On Tue, Jul 19, 2011 at 09:18:46PM +0200, Axel Dörfler wrote: > pete.goodeve@xxxxxxxxxxxx wrote: > > Some installations will not produce sound immediately after boot up, > > but > > sound will start working later. This appears to somehow be a result > > of > > a large (default) buffer size in the driver. [....] > > The large buffer size also results in a latency of ~1/5 sec. > > Since HDA has similar issues, I guess it's not driver related at all. Yes, I think so too. What I'm wondering is if/how buffer size relates to the problems. Are buffers large for other drivers -- preventing crackles, but implying longer latency (and somehow not causing startup silence)? Or are the buffers small, but the crackles don't show up for some reason? > I would also assume that auich worked nicely on BeOS a while back - > Jérôme could clarify. What's the history here? It was written for BeOS originally? > I have HDA and didn't have any issues for a long time, but since a few > months, audio has problems like being silent at first. This is also a bit odd, because the problem has been around for me ever since I first installed Haiku -- over two years on Scott's old Thinkpad that requires OSS, and about a year, I guess, for the auich box (Dell). > > I think the main issue is somewhere in the media kit, possibly some > broken latency computation or something similar. (Whimper... (:-/)) This is the battle I was fighting for most of this year. [see the saga in #7285...] Latency is certainly currently broken -- if I could only convince anyone else...! The mods I made to the latency computation improve things on my machine, but they don't cure the essential problem, which is that 'blackouts' of ~8-10 msec occur, and these can delay a buffer anywhere in the chain. If a buffer doesn't arrive in time, you're going to get a crackle. (My fixes just prevent latency increasing without limit when a buffer is late.) The blackouts seem highly causally related to the display. I now have a command-line only version of ToneProducer (with latency reporting); if I run this remotely, so the display is not being touched, it is quite clean -- I can run forever without a glitch. If I run it from a Terminal, it's usually the same. Any slight movement of any window though -- or possibly even just the mouse (?) -- can cause a blackout. Interestingly, it may be strictly window related. If I run the teapot and am careful not to move any window at all, I never seem to get a blackout. Even if I use the mouse strictly within the teapot window, it doesn't seem to glitch. That's all I have for now. Not sure where to look from here. -- Pete --