[LAD] [Jack-Devel] more jack/qjackctl madness

[Rui Nuno Capela]

Fons Adriaensen wrote:
> On Sun, May 17, 2009 at 11:48:57AM +0100, Rui Nuno Capela wrote:
> 
>> qjackctl only writes to ~/.jackdrc *iif* you opt to (see Setup/Misc/Save
>> JACK audio server configuration). otherwise it *never* makes use of that
>> file ie. it never auto-starts jackd implicitly; it only does start jackd
>> explicitly if none is found running atm.
> 
> If the file is ignored by jackd, and by qjackctl, why
> does qjackctl write it at all ?
>  
>> if you really want several jackd server running simultaneously
> 
> I don't want that. I want to get rid of one that was 
> started automatically, terminated, but restarted by
> dbus (why on earth ?) when I run qjackctl. The net
> result is that running qjackctl starts a server with
> parameters that have no relation at all to its setup
> (making it appear as if qjackctl has gone nuts), and
> starting the one corresponding to qjackctl's setup
> becomes impossible. It's very probably not qjackctl's
> fault.
> 
> Two other things _are_:
> 
> 1. I definitely want to be able to terminate qjackctl
> without stopping jackd, even if that jackd was started
> by qjackctl. This used to be possible.
> 

unfortunately, Qt4's class which is being used for wrapping the jackd
process (QProcess) does kill it on its destructor. afaict, this behavior
wasn't present in Qt3 and that's why there's no option to keep jackd
running upon quitting qjackctl anymore.

> 2. Qjackctl-0.3.4 seems to have a bug handling the port
> creation and destruction callbacks, it continues to show
> ports that have been deleted and does not show some new
> ones. This happens when the delete/create calls are close
> together. Jack_evmon shows the right events in the right
> order, so it must be qjackctl getting it wrong.
> 

this might just be a side-effect of the jack_port_t* reuse issue that i
think has been discussed recently.

qjackctl does not handle port (un)registration callbacks but does its
book-keeping with this jack_port_t * handle as a unique key identifier.

if, for instance, one client unregisters one given port and registers
another in a quick sequence, it might just happen that the new port has
the very same jack_port_t* address value which got just reclaimed and
reused. alas, qjackctl assumes it's the very same port object as it
keeps jack_port_t*s as its unique port identifiers, so nothing changes.

i suppose i'll have to fix this soon. by properly handling port
unregistration callbacks, which will be a novelty after 5+ years of
doing just fine without on jack1 at least ;)

cheers
-- 
rncbc aka Rui Nuno Capela
rncbc at rncbc.org