Re: bug: "gcc -std=gnu99" passes AC_PROG_CC_C11

[Adrian Bunk]

On Tue, Sep 25, 2012 at 04:04:27PM -0700, Paul Eggert wrote:
> On 09/25/2012 02:22 PM, Adrian Bunk wrote:
> > #if !defined(__STDC_VERSION__) || (__STDC_VERSION__ < 201112L)
> 
> If that's what the user wants, they have that already;
> they don't need Autoconf for it.

We are talking about the "put the compiler into the highest mode 
possible" logic in AC_PROG_CC.

If a future gcc version
- defaults to gnu99 mode (defaulting to a higher C mode is
  documented to happen in gcc in the future) and
- supports C11 features in gnu11 mode it doesn't support in
  gnu99 mode (likely)
then this bug will cause problems.

Testing for features not documented to be tested for ac_cv_prog_cc_c11
is a duty of the user, but the user shouldn't have to fiddle with the
compiler mode after AC_PROG_CC.

> Surely it's more
> useful if Autoconf tests for actual features rather
> than testing __STDC_VERSION__.

I am not suggesting to remove the existing parts of the C11 check.

>  Historically, defining
> __STDC_VERSION__ has often meant "I support the
> standard and nothing more -- I'm being pedantic"
> which is not typically what's really wanted.

Historically, compilers were not supporting C11.

Currently gcc is the only compiler where autoconf knows how to put it 
into C11 mode, and __STDC_VERSION__ is defined in gnu11 mode.

If __STDC_VERSION__ becomes a problem when C11 options for other 
compilers get added, then the __STDC_VERSION__ test could be put inside 
an "#ifdef __GNUC__".

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed