[xine-user] RE: [xine-devel] 0.4.0alpha is in CVS

[Guenter B. <bar...@st...>]

Hi Joachim,

On Mon, 12 Feb 2001, Joachim Koenig wrote:

> > - syncfb video driver
> 
> I will check and correct this beast tonight.

great :-) there hasn't changed much with the video output drivers, the
only thing to watch out for is that the video driver had to free the
images given to it via display_frame () as soon as they're no longer
needed so they can be re-used by the decoder.

This means you have to call 

vo_image_drawn (  (vo_image_buffer_t *) gXv.cur_image);

on every image you get delivered as soon as you're finished drawing it.

One easy way to implement this is to remember the image currently
displayed (gXv.cur_image for example) and free it at soon as you get a new
image (take a look at display_frame_xv to see what I mean). In the
driver's reset function you should then release the currently displayed
image too (this function is called on a resolution change for example).

Apart from that, make sure alloc_image_buffer () initialized the mutex
in the image buffer structure

  pthread_mutex_init (&img->mImageBuffer.mutex, NULL);

> I will also look into alsa sound driver and avi demux. Do you want patches
> or  shall I update CVS directly.

you can directly update CVS :-)   ... if you find out anything about
synchronization in AVIs I'd be very interested in hearing about out - so
please drop me a line. The trouble with the avi demuxer is that it has to
generate PTS for the video and audio packages it deliveres. For video,
that doesn't seem to be a big problem, we know the frame number and the
exact frame rate from the avi header. But I'm not sure about audio - the
audio packages contain mp3 audio so we can't simply count samples since we
don't know how many samples a packages will generate when decoded. I tried
two different approaches, first I used nAvgBytesPerSec from the AVI header
- that works for most AVIs I have here, but not for all of them (with two
test streams I have here audio and video drift apart). Then I tried to
calculate that value myself

  gDemuxAvi.pts_per_byte = (((double) gDemuxAvi.avi->video_frames /
(double) gDemuxAvi.avi->fps) * 90000.0) / (double) gDemuxAvi.avi->audio_bytes;

that works better for the two "bad streams" bad isn't exact enough.

> PS: By what early hours do you work ??

*grin* .. yeah, got a bit late last night ;))

Regards,

   Guenter