[vlc-devel] [PATCH 3/3] mediacodec: implementation of MediaCodec direct rendering based on the work by Martin Storsjö.

[Felix Abecassis]

I was able to reproduce this deadlock on N5. Thank you for pointing it.

Could we promote the static function DecoderIsExitRequested from
src/input/decoder.c to the public API in include/vlc_input.h?
We could use this function instead of checking liveness of the decoder
object and this should work.

2013/12/23 Rémi Denis-Courmont <remi at remlab.net>:
> Le lundi 23 décembre 2013, 20:35:21 Martin Storsjö a écrit :
>> Hmm, I'm not familiar enough with stream output in VLC to know how it
>> would interfere here.
>>
>> For the race condition, I guess you mean that when closing/finishing, it's
>> undefined whether the last few frames get decoded or not.
>
> Generally, I mean you don't really have a warranty that alive ever returns
> false until pf_deactivate() is called. You are making not always correct
> assumptions about the way decoders may be used. Simply put, a decoder relying
> on that to return from its pf_decode() is a buggy decoder.
>
>> If that's the case, I think it is an acceptable tradeoff for the direct
>> rendering/opaque output buffers case - we could do this only in that case
>> (not in the currently default YUV buffer output mode).
>
> I thought so too when I started VLC development. But a lot of users seem to
> notice. So I don't think so anymore.
>
> --
> Rémi Denis-Courmont
> http://www.remlab.net/
>
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> https://mailman.videolan.org/listinfo/vlc-devel



-- 
Félix Abecassis
http://felix.abecassis.me