[FFmpeg-cvslog] avcodec: Add av_fast_padded_malloc().

[Michael Niedermayer]

On Fri, Feb 03, 2012 at 08:24:34PM +0100, Reimar Döffinger wrote:
> On Fri, Feb 03, 2012 at 08:13:34PM +0100, Michael Niedermayer wrote:
> > On Fri, Feb 03, 2012 at 07:11:34PM +0100, Reimar Döffinger wrote:
> > > On Fri, Feb 03, 2012 at 04:55:12PM +0100, Michael Niedermayer wrote:
> > > > On Thu, Feb 02, 2012 at 09:50:44AM +0100, Reimar Döffinger wrote:
> > > > > On Thu, Feb 02, 2012 at 02:46:27AM +0100, Janne Grunau wrote:
> > > > > > ffmpeg | branch: master | Janne Grunau <janne-libav at jannau.net> | Tue Jan 31 15:40:11 2012 +0000| [316fc7443b05f8c8b494443f2dfe590434796902] | committer: Janne Grunau
> > > > > > 
> > > > > > avcodec: Add av_fast_padded_malloc().
> > > > > > 
> > > > > > Wrapper around av_fast_malloc() that keeps FF_INPUT_BUFFER_PADDING_SIZE
> > > > > > zero-padded bytes at the end of the used buffer.
> > > > > > 
> > > > > > Based on a patch by Reimar Döffinger <Reimar.Doeffinger at gmx.de>.
> > > > > 
> > > > > No idea if you merged this, but this variant will not fix
> > > > > valgrind failures mine did since it does not always ensure
> > > > > all data in the buffer has been initialized at some point.
> > > > 
> > > > I mixed and shaked them together IIRC seems fate valgrind is still
> > > > green so i naively assume i didnt mess up
> > 
> > This diff that went in:
> > 
> > diff --git a/libavcodec/utils.c b/libavcodec/utils.c
> > index 0df3d7f..f21e36f 100644
> > --- a/libavcodec/utils.c
> > +++ b/libavcodec/utils.c
> > @@ -91,7 +91,7 @@ void av_fast_padded_malloc(void *ptr, unsigned int *size, size_t min_size)
> >  {
> >      uint8_t **p = ptr;
> >      if (min_size > SIZE_MAX - FF_INPUT_BUFFER_PADDING_SIZE) {
> > -        *p = NULL;
> > +        av_freep(p);
> >          *size = 0;
> >          return;
> >      }
> > 
> > and this should fix some memleaks in the failure case
> 
> Ups. Yes, embarrassing, that definitely is right.
> But nothing will have broken then, great.
> And maybe we'll get another valgrind machine that
> runs the full program somewhen.

btw, this reminds me, the ppc64 machines seem stuck

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

DNS cache poisoning attacks, popular search engine, Google internet authority
dont be evil, please
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-cvslog/attachments/20120204/eda17b1e/attachment.asc>