[MPlayer-dev-eng] [PATCH] further dvr-ms playback improvements

[John Donaghy]

> You can use the framerate in the demuxer, either always returning
> previous_pts+frame_time as the pts of the next frame or with some kind
> of adjustment towards the stored pts (if that is needed for some files).

I had no idea that it would be acceptable to calculate more accurate
pts values using the average frame time. This patch attempts to do
that and seems to work pretty well for my NTSC recordings - an average
frametime of 0.033ms is picked up which gives a framerate of
approximately 30fps.

However there is an issue with the (only) HDTV sample I have in that
the average frametime recorded in the file seems to be wrong. In the
'today.dvr-ms' HD sample (which I posted before), the average frame
time is reported as 0.017 which gives an fps of 59.94 - yet when
playing the file with '-demuxer 35' the fps is reported as 25 (from
the codec). Playback looks right with '-demuxer 35' so I believe the
average frame time is just wrong in at least this file. I guess we
need more HDTV samples to confirm if this is a widespread problem
(this file was already unusual in that the audio was unplayable by
mplayer though M$ software can play it correctly).

I just discovered one more thing about the 'today.dvr-ms' HD sample
which is that there is an extra piece of data available in each
segment that is not in any of my NTSC files. Each segment contains
some aditional data identified by mostly unknown GUIDs. In the HD file
one of the GUIDs is actually a known GUID that is supposed to contain
the duration of the media object (in milisecs) and this might prove
useful here.

So it could be the case that the frametime can obtained accurately
using this new known GUID when it is present, and when it isnt present
then the 'average frame time' as obtained by my patch will suffice.
But I dont know without more samples plus I need to write some more
code to extract the frametime value from this HD sample I have.

So I'm not sure what should be done here. Assuming the patch is ok, is
it still acceptable to use it given the above?  Or do we need more PAL
and HDTV samples and if so, from where? I'm guessing we need more
samples but please let me know.

Any other ideas welcome too.

Thanks,

John
-------------- next part --------------
A non-text attachment was scrubbed...
Name: asfpts.patch
Type: application/octet-stream
Size: 4486 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/mplayer-dev-eng/attachments/20060918/d3ce38f8/attachment.obj>