[FFmpeg-devel] [PATCH] avcodec: increase AV_INPUT_BUFFER_PADDING_SIZE to 64

[Reimar Döffinger]

On Mon, Jan 08, 2018 at 04:14:22AM +0100, Michael Niedermayer wrote:
> On Sun, Jan 07, 2018 at 11:42:25PM -0300, James Almer wrote:
> > On 1/7/2018 11:06 PM, Michael Niedermayer wrote:
> > > On Sun, Jan 07, 2018 at 01:22:38AM -0300, James Almer wrote:
> > >> AVX-512 support has been introduced, and even if no functions currently
> > >> use zmm registers (able to load as much as 64 bytes of consecutive data
> > >> per instruction), they will be added eventually.
> > >>
> > >> Signed-off-by: James Almer <jamrial at gmail.com>
> > >> ---
> > >> Same rationale as when it was increased to 32 back in commit
> > >> 67d29da4bd23057a1f646568442a77b844cb2d1b.
> > >>
> > >>  libavcodec/avcodec.h              | 2 +-
> > >>  libavcodec/x86/hevc_sao.asm       | 2 +-
> > >>  libavcodec/x86/hevc_sao_10bit.asm | 2 +-
> > >>  3 files changed, 3 insertions(+), 3 deletions(-)
> > > 
> > > causes assertion failure:
> > > ./ffmpeg -i bgc.sub.dub.ogm x.webm
> > > 
> > > Assertion 64 <= 52 failed at libavformat/oggparseogm.c:109
> > > Aborted (core dumped)
> > 
> > How do you say this should be fixed? Why should
> > AV_INPUT_BUFFER_PADDING_SIZE be lower or equal to 52?
> > 
> > I have no idea about OGM, so i could use some help.
> 
> cc-ing reimar, maybe he remembers, the assert seems to come from:
> 
> commit 7c8d477299c9b5e89fc30ed22f9e42b50761342c
> Author: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
> Date:   Mon Feb 6 22:04:46 2012 +0100
> 
>     Make AAC in Ogg (ogm) work.

The assert probably never made sense (probably left-over from an early
try that would do the allocation earlier on, in that case we wouldn't have
needed to reallocate the data to add the padding before storing in
extradata), but it should have been removed at the latest when the code
was switched to use ff_alloc_extradata.
Either way, it should no longer be relevant.