libc: Don't use a custom wrapper macro around __has_include.

This causes issues when using clang with -frewrite-includes to e.g., submit the
translation unit to a distributed compiler.

In my case, I was building Firefox using sccache.

See https://bugs.llvm.org/show_bug.cgi?id=43982 for a reduced test-case since I
initially thought this was a clang bug.

Apparently doing this is invalid C++ per:

 * http://eel.is/c++draft/cpp.cond#7.sentence-2

(See https://bugs.llvm.org/show_bug.cgi?id=37990)

Signed-off-by: Emilio Cobos Álvarez <emilio@crisal.io>
---
 misc/sys/cdefs.h                     |  8 --------
 sysdeps/unix/sysv/linux/bits/statx.h | 14 ++++++++------
 2 files changed, 8 insertions(+), 14 deletions(-)

* Emilio Cobos Álvarez:

> This causes issues when using clang with -frewrite-includes to e.g., submit the
> translation unit to a distributed compiler.

Thanks for the patch.  The substance of the patch is okay.  I would
prefer if you could repost it with the minor issues mentioned below
addressed, then I can push it on your behalf.

I've tried to push this into Gerrit, but we can't Cc: people who do not
have accounts in the tool.  Not your fault.  We can continue on the
mailing list.

> In my case, I was building Firefox using sccache.
>
> See https://bugs.llvm.org/show_bug.cgi?id=43982 for a reduced test-case since I
> initially thought this was a clang bug.
>
> Apparently doing this is invalid C++ per:
>
>  * http://eel.is/c++draft/cpp.cond#7.sentence-2

I think it's probably better to quote the actual language in the
standard.

> (See https://bugs.llvm.org/show_bug.cgi?id=37990)
>
> Signed-off-by: Emilio Cobos Álvarez <emilio@crisal.io>

We do not use Signed-off-by, but copyright assignments, but we have been
advised that this is not necessary for changes of this size.

> diff --git a/sysdeps/unix/sysv/linux/bits/statx.h b/sysdeps/unix/sysv/linux/bits/statx.h
> index ff3f2e8973..14d5992226 100644
> --- a/sysdeps/unix/sysv/linux/bits/statx.h
> +++ b/sysdeps/unix/sysv/linux/bits/statx.h
> @@ -26,12 +26,14 @@
>  
>  /* Use "" to work around incorrect macro expansion of the
>     __has_include argument (GCC PR 80005).  */
> -#if __glibc_has_include ("linux/stat.h")
> -# include "linux/stat.h"
> -# ifdef STATX_TYPE
> -#  define __statx_timestamp_defined 1
> -#  define __statx_defined 1
> -# endif
> +#ifdef __has_include
> +#  if __has_include("linux/stat.h")
> +#   include "linux/stat.h"
> +#   ifdef STATX_TYPE
> +#    define __statx_timestamp_defined 1
> +#    define __statx_defined 1
> +#   endif
> +#  endif
>  #endif

The indentation is slight off here, I think.  Each nested conditional
adds one space after the #.  Please also include a space after
__has_include.

Thanks,
Florian

On 11/21/19 2:00 PM, Florian Weimer wrote:
> * Emilio Cobos Álvarez:
> 
>> This causes issues when using clang with -frewrite-includes to e.g., submit the
>> translation unit to a distributed compiler.
> 
> Thanks for the patch.  The substance of the patch is okay.  I would
> prefer if you could repost it with the minor issues mentioned below
> addressed, then I can push it on your behalf.

I think I did this right, but let me know otherwise :)

> I've tried to push this into Gerrit, but we can't Cc: people who do not
> have accounts in the tool.  Not your fault.  We can continue on the
> mailing list.

Ah, cool. Are there docs for pushing to Gerrit? I saw the revision, and 
it seems I can edit the revision description, but not sure how to set up 
the instance for glibc :)

I am familiar with Gerrit itself so I'd be ok with following-up there if 
needed. Anyhow hopefully the second version of the patch looks good.
>> In my case, I was building Firefox using sccache.
>>
>> See https://bugs.llvm.org/show_bug.cgi?id=43982 for a reduced test-case since I
>> initially thought this was a clang bug.
>>
>> Apparently doing this is invalid C++ per:
>>
>>   * http://eel.is/c++draft/cpp.cond#7.sentence-2
> 
> I think it's probably better to quote the actual language in the
> standard.

Rephrased the commit message to be a little cleaner and quote the 
standard directly.

> 
>> (See https://bugs.llvm.org/show_bug.cgi?id=37990)
>>
>> Signed-off-by: Emilio Cobos Álvarez <emilio@crisal.io>
> 
> We do not use Signed-off-by, but copyright assignments, but we have been
> advised that this is not necessary for changes of this size.

Removed.

Thanks!

  -- Emilio

>> diff --git a/sysdeps/unix/sysv/linux/bits/statx.h b/sysdeps/unix/sysv/linux/bits/statx.h
>> index ff3f2e8973..14d5992226 100644
>> --- a/sysdeps/unix/sysv/linux/bits/statx.h
>> +++ b/sysdeps/unix/sysv/linux/bits/statx.h
>> @@ -26,12 +26,14 @@
>>   
>>   /* Use "" to work around incorrect macro expansion of the
>>      __has_include argument (GCC PR 80005).  */
>> -#if __glibc_has_include ("linux/stat.h")
>> -# include "linux/stat.h"
>> -# ifdef STATX_TYPE
>> -#  define __statx_timestamp_defined 1
>> -#  define __statx_defined 1
>> -# endif
>> +#ifdef __has_include
>> +#  if __has_include("linux/stat.h")
>> +#   include "linux/stat.h"
>> +#   ifdef STATX_TYPE
>> +#    define __statx_timestamp_defined 1
>> +#    define __statx_defined 1
>> +#   endif
>> +#  endif
>>   #endif
> 
> The indentation is slight off here, I think.  Each nested conditional
> adds one space after the #.  Please also include a space after
> __has_include.
> 
> Thanks,
> Florian
>