On 24.11.2014 00:52, Michael Niedermayer wrote:
i think calling decode_end() is better than duplicating what it does
if it works
I have no opinion about that, but changed. I had to move decode_end above.
I could add prototype but from these 2 I prefer move it as it is not long.
Perfectly decode end callback could be called in avcodec_open2 on
fail but I'm not sure every codec under any circumstances is secure
to call it twice.
maybe a libavcodec private flag could be added to AVCodec.capabilities
to indicate that calling "end" is safe/needed for open failures
and then call it in case of error & that flag being set
Maybe, but not sure it is worth the effort. Many codec seems to do that.
The leaking one doesn't :)
It is not to hard call one function in implementation and it is easier
to read for others.
I'm trying to write some tests that allocs/deallocs codecs with "random"
options to find leaks/crashes and so far only few codecs leaked. Maybe
after this patch is created it will be safer to introduce that auto
deinit mechanism.
You didn't comment patch itself.
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
