[vlc-devel] [RFC PATCH 0/2] Decoder fallback (new question inside)

[Thomas Guillem]

On Mon, Apr 11, 2016, at 11:48, Rémi Denis-Courmont wrote:
> Le 2016-04-08 16:44, Thomas Guillem a écrit :
> > What do you think ?
> 
> I think you are overthinking this. I don't think that fallback could be 
> smoothless even if you preserved the block.

This is not really an issue for Video decoders: it's OK to wait for a
ref frame or have some video glitches in case of a HW decoder error.
These errors are not that common.

My main concern was about audio. An other goal of the decoder fallback
is to move some audio filters to audio decoders, like a52tospdif and
a52tofloat. So if the a52tospdif filter become a decoder, it will be
always tried before a52tofloat, and fails in decoder_UpdateAudioFormat
if there is not audio_output that can handle spdif. The drawback is
that, in most case, a first block (or more ?) will be always dropped if
you want to play PCM from a52.

I don't see any others solution than the current audio-filter behaviour
for now.

> 
> Potentially you would also need a few previous block (depending on 
> coding and packetizer). Also in the original use case, lost S/PDIF 
> format, audio glitching is unavoidable due to buffering and hardware 
> constraints.
> 
> -- 
> 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