Re: [RFC PATCH 1/1] io: Add FORTIFY_SOURCE check for fcntl arguments

[Florian Weimer <fweimer@redhat.com>]

* Sergey Bugaev:

> On Wed, May 24, 2023 at 3:18 PM Florian Weimer <fweimer@redhat.com> wrote:
>> We also generally do not prevent the use of compiler extensions in
>> strict feature modes.  I think this should use a compiler version check
>> and maybe a fallback check against C17 compiler support if we do not
>> need any extensions (assuming that all C17 compilers implement _Generic,
>> so that an exception like the one you mentioned for GCC does not apply
>> vis-a-vis C17).
>
> Do I understand this correctly? -- you want the check to be
>
> #if  (__STDC_VERSION__ >= 201710L) || (__GNUC_PREREQ (4, 9) &&
> __STDC_VERSION__ >= 201112L)
>
> right?

More like this:

#if  __STDC_VERSION__ >= 201710L) || __GNUC_PREREQ (4, 9)

If you have GCC 4.9, __STDC_VERSION__ does not matter if you use
__extension__ in front of _Generic (untested).  Clang would benefit from
another version conditional.

But we may not need _Generic after all.

> Why even care for compilers other than modern GCC (and Clang
> pretending to be GCC), considering the fortifications use very
> GCC-specific intrinsics (__builtin_*, attributes)?

Some of us want to reproduce things with older compilers, for bisecting
regressions for example.  For that, the headers better remain
compatible.  And GCC 4.8 is still very widely used (although not with
current GCC headers, admittedly).

Thanks,
Florian