On Mon, 28.04.08 09:53, Colin Guthrie (gmane at colin.guthr.ie) wrote: > > Sjoerd Simons wrote: > > The attached patch extends the alsa pulse plugin set with a alsa > > configuration hook. Allowing one to specify some configuration parameters > > that only come into effect when pulseaudio is running. > > I was waiting with baited breath after your chat on IRC as to what your > solution would actually be! > > I like it! > > I think the overhead at initialisation is probably OK, although I guess > it depends on how an application works - e.g. if it constant opens and > closes alsa connections then does this overhead mount up? If not then > this looks like a good solution. > > That said Lennart has said in the past that he didn't like the idea of > Alsa automatically falling back to h/w access if pulse isn't > responding. (this was in reference to tweaking the alsa plugin itself, > rather than this approach which is slightly different conceptually, > albeit with the same end goal). I think his comment at the time was "it > masks configuration problems". > > Do you have any thoughts on this Lennart? I think that the patch is OK although it does have some problems as Sjoerd already pointed out. My plan for the future is to enable auto-spawning by default in addition to starting PA by gnome-session. This will tape over these problems, too. The reason for this is that as it stands right now, PA doesn't really work from the console. Since console logins don't have anything like a session manager, we cannot really start PA from there. Same with SSH logins or System services using PA. Unfortunately console logins are still very important for a11y people, and audio happens to be too. So, PA will be started as user-daemon (not session daemon!) from session managers, in addition to be started by auto-spawning. This will tape over configuration problems, but heck, I see no better way. Then I will write a small module that terminates PA if a) no client is connected anymore and b) the user owning it is not logged in anymore according to CK and c) some time has passed. Then I'll add some code that PA won't terminate anymore when an X11 connection dies. And then everything is bright an shiny. That way PA will probably feel more robust to many people and we'll win the hearts of our a11y people ;-) So, in summary: I don't think Sjoerd's patch is flawless, or that it is the way to the future. But it's an OK temporary fix on our way to world domination. It get's the "Approved by Lennart" stamp. ;-) Lennart -- Lennart Poettering Red Hat, Inc. lennart [at] poettering [dot] net ICQ# 11060553 http://0pointer.net/lennart/ GnuPG 0x1A015CC4
