Re: [PATCH] add attribute none to pthread_setspecific (BZ #27714)

[Martin Sebor <msebor@gmail.com>]

On 4/23/21 2:19 PM, Paul Eggert wrote:
> On 4/23/21 8:24 AM, Martin Sebor wrote:
>>
>> I think we considered variadic macros
>> when we first introduced the attribute but rejected it for some
>> reason that I'm not sure I remember.  Maybe because they're a C99
>> feature and Glibc supports older compilers?
> 
> That shouldn't be an issue here, since the suggested change uses 
> variadic macros for GCC 10 only.
> 
> Unless people were worried about running something like 'gcc -ansi' or 
> 'gcc -std=c89? To head that off at the pass, we can do the GCC 10 stuff 
> only if !__STRICT_ANSI__. Also, while we're at it we should be 
> C99-compatible in the variadic part (i.e., at least one named argument). 
> Something like the attached (untested) patch, say.

I'm open to your suggestion to use the variadic macros if no one
objects to it.

GCC doesn't let language conformance modes affect unrelated warnings
(like -Wuninitialized) and I am not in favor of introducing such
a distinction in Glibc.  We routinely get bug reports about false
negatives that boil down to the sensitivity of these warnings
(especially -Wuninitialized) to optimization settings.  I wouldn't
want to compound the problem.

Martin