 2 */ int main () { '$'; } Any thoughts? Looking at this from a different perspective, must a conforming implementation on an 8-bit-byte machine be prepared to accept all 256 possible bytes in a string literal? Must it map them to 256 distinct values? I think not. The mapping of source file characters to the source character set in translation phase 1 is not defined by the standard. Walter Murray -------------
gwyn@smoke.BRL.MIL (Doug Gwyn) (06/12/90)
In article <16490021@hpcllca.HP.COM> walter@hpcllca.HP.COM (Walter Murray) writes:
-I don't see how this can be a strictly conforming program. The source
-character set is implementation-defined (page 11, lines 3-4), and there
-is nothing in the standard that requires $ to be a member. (See also
-page 11, lines 29-32.)
Quite right; a strictly conforming program cannot use '$' in identifiers.
-In fact, I would have trouble convincing myself that either of the
-following is a strictly conforming program:
- /* Program 1 */ int main () { /* $ */ }
- /* Program 2 */ int main () { '$'; }
I agree. Indeed, I stopped using ` characters in comments when I realized
that they could impair portability.
-Looking at this from a different perspective, must a conforming
-implementation on an 8-bit-byte machine be prepared to accept all
-256 possible bytes in a string literal? Must it map them to 256
-distinct values? I think not.
I agree again. Further note that multibyte considerations mean that
even the sequence '*' '/' may not be seen in a comment, if through
use of funny characters the parser has entered some shift state.karl@haddock.ima.isc.com (Karl Heuer) (06/14/90)
In article <13082@smoke.BRL.MIL> gwyn@smoke.BRL.MIL (Doug Gwyn) writes: >In article <16490021@hpcllca.HP.COM> walter@hpcllca.HP.COM (Walter Murray) writes: >>I don't see how this can be a strictly conforming program. [`$' need not be >>a member of the source character set.] I'll reluctantly concede that point. (But see below.) >Quite right; a strictly conforming program cannot use '$' in identifiers. Not relevant; the program in question wasn't using it in an identifier. That was, in fact, the point: the Standard requires that any implementation in which `$' is in the source character set (in practice, any implementation in the ASCII-or-superset world) *must* accept the program and generate a result contrary to what is produced by the traditional DEC or Apollo compilers. >I stopped using ` characters in comments when I realized that they could >impair portability. I figure it's a Quality of Implementation issue. An implementation that disallows "`" in a comment probably won't survive long in a free market. (And if some of the characters in my source file cause a change in shift state when interpreted in the implementation's character set, then the vendor had better supply a translator that maps my character set into his.) Karl W. Z. Heuer (karl@ima.ima.isc.com or harvard!ima!karl), The Walking Lint