[vlc-devel] commit: CPU: get rid of signal(SIGILL) - not thread-safe (in LibVLC) ( Rémi Denis-Courmont )

[xxcv]

Rémi Denis-Courmont wrote:
> Le vendredi 21 août 2009 23:15:09 Pierre d'Herbemont, vous avez écrit :
>   
>> On Aug 21, 2009, at 9:31 PM, git version control wrote:
>>     
>>> +# warning FIXME!
>>> +# define fork() (errno = ENOSYS, -1)
>>> +    (void)pid;
>>> +    (void)psz_capability;
>>> +    return true;
>>> +#endif
>>>       
>> I used to have no opinion on this, but I think that VLC_UNUSED(arg) is
>> a lot better than (void)arg. Mostly because this is more explicit, and
>> then because it could be instrumented.
>>     
>
> I find it VLC_UGLY and pointless.
>
>   
>> As a side node __attribute__((unused)) is probably better since it
>> could allow certain optimization from the compiler. We could switch to
>> it.
>>     
>
> The compiler is quite stupid if it cannot optimize (void)foo; to nothing.
>
>   
>> Hence VLC_UNUSED makes it a bit easier if we wanted, for instance, to
>> switch to __attribute__((unused)), because grep wouldn't return false
>> positive.
>>     
>
> I disagree. (void)foo; is more legible if only because it is standard 
> practice. I don't like to reinvent C.
>
>   
[0xa51040] main libvlc debug: CPU has capabilities 486 586 MMX MMXEXT 
SSE SSE2 FPU
[0xa51040] main libvlc debug: looking for memcpy module: 3 candidates
[0xa51040] main libvlc debug: using memcpy module "memcpymmxext"
My CPU supports 3DNow! (+) instruction and it is not been detected by vlc.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/vlc-devel/attachments/20090823/224c42fa/attachment.html>