[FFmpeg-devel] [PATCH] H.264/AVCHD interlaced fixes

[Alexander Strange]

On Feb 9, 2009, at 5:34 AM, Michael Niedermayer wrote:

> On Mon, Feb 09, 2009 at 12:52:46AM +0100, Ivan Schreter wrote:
>> Hi Michael,
>>
>> Ivan Schreter wrote:
>>> Ivan Schreter wrote:
>>>
>>>> Michael Niedermayer wrote:
>>>>
>>>>> keyframe |= ...
>>>>>
>>>> Do you mean I should add additional "keyframe" flag? I can do  
>>>> that...
>>>>
>>> Ah, I see that the key_frame flag is already set on current  
>>> picture. So it
>>> was OK as it was, the problem is elsewhere. If recovery_frame_cnt  
>>> >= 0, it
>>> should OR a key frame flag, not set it to recovery_frame_cnt >= 0.  
>>> I will
>>> correct that.
>>>
>>>
>> I suppose the attached patch regarding SEI recovery point will now  
>> suffice
>> your critical eye :-)
>>
>> I made it independent of #1, so it can be applied as-is.
>>
>> Please review & if you are happy with it now, apply it.
>
> [...]
>> @@ -7340,6 +7344,8 @@
>>     int context_count = 0;
>>
>>     h->max_contexts = avctx->thread_count;
>> +    h->sei_recovery_frame_cnt = -1;
>> +
>
> i suspect that this will misbehave if CODEC_FLAG2_CHUNKS is set

Is that flag still needed for anything? AFAIK it was only added for  
one use, and can be replaced by using a parser; there isn't any code  
using it anymore according to Google Code Search. If it's not needed,  
I have a patch to remove it from h264.c.