On Fri, Jun 29, 2012 at 04:59:48PM -0700, Urias McCullough wrote: > On Fri, Jun 29, 2012 at 4:34 PM, <pete.goodeve@xxxxxxxxxxxx> wrote: > > [Even more unfortunately, when I got r44242 running on my own desktop > > box, it turns out to be useless for me there, too! As soon as I start > > feeding the audio chain (auich) with anything, the screen starts to > > freeze! Adding a second mixer input locks the whole OS completely! > > These regressions need to be documented, and hopefully the source > identified (via binary-search of revisions between A3 and current). Yes, I intend to do so. At the moment, though,I'm a bit foggy as to exactly what's happening. On some further probing I found that MediaPlayer still works fine. It's only my own code (that was also fine under the earlier rev) that causes the lockup. (:-() If I go into KDB, I see either the mixer thread or the auich one is the "running" -- but locked -- thread. It may be that it's because I have to use Recording Mode to avoid the increasing latency problem. Otherwise my code is pretty standard BSoundPlayer usage. (Also, I didn't see any lockup on the ZaReason box. (May not have tried turning off the extra cores, though. I did for other tests -- no noticeable effect -- but I didn't know about the lockup problem then.) I'll mount a full innvestigation... > Sounds like it could be an interrupt issue, or it could be related to > some recently discussed priority inversion issue in the scheduler. I think it must be something like that. It has to be a deadlock somewhere. -- Pete -- [BTW, I did get the new rev to boot. Apparently there's been a change in the bootloader at some point. When I used the BootManager from the new USB-stick to install its MBR on the hard drive, it now works -- and so do the older revs still, thank goodness! I won't do any more binary searching on that, whioch was *extremely* painful, taking several hours for each cycle. As I said, writing to USB also seems to be slow...]