Re: [Jack-Devel] buffer size callback, revisit

Am Dienstag, den 15.02.2011, 13:41 -0600 schrieb Jack O'Quin:
> On Tue, Feb 15, 2011 at 1:25 PM, Paul Davis <[hidden]> wrote:
> > On Tue, Feb 15, 2011 at 2:22 PM, Jack O'Quin <[hidden]> wrote:
> >
> >> If JACK does not, the best hope is not to change the buffer size.
> >> That's been the de facto solution for the past ten years or so. I
> >> doubt very many JACK clients ever worked correctly with this
> >> interface.
> >
> > so, if i understand correctly, you are suggesting that we simply
> > convert the implementations so that the callback is always called from
> > jack_activate(), and tell client authors "you should assume that its
> > called from jack_activate(), though it may not be, and that's a bug in
> > older versions of JACK" ?
> 
> Not exactly. The issue is complex, because one needs to look at it
> from both sides. Existing clients will want to work as best they can
> with old JACK versions. Removing the initialization code would not
> serve that purpose well, despite being subject to a race condition.
> 
> I am not against adding new, cleaner interfaces. But, let's not hold
> our breath waiting for everyone to adopt them. That takes years.
> 
> I was just asking: what's wrong with fixing the bug in the existing one?
> 
> I suspect I may have missed some earlier message where that problem
> was described already. If so, I apologize.

Many jack clients cant handle buffersize changes anyway, mostly because
the jack_set_buffer_size_callback() isn't involved.  

But anyhow, clients witch handle it, and use internal buffers, have
allocate the buffers before they called jack_activate(), so, a change in
buffersize in between this time, leads anyway to a new allocation, when
the client notice the change, isn't it ?  

so wouldn't a call like 

if (!jack_activate(jack_client))
 if(sz != jack_get_buffer_size (jack_client)){buffersize change handle }

handle the race condition without change anything in jack ?

regards
hermann

1297830242.9505_0.ltw:2,a <1297830129.2052.30.camel at box>