[FFmpeg-devel] [PATCH] alloc_picture needs to free buffer when it's not going to use it

[Daniel Kristjansson]

I'm resending this patch because my last message seems
to have been moderated and rejected due to my e-mailing
from the wrong e-mail account.

I've also added an 2nd patch which takes care of the
CHECKED_ALLOCZ case.

== original message below ==

If there is some problem with the buffer returned from
s->avctx->get_buffer(...) mpegvideo.c:alloc_picture(...)
returns an error code but never returns the picture with
s->avctx->release_buffer(...).

The attached patch fixes this for the case when the
linesizes are wrong, or the age type or data are empty.

We might want to also do this when CHECKED_ALLOCZ fails,
but I'll leave that to the experts.

The reason I noticed this was because MythTV locks a frame
when get_buffer is called and releases it when release_buffer
is called. When a user played a video where the resolution
changed but the decoding library didn't pick up on it the
buffers were quickly exhausted. (This trigger is a separate
problem, which may even be MythTV specific.)

I may have misunderstood who picks up the pieces when 
alloc_picture() fails, if so I would appreciate a pointer
in the right direction.

-- Daniel


-------------- next part --------------
A non-text attachment was scrubbed...
Name: alloc_picture_leak.patch
Type: text/x-patch
Size: 1101 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20070720/f0b2ce4e/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: alloc_picture_leak_with_allocz.patch
Type: text/x-patch
Size: 1704 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20070720/f0b2ce4e/attachment-0001.bin>