Re: G_STMT_START / G_STMT_END on Sun

[Tim Janik <timj imendio com>]

On Fri, 11 Nov 2005, ANDREW PAPROCKI, BLOOMBERG/ 731 LEXIN wrote:

> Can anyone tell me what Sun compiler required that G_STMT_START/G_STMT_END be
> written as follows:
>
> <inside gmacros.h>
> #    if (defined (sun) || defined (__sun__))
> #      define G_STMT_START      if (1)
> #      define G_STMT_END        else (void)0
> #    else
>
> This is causing me headaches now and no longer seems to be any kind of
> requirement when compiling with even semi-recent Sun compilers:
>
> $ cat a.c
> #include <stdio.h>
> int main(void)
> {
>    do { printf("foo\n"); } while(0);
> }
> $ /opt/SUNWspro61/bin/cc -o a a.c
> $ /opt/SUNWspro8/bin/cc -o a a.c
> $
>
> Also, the macros are written in such a way that if you are using GCC in strict
> ANSI mode on a Sun machine, you will get the silly "if (1) else (void) 0" macros
> instead of the normal "do while(0)".
>
> The end result of all of this is that the following code generates warnings on
> recent GCC versions on Solaris:
>
> if (1)
>    g_assert_not_reached();
>
> $ /opt/gcc-3.4.3/bin/gcc <options here> -ansi test.c
> warning: suggest explicit braces to avoid ambiguous `else'
>
> If someone knows why those Sun checks existed in the first place, I'll post a
> patch to fix these macros to prevent this issue on Sun machines. The comment
> indicates they came from Perl code, but does not mention any version numbers.
> Next step, dig out the Perl src to see if they have an idea :)

the first mentioning of that is in a 1997 email from owen, i have
appended it. so if you're looking for a particular perl version, it's
one from before that date ;)

> Thanks,
>
> Andrew Paprocki
> Bloomberg LP
>
>
>
> _______________________________________________
> gtk-devel-list mailing list
> gtk-devel-list gnome org
> http://mail.gnome.org/mailman/listinfo/gtk-devel-list

---
ciaoTJ



> From owt1 cornell edu Thu Dec  4 00:27:11 1997
Date: Wed, 03 Dec 1997 09:25:56 -0500
From: Owen Taylor <owt1 cornell edu>
Reply-To: gtk-list redhat com
To: gtk-list redhat com
Subject: [gtk-list] Re: g_assert() and consorts into {}?
Resent-Date: 3 Dec 1997 16:30:02 -0000
Resent-From: gtk-list redhat com
Resent-cc: recipient list not shown: ;


> i'd like to see all the macros around g_assert () (return_if_fail...),
> to be enclosed by {}.

[ ... ]

> any objections?

A good idea, but just {} is a bad idea - it can lead to warnings
(errors?) on some systems.

Here's how Perl does it:

/*
 * STMT_START { statements; } STMT_END;
 * can be used as a single statement, as in
 * if (x) STMT_START { ... } STMT_END; else ...
 *
 * Trying to select a version that gives no warnings...
 */
#if !(defined(STMT_START) && defined(STMT_END))
# if defined(__GNUC__) && !defined(__STRICT_ANSI__) && !defined(__cplusplus)
#   define STMT_START	(void)(	/* gcc supports ``({ STATEMENTS; })'' */
#   define STMT_END	)
# else
   /* Now which other defined()s do we need here ??? */
#  if (VOIDFLAGS) && (defined(sun) || defined(__sun__))
#   define STMT_START	if (1)
#   define STMT_END	else (void)0
#  else
#   define STMT_START	do
#   define STMT_END	while (0)
#  endif
# endif
#endif

We can probably just adopt this pretty much wholesale.
VOIDFLAGS is a #define which is true if 'void' is defined.
Since we don't care about pre-ANSI systems, we probably
can omit that. I'm not sure why SunOS (?) systems warn
about do { } while (0), but it doesn't cost much to
leave that in.

Regards,
                                        Owen

--
To unsubscribe: mail -s unsubscribe gtk-list-request redhat com < /dev/null