[FFmpeg-devel] [RFC v2 3/3] daaladec: Implement a native Daala decoder

Andreas Cadhalpun andreas.cadhalpun at googlemail.com
Wed Dec 30 01:09:37 CET 2015


On 29.12.2015 22:34, Rostislav Pehlivanov wrote:
>> This is an infinite loop that can hang the decoder.
> Will try to get the upstream to accept a fix.

Thanks.

>> overflows in the entropy decoding system
> The assertions should be able to catch them, so they should be fine.

Are you talking about av_assert0 here?
None of them was triggered, which is good, as triggering an assert is
usually much worse than an overflow.

> They'll trigger if there's a desync with the bitstream. I might add more in
> case they're not enough.

I don't think that more asserts will fix the crash caused by one of these
overflows. That has to be fixed differently.

>> What is the point of having DaalaSharedContext in addition to
> DaalaBitstreamHeader?
> I was looking at VP9 when starting to write this and it had this structure.
> I'll  remove the shared context one.

I see.

>> Third, the decoder seems to return only the first frame, repeatedly,
>> instead of the following frames.
> You did read that it can only decode I-frames (keyframes). So you need to
> encode your files with -k 1 to make all frames keyframes.

I must have missed that, but I hope this decoder will support other frame
types in the future. ;)

>> First, this fails building with --enable-shared, because
> daala_find_p_format
>> is not static.
> Fixed yesterday, but sent the same patch twice by mistake. It's been fixed,
> but I'd like to post a v3 of the patch to address some of the issues found.

OK.

Best regards,
Andreas



More information about the ffmpeg-devel mailing list