[libvirt] Compiler warnings

[Wen Congyang]

On 10/18/2011 10:02 AM, Wen Congyang wrote:
> At 10/18/2011 01:03 AM, Eric Blake Write:
>> On 10/17/2011 10:17 AM, Alex Jia wrote:
>>> Eric,
>>> It's latest libvirt upstream, current commit id is commit 0a71c79.
>>>
>>> # rpm -q glibc
>>> glibc-2.12-1.42.el6.x86_64
>>>
>>
>>> 430         if (strchr(toescape, *cur))
>>
>>>>>       CC     libvirt_util_la-buf.lo
>>>>> util/buf.c: In function 'virBufferEscape':
>>>>> util/buf.c:430: warning: logical '&&' with non-zero constant will
>>>>> always
>>>>> evaluate as true [-Wlogical-op]
>>
>> Something's screwy.  There's no&&  in that line, unless strchr() is a
>> macro based on your particular compilation flags.  I'm suspecting that
>> this is a case of -D_FORTIFY_SOURCE going awry.  I don't see how this
>> could possibly be a libvirt issue; more likely, it is an issue with the
>> choice of compiler flags you are using, coupled with an issue in the
>> system headers causing the preprocessor to expand strchr() into
>> something that tickles a spurious gcc warning.
>>
>
> Yes, strchr() is a macro, I use -E, and get the following:
>
>      while (*cur != 0) {
>          if ((__extension__ (__builtin_constant_p (*cur)&&  !__builtin_constant_p (toescape)&&  (*cur) == '\0' ? (char *) __rawmemchr (toescape, *cur) : __builtin_strchr (toescape, *cur))))
>              *out++ = '\\';
>          *out++ = *cur;
>          cur++;
>      }
>      *out = 0;

I reproduce this problem on RHEL6.

Hmm, I test it on Fedora15, and the problem does not happen.
I check the output with -E, and the macro strchr()'s expansion
is the same as its expansion on RHEL6.

I think it maybe gcc's problem.

Thanks
Wen Congyang

>
> Thanks
> Wen Congyang
>
> --
> libvir-list mailing list
> libvir-list at redhat.com
> https://www.redhat.com/mailman/listinfo/libvir-list
>