[vlc-devel] commit: memcpy: Flag an unused param in memcpyaltivec. (Pierre d'Herbemont )

[Pierre d'Herbemont]

On Aug 21, 2009, at 7:56 AM, Rémi Denis-Courmont wrote:

> Le vendredi 21 août 2009 00:06:30 git version control, vous avez  
> écrit :
>> vlc | branch: master | Pierre d'Herbemont <pdherbemont at free.fr> |  
>> Thu Aug
>> 20 23:05:06 2009 +0200| [3d82d569c4030a56bb25baffe8ebd0f303fc31c2] |
>> committer: Pierre d'Herbemont
>>
>> memcpy: Flag an unused param in memcpyaltivec.
>
> AGAIN this a bug. You CANNOT register a memcpy optimization without  
> using the
> object pointer. You are really being annoying with all that crap.

> PLEASE STOP THIS MADNESS!!!!


This is simply not true,. You don't need the object pointer to  
register a memcpy.

If your point is to say that vlc_fastmem_register() is broken, I don't  
think that this warning is relevant.
- It doesn't indicate what you are saying - apart probably to you.
- The problem you are (not) explaining doesn't lay in the module but  
in the API. If there is to be a warning it should be here.
- You haven't been seeing any issue here for years until now, else you  
would have fixed it, I assume.
- No one has either.
- We want to reduce the number of warnings to catch important one.

This warning should to be silenced if we want to catch real warnings,  
and prevent future one. That's the point of those commits.

I don't think current warning approach has been proven to be useful at  
all. A hudge number of warnings are benign, indicating nothing and  
spamming. Look at the number of real issue hidden behind warnings.  
Look at the number of submission that could be fixed just by looking  
at the warnings. I think others share this point of view here.

This will only raise a bit the code quality of VLC, and ease  
developers in their work.

Your tone is not appropriate for this list, please keep it technical  
and less dramatic. (nth time, come on!)

Pierre.