David Henningsson wrote: >For the record, I received some test code from Joerg a few weeks ago, >and tested it on Ubuntu 12.04 (based on PulseAudio 1.1 + my jack >detection), and I was unable to find something as serious as the >buffer_size and avail values going havoc. Please keep in mind that out of range avail values was just one of my gripes. Equally important is to observe the dynamic changes of these values: - snd_pcm_delay is amenable to a speaker position. As such, it should change very regularly. However, on my Ubuntu PA system, it sometimes purely and simply stops for as long as 200ms! (without intervening underrun) The deltas in snd_pcm_delay do not correlate with elapsed time. - snd_pcm_avail also changes in a bizarre way. >IIRC, Joerg uses an older version of PulseAudio. Ubuntu Intrepid and Ubuntu Lucid. I was glad to hear from David that he could not observe the bad behaviour that I mention on his machine with the latest PA/alsa_plugins. However, I don't know what to tell users of the old Ubuntu systems, some of which use the long term support (LTS) distributions. My experiments with PA on those old systems make me wonder how it can manage to play anything or why even speaker-test works. The behaviour of avail and delay is so irregular and non predictable that I simply don't know how to write code for Wine that would reliably allow to play a simple continuous tone. I simply don't understand why some apps on my Ubuntu systems appear to play audio nicely for a few hours when I can't for a few seconds. Note that MS's mmdevapi wants to write data every 10ms. It's not the "write 2s and be done" style which old PA seems to prefer. In random cases, PA will report an underrun shortly after the initial write. snd_pcm_recover gets called and I can hear an audible pause. So what am I going to tell the users of such systems? David, not me, is from Canonical, what are you going to tell LTS users? Regards, J?rg H?hle
