[PATCH] D17444: PR26672: [MSVC] Clang does not recognize "static_assert" keyword in C mode

[Richard Smith via cfe-commits]

On Thu, Mar 24, 2016 at 9:23 AM, David Majnemer
<david.majnemer at gmail.com> wrote:
>
>
> On Thu, Mar 24, 2016 at 9:09 AM, Aaron Ballman <aaron at aaronballman.com>
> wrote:
>>
>> On Thu, Mar 24, 2016 at 11:49 AM, Reid Kleckner <rnk at google.com> wrote:
>> > On Thu, Mar 3, 2016 at 10:40 AM, Aaron Ballman <aaron at aaronballman.com>
>> > wrote:
>> >>
>> >> That was what I meant by "justification". I would say it has to be
>> >> reasonably compelling code (win32 headers, boost, some other major
>> >> library) as that's our usual bar for these sort of bug-for-bug
>> >> compatible things, as I understand it.
>> >
>> >
>> > I'd rather apply this patch now than wait for ffmpeg or someone to try
>> > to
>> > use static_assert and then have to hustle to get this into clang. Many
>> > many
>> > C projects have COMPILE_ASSERT macros that are just a small change away
>> > from
>> > relying on (_S|s)tatic_assert.
>>
>> I don't find "someone might rely on this bug" to be compelling. Put
>> another way, why do we lower the bar for this bug but not others given
>> that no large projects appear to need this behavior?
>
>
> Very few people are using clang-cl with C code.  I'm sure it would come up a
> lot more often if it were better tested.
>
>>
>>
>> >>
>> >> Agreed, we have a way forward if we need it. I mostly just want to
>> >> avoid the burden of supporting that because this is sufficiently weird
>> >> (being a non-conforming keyword).
>> >
>> >
>> > It's not conforming, but it's not that weird to define our own keywords.
>> > The
>> > C++ committee chose the keyword "static_assert" because it was unlikely
>> > to
>> > conflict with existing code. MSVC has made this a keyword in C mode and
>> > the
>> > world hasn't burned.
>>
>> Correct, we have a way around it, I am just not convinced that we
>> should put forth the effort of supporting another compiler's bug
>> without a compelling use-case.
>
>
> A compelling use case is that users who wish to use static_assert in a
> conforming C program can do so.
> Today, it is impossible to do this with clang-cl & MSVC's CRT.

If MSVC's assert.h is failing to provide the required

  #define static_assert _Static_assert

then perhaps we should fix that instead. I have very little sympathy
for people who use static_assert in C without including the header
that provides it (it's harmless for MSVC and makes their code
portable).