Le 14 mars 08 à 16:32, Salvatore Di Pietro a écrit : > Hi all, > I just fitted Sean Meiners' alsa device disconnect patch for JACK > 0.109.2. > > It enables jackd to detach from the ALSA device, letting its use by > other non-JACK programs whilst in the middle of a jack session, > without closing jackd and its clients. IMHO very useful for i.e > common builtin laptop soundcards which don't support more than one > alsa-direct audio app at once (also commonly known as "no hardware > mixing soundcards"). > > You get an additional example client: jack_device_connect (based on > jack_freewheel, its arguments are "y" or "n") which you use to make > jack release and re-hook to the audio card. Time is paused and all > audio from jack is of course muted and jackd and its client > applications are put into this fake-freewheel mode while jack has > released the device; they continue producing and consuming audio > upon jackd re-attaching to the audio device, with the command > "jack_device_connect y". > > The original version was for jack 0.100.0, I just modified it to > make it patch successfully on jack 0.109.2. I don't have so > advanced programming skills to understand its inner working, but > from a quick read of the code, for what I can understand, I saw it > mainly adds some functions in jackd (jack_device_disconnect() and > jack_device_connect()in jackd/engine.c) and libjack > (jack_set_connected() in client.c, jack_driver_nt_disconnect() and > jack_driver_nt_connect() in driver.c) to dis/connect to the alsa > device. > > I was asked on this list about the interaction of this patch with > realtime watchdogs. I didn't encounter any problem, they were never > triggered (and no crashes or weird things ;) ) as in normal > operation, because they enter this freewheeling-like mode. > I contacted Sean and asked him if it was this way, and he replied: > > >As I remember, that's pretty much it. It puts jack into something > like >freewheel mode (so that jack clients don't freeze up). And > then shuts >down the alsa driver. As for the watchdog, I don't > remember anything >about that, but if I had to guess I'd say it's > because jack is put into >a fake-freewheel mode. > > Hope it is useful to who wants to try it: just use > > jack_device_connect n > > and voila, you can watch youtube videos or use that strange non- > jack aware application you need in that moment, _without_ closing > any jack program. > When you're done, just use > > ja Hi, I don't think this patch should go for jack 1.0 for several reasons: - it is quite intrusive and add a new API jack_set_connected on the client side. - dynamically changing backend is something that was discussed at LAC and that should be better done in a new "server control" API that will be part of jack2. Changing the backend is conceptually more a server control operation. This API is yet to be completely defined but should allow much better control of the server, like dynamically gettting/setting backend parameters, starting/stopping the server, dynamically changing backend.... Stephane _______________________________________________ Linux-audio-user mailing list Linux-audio-user@xxxxxxxxxxxxxxxxxxxx http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
