mojo: Use static_assert for MOJO_COMPILE_ASSERT in C++11 mode.

mojo: Use static_assert for MOJO_COMPILE_ASSERT in C++11 mode, like in base/.

BUG=321833
NOTRY=true

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=268098

[Nico, 2014-05-04 01:32:40]

First to stamp wins!

https://codereview.chromium.org/267963003/diff/20001/mojo/public/c/system/mac...
File mojo/public/c/system/macros.h (right):

https://codereview.chromium.org/267963003/diff/20001/mojo/public/c/system/mac...
mojo/public/c/system/macros.h:14: #define MOJO_ALLOW_UNUSED
__attribute__((unused))
(Note that base's ignore_result() is a better way to handle this most of the
time.)

[darin (slow to review), 2014-05-04 03:15:14]

https://codereview.chromium.org/267963003/diff/20001/mojo/public/c/system/mac...
File mojo/public/c/system/macros.h (right):

https://codereview.chromium.org/267963003/diff/20001/mojo/public/c/system/mac...
mojo/public/c/system/macros.h:31: #if __cplusplus >= 201103L
this header is intended to be seen by a C compiler as well. __cplusplus would
not be defined in that case.

[Nico, 2014-05-04 04:20:12]

https://codereview.chromium.org/267963003/diff/20001/mojo/public/c/system/mac...
File mojo/public/c/system/macros.h (right):

https://codereview.chromium.org/267963003/diff/20001/mojo/public/c/system/mac...
mojo/public/c/system/macros.h:31: #if __cplusplus >= 201103L
On 2014/05/04 03:15:15, darin wrote:
> this header is intended to be seen by a C compiler as well. __cplusplus would
> not be defined in that case.

I think that's fine; undefined symbols evaluate to 0 in comparisons as far as I
can tell:

thakis$ cat macrotest.c 
#if __cplusplus >= 201103L
#error c++11
#elif defined(__cplusplus)
#error c++
#else
#error c
#endif

thakis$ clang -c macrotest.c 
macrotest.c:6:2: error: c
#error c
 ^
1 error generated.


If you don't have that yet, you should add a target with a .c file that includes
your c-compatible headers, so that the trybots will notice if someone checks in
something that breaks c compat. (But I think this change should be fine as is.)


(more tests:


thakis$ cat macrotest.c 
#if asdfasfasfasdf >= 0
#error yes
#else
#error no
#endif
thakis$ clang -c macrotest.c 
macrotest.c:2:2: error: yes
#error yes
 ^
1 error generated.

and
thakis$ cat macrotest.c 
#if asdfasfasfasdf > 0
#error yes
#else
#error no
#endif
thakis$ clang -c macrotest.c 
macrotest.c:4:2: error: no
#error no
 ^
1 error generated.
)

[Nico, 2014-05-04 04:27:11]

(see also
http://stackoverflow.com/questions/5085392/what-is-the-value-of-an-undefined-...
)

On Sat, May 3, 2014 at 9:20 PM,  <thakis@chromium.org> wrote:
>
>
https://codereview.chromium.org/267963003/diff/20001/mojo/public/c/system/mac...
> File mojo/public/c/system/macros.h (right):
>
>
https://codereview.chromium.org/267963003/diff/20001/mojo/public/c/system/mac...
> mojo/public/c/system/macros.h:31: #if __cplusplus >= 201103L
> On 2014/05/04 03:15:15, darin wrote:
>>
>> this header is intended to be seen by a C compiler as well.
>
> __cplusplus would
>>
>> not be defined in that case.
>
>
> I think that's fine; undefined symbols evaluate to 0 in comparisons as
> far as I can tell:
>
> thakis$ cat macrotest.c
> #if __cplusplus >= 201103L
> #error c++11
> #elif defined(__cplusplus)
> #error c++
> #else
> #error c
> #endif
>
> thakis$ clang -c macrotest.c
> macrotest.c:6:2: error: c
> #error c
>  ^
> 1 error generated.
>
>
> If you don't have that yet, you should add a target with a .c file that
> includes your c-compatible headers, so that the trybots will notice if
> someone checks in something that breaks c compat. (But I think this
> change should be fine as is.)
>
>
> (more tests:
>
>
> thakis$ cat macrotest.c
> #if asdfasfasfasdf >= 0
> #error yes
> #else
> #error no
> #endif
> thakis$ clang -c macrotest.c
> macrotest.c:2:2: error: yes
> #error yes
>  ^
> 1 error generated.
>
> and
> thakis$ cat macrotest.c
> #if asdfasfasfasdf > 0
> #error yes
> #else
> #error no
> #endif
> thakis$ clang -c macrotest.c
> macrotest.c:4:2: error: no
> #error no
>  ^
> 1 error generated.
> )
>
> https://codereview.chromium.org/267963003/

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.

[darin (slow to review), 2014-05-04 04:50:35]

OK, thanks for explaining. I was sure I had encountered compilers complaining
about #if used with undefined macros, but perhaps that is only an issue with
ancient compilers or only when undefined macros are not combined with relational
operators :-/

Anyways, it looks like the try bots would have complained if this were a
problem. We have a .c test file that includes this header (and others).

LGTM

[Nico, 2014-05-04 04:52:57]

The CQ bit was checked by thakis@chromium.org

[commit-bot: I haz the power, 2014-05-04 04:53:01]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/thakis@chromium.org/267963003/20001

[commit-bot: I haz the power, 2014-05-04 05:21:34]

Change committed as 268098