[FFmpeg-cvslog] r12329 - in trunk/tests: ffmpeg.regression.ref rotozoom.regression.ref seek.regression.ref

[Panagiotis Issaris]

M?ns Rullg?rd schreef:
> Panagiotis Issaris wrote:
>> Michael Niedermayer schreef:
>>> On Thu, Mar 06, 2008 at 10:39:01AM +0100, Panagiotis Issaris wrote:
>>>> michael schreef:
>>>>> Author: michael
>>>>> Date: Wed Mar  5 13:45:32 2008
>>>>> New Revision: 12329
>>>>>
>>>>> Log:
>>>>> The last flag fix, fixed mpeg4-qprd (there the flags ended up being trashed
>>>>> through lrintf(), that is gcc puted the 32bit int flags in a 32bit float
>>>>> which caused some to be lost ...).
>>>>> I wonder why FATE didnt pick this up ?
>>>> My website reports such failures, it showed clearly that 12324 broke the
>>>> regression tests:
>>>> http://issaris.org/~takis/projects/ffmpeg/list.php
>>> competition for fate :)
>> :)
>>
>> And in a few clicks I've noticed that commit 12300 also increased the
>> number of warnings and introduced this new warning:
>> imgresample.c:530: warning: assignment discards qualifiers from pointer
>> target type
>>
>> Should the const go away as in attached patch?
> 
> No, whatever is the target of the assignment should be made const.

Ah, OK.  Normally I am also inclined to add consts not remove them, but 
the I'd checked and noticed that AVCodecContext->av_class and 
SwsContext->av_class were not const either. I figured this was for a reason.

Afterwards I noticed that AVFormatContext->av_class was indeed const, 
so, I've send a patch to ffmpeg-devel making all of them const.

Thanks for reviewing!

With friendly regards,
Takis
--
vCard: http://issaris.org/pi.vcf
PGP key: http://issaris.org/pi.key