Re: [xine-user] Re: Downsampling solution...

[Bill F. <bil...@mi...>]

Hi Guenter,

On Sun, 14 Oct 2001, Guenter Bartsch wrote:

> [ broken patches ]
>
> > I wonder if something weird happened to blanks or tabs.  Maybe you
> > could try patch with the --ignore-whitespace if you have any problems
> > with future patches of mine, such as I'm enclosing with this, to see
> > if that helps.
>
> humm - perhaps it would also help if you'd send your patches as
> attachments and not included in the mail text - I suspect pine changing
> the whitespaces somehow.

OK.  I'll try using an attachment next time.

> > AV sync with the latest CVS still didn't work quite right for me.
> > I've attached a small patch that does get it to work.  I'm wondering
> > if the GAP_NONRT_TOLERANCE should be a configurable variable.
>
> what was the problem there? a big GAP_TOLERANCE should help in-accurate
> audio timer, but at least the system clock should be accurate enough (even
> on PowerPC).

I'm not seeing the problem any more with the standard GAP_TOLERANCE
of 5000.  I'm not sure what's different now.

> > Perhaps you could try forcing the use of SOFTSYNC by setting
> > oss_audio_sync to 3 in the .xinerc to see how it works for you (make
> > sure to also try with both forward and backward seeking).

> yep - just tried it and it works perfectly for me - in fact, I never
> thought softsync could work that nicely :)
>
> I implemented this syncing method for esd now too and the results are
> great :)

That's definitely great news!

> > The other part of the change was to remove the call to gettimeofday
> > that gets the start_time from ao_oss_open, since the setting of start_time
> > should only be done once, which is done in init_audio_out_plugin.
>
> err - I think it should be exactly the other way round, start_time should
> be resetted in ao_oss_open, setting it in init_... doesn't make sense,
> right?

You may be right that it doesn't make sense, but all I know is it doesn't
work right when done in the open rather than the init.  I just pulled
down a fresh latest CVS and it's broken again.  If you play from the
beginning it's nicely synced, but if you seek into a new chapter, which
I believe causes the open function to be called, the audio once again
lags the video by about 0.5 second.  If I move the gettimeofday that
gets the start_time from the open to the init, then the AV sync is fine
regardless of whether playing from the beginning or seeking either forward
or backward.

						-Thanks

						-Bill