[mythtv] [mythtv-commits] mythtv commit: r24384 by jyavenard

[Jean-Yves Avenard]

Hi

On 4 May 2010 21:14, Colin Guthrie <mythtv at colin.guthr.ie> wrote:
> That's a shame, but there are very real and practical reasons for not
> changing the sample rate on the fly (people pop up about every six
> months and ask us why!)

I perfectly understand why pulse would open the sound card once , and
not let applications directly change it ; after all it is a shared
resource , why would one have priority over another.

This only became an issue as we were trying to bypass problems when
pulse was doing the resampling.

At some stage, the code worked every time pulse was resampling, but
*not* when it wasn't resampling. Very weird


>
> I'd be very interested in the problems that occur when PA does the
> resampling as the best route forward is to make that work as it would
> save doing similar workarounds in all apps.

Another issue is related to ALSA showing buffer underruns issues ;
pulse starts to behave very erratically when this occur. Finding why
it behaved in such matter was a long a tedious task.

pulse expects 24 bits fixed be stored in the LSB of an int, this is
different to what ffmpeg feeds us (or what ALSA takes too).
Not that it matters much in the end, as the fixed -> float conversion
can do the shift in one cycle...

Sometimes Pulse remapped channels, sometimes it didn't etc..

But FooBar will be able to explain better than me here.

>
> Feel free to give FooBar my email address or tell hime to hassle me on
> IRC (coling) so that he can discuss the problems and hopefully work out
> a solution.

He subscribes to this distribution list, so he will be reading it all,
and hopefully answer.
I had told him that he should be contacting you when he hit all those
issues, but he wanted to find the problem on his own.

>
> But awesome work all round - KUTGW both of you :D
>

Thanks..

it has been a few very long weeks , with lots of sleepless nights.

Jean-Yves