[Ffmpeg-devel] [Bug] h264 decoder crash (sample included)

Loren Merritt lorenm
Sun Feb 11 00:13:25 CET 2007

On Sat, 10 Feb 2007, Christopher Lux wrote:
> Michael Niedermayer <michaelni at gmx.at> wrote:
>> the crash will be fixed when i have time or someone else does
>> as i said be patient, asking every few days wont speed that up, the decoding
>> errors, i assume the stream uses PAFF will not be fixed by me, maybe someone
>> else will but i wont touch the messy h.264 interlacing
>> if you seriously want PAFF support then your choices are
>> * implement it yourself, this is not fun the h.264 spec is a mess which
>>   generally needs to be read 3 times before you understand it, though OTOH
>>   it doesnt need any magic. knowledge of C and lots of time is all thats
>>   needed and many people will be happy if you add PAFF support ...
>> * pay someone else to do it (your chances here are not that bad, as there are
>>   many people who want PAFF who might contribute some money) but dont expect
>>   me to be the one doing the PAFF coding ...
> is PAFF really the problem?

Yes, because lavc spews a bunch of
"[h264 @ 0xc6d430]PAFF interlacing is not implemented"
just before crashing.

Well, the fact that it crashes is due to some missing input validation 
somewhere and could also be triggered by a corrupt stream without PAFF. 
But fixing that would just make it display garbage or reject the frames 
instead. You won't get a correct decode without implementing PAFF.

--Loren Merritt

More information about the ffmpeg-devel mailing list