Re: [PATCH] Check for ISO C compilers should also allow C++

[Fangrui Song <maskray@google.com>]

On 2022-05-13, Jonathan Wakely via Libc-alpha wrote:
>On Fri, 13 May 2022 at 18:28, Joseph Myers wrote:
>>
>> On Fri, 13 May 2022, Jonathan Wakely via Libc-alpha wrote:
>>
>> > Tested x86_64-linux.
>> >
>> > There is a suggestion[1] to make clang++ stop defining __STDC__ but it's
>> > blocked by this. The clang++ change will still be blocked while old
>> > Glibc versions are still in use, but we can start the clock ticking by
>> > fixing Glibc now.
>> >
>> > [1] https://discourse.llvm.org/t/rfc-stop-defining-the-stdc-and-related-macros-in-c-mode/62468/1
>> >
>> > >From a quick grep, I don't think any other places in Glibc need a
>> > similar change. Florian pointed out that gnulib overrides
>> > <sys/cdefs.h> and so should make the same change in its version.
>>
>> The change seems sane.  While we don't document what extensions (beyond
>> C90 / C++98) are needed to use glibc's headers (e.g. long long, flexible
>> array members, anonymous structs / unions; in at least some cases,
>> alignment attributes and asm redirection of functions), I see no reason
>> defining __STDC__ in C++ should be such a required extension.
>
>Right. Its meaning in ISO C++ is entirely implementation-defined, and
>G++ doesn't document any particular meaning for it, and clang++
>doesn't either AFAIK, and follows a slightly different set of rules
>for when it's defined (see the link above). So in C++ the only thing
>that __STDC__ being defined tells you is that __STDC__ is defined :-)
>That's not a very useful thing to test for in isolation.
>

I agree with the spirit of the clang RFC but I am sad to know that glibc sys/cdefs.h blocks it.
Thanks for sending the patch!

It seems that the subject may be clearer if starting with "<sys/cdefs.h>: ..."

In addition, It will be clearer if the token `__STDC__` is mentioned on the subject line.

Reviewed-by: Fangrui Song <maskray@google.com>