[MPlayer-cvslog] r30109 - trunk/libmpdemux/extension.c

[Reimar Döffinger]

On Thu, Dec 24, 2009 at 08:57:48PM -0500, compn wrote:
> On Fri, 25 Dec 2009 00:59:42 +0100, Reimar Döffinger wrote:
> >On Thu, Dec 24, 2009 at 07:20:04PM +0100, compn wrote:
> >> Author: compn
> >> Date: Thu Dec 24 19:20:03 2009
> >> New Revision: 30109
> >> 
> >> Log:
> >> force cdg to lavf demuxer in extensions.c, misdetected as mpeg otherwise
> >> 
> >> Modified:
> >>    trunk/libmpdemux/extension.c
> >> 
> >> Modified: trunk/libmpdemux/extension.c
> >> ==============================================================================
> >> --- trunk/libmpdemux/extension.c	Thu Dec 24 11:56:25 2009	(r30108)
> >> +++ trunk/libmpdemux/extension.c	Thu Dec 24 19:20:03 2009	(r30109)
> >> @@ -81,6 +81,7 @@ static struct {
> >>          { "eac3",DEMUXER_TYPE_LAVF },
> >>          { "mac", DEMUXER_TYPE_LAVF },
> >>          { "str", DEMUXER_TYPE_LAVF },
> >> +        { "cdg", DEMUXER_TYPE_LAVF },
> >
> >You should only use this as a very last resort.
> 
> put a note in extension.c for future?

I did.

> you want me to move it to demux_lavf?
> i should check str too...

Actually I think all with DEMUXER_TYPE_LAVF should go there, none of
those currently there fail probetest (even after removing the MAX/4
limit where it would normally start warning).

> >For example adding this format to lavf_preferred in demux_lavf is a
> >better solution.
> >This assumes of course that the auto-detection in lavf is reliable -
> 
> i was only testing the one sample. i'll dig up some more.
> 
> >if you have FFmpeg SVN at hand, check if it passes tools/probetest.
> 
> dont have it atm.

I tried probetest, it is fine. You don't need more samples, all this
does not make a difference for formats that FFmpeg incorrectly rejects,
you only have to check that FFmpeg will not incorrectly detect something
as CDG that isn't - and probetest tells me it won't.

> btw cdg looks corrupted in mplayer, i couldnt narrow it down to any
> one particular problem. someone said that it might be a pallete
> issue (-vf palette gives diff output) ? any ideas?

No, since it does not use the deprecated palette_control there should
be no issue.
Do we have CDG samples on samples.mplayerhq.hu? Once I get X working
again here I might look into it (the OpenSource ATI drivers still just
so don't work, not even 2D-only).

> also extension.c maps ogg/ogm to demuxer_ogg, should ogv be done
> this way too?

That's probably a case of optimization for streaming, so probably not.
However if there are samples of ogg files that are misdetected as
something else when they don't have the right extension that would
preferably be fixed.