Re: Media PLayer 9, won't play with my video decode filter...

Tech-Archive recommends: Repair Windows Errors & Optimize Windows Performance

From: Chris P. [MVP] (msdn_at_chrisnet.net)
Date: 06/01/04 

Date: Tue, 1 Jun 2004 12:27:15 -0400

illiminable wrote:

>
> I do have one, but it's only a result of connected a graph like
> this... is this what you tried or did you try and use the other xiph
> tools to create one ?
>
> Connect the demux source to 2 muxers, one for the vorbis one for the
> theora. Though this will make files that have some wrong timestamps
> (i haven't properly tested it), due to the lack of timestamps of most
> ogg packets (ie on avergae only every 20th packet in ogg gets a
> timestamp, or the stamps are at page level not packet level) An easy
> way to strip it is to look at the source for OOOggDump in my
> project... it only has one dependant library libOOOgg (my demux code)
> and it should be fairly obvious how to modify it to write out one
> stream of pages and drop the other. This allows you to split at the
> page level, whereas the filters all operate at the packet level.
>
> I've put one on my webspace that is split from one of my test
> files... the vorbis only stream works, but the theora only stream
> doesn't. I'm fairly convinced i've got some subtle bug in my decode
> filter somewhere, i have also got some new errors, Size not Set from
> calling getdeliverybuffer, which would indicate to me that somehow
> the commit or setpropeerties didn't work properly, and it has nulled
> allocator properties, and also i've found if i persistently force it
> data at it even after it rejects it, it eventually drops back to a
> Not committed error.
>
> But that still doesn't explain what the exact difference is between
> graphedit and WMP 9... i can't seem to find any documentation
> anywhere about how wmp9 operates differentely if it does, or whether
> it has different requriements.
>
> Here's the theora only and vorbis only stripped from the movie on the
> site linked to below. www.illiminable.com/th/anc_theora.ogg ...(about
> 8 megs) just note it is not properly mxued because of the packet
> stamping issues i mentioned before, but in graphedit you will at
> least get the first 10 seconds of it, which is enough to show it
> play... but this doesn't happen in wmp, it exhibits the same problem
> as the fully muxed stream.
>
> Also another thing, if you have a ogg/ogm/divx video in one stream and
> vorbis in the other stream, it plays fine. For example (you also need
> ffdshow to provide divx decoding), this file...
> http://www.polycrystal.org/lego/movies.html lpaying the divx/vorbis
> in ogg file works fine in WMP9... and i've looked through the source
> of ffdshow to see if they do anything different and it doesn't appear
> so... they do derive from CVideoTransform, whereas i created my own
> intermediate audio and video decoder abstractions, and basicly derive
> from CBaseFilter, CBaseOutputPin and CBaseInputPin and from nothing
> deeper in the inheritance hierarchy.

I made a demuxed file with GraphEdit but it didn't work at all. I connected
the Vorbis side to a Null renderer.

I experienced the same thing with your file, plays (mostly) in GraphEdit but
not in WMP. It also works fine in MPlayer 6.4 which should use a standard
filter graph.

The only thing I can think of is that it's a problem with the allocator, I
haven't been able to delve into the source as I'm on a deadline with another
project. There are some situations when using a different combination of
filters, that the negotiation process may occur a little differently. I had
a case recently where the buffer wasn't allocated large enough and the
symptoms were almost the same.

You mark your output as YV12 but I'm unable to connect it to a Color Space
Converter in order to connect it to another renderer. Only the XP default
render and the VMR9 seem to be accepting a connection.

Thore, do you have any words of wisdom here?

Relevant Pages