Hi,
I am working on an application written in C# with the Mono framework on
Linux. I have built a managed/unmanaged interface and successfully
played back audio with the PulseAudio client library in a GTK
application with libpulse-mainloop-glib.
For performance reasons I would like to use the threaded main loop API
instead. I set up the thread and call pa_threaded_mainloop_start and all
seems well. However, when I hit a Mono DNS lookup in a completely
unrelated thread the application ceases to respond. An strace reveals
these as the last few events:
poll([{fd=4, events=POLLIN}, {fd=3, events=POLLIN}, {fd=15,
events=POLLIN|POLLPRI}, {fd=17, events=POLLIN|POLLPRI}, {fd=18,
events=POLLIN|POLLPRI}, {fd=19, events=POLLIN|POLLPRI}, {fd=25,
events=POLLIN}], 7, -1) = ? ERESTART_RESTARTBLOCK (To be restarted)
--- SIGPWR (Power failure) @ 0 (0) ---
futex(0x8941a0, 0x81 /* FUTEX_??? */, 1) = 1
rt_sigsuspend(~[INT QUIT ABRT TERM XCPU RTMIN RT_1] <unfinished ...>
As explained on this page Mono makes use of the SIGPWR signal:
http://www.mono-project.com/Debugging
I suspect that the PulseAudio threaded main loop API is stealing signals
that Mono needs to rely upon, as this does not occur if I omit the
pa_threaded_mainloop_start call. Would this analysis be correct and is
there a way to make PulseAudio act more conservatively in its thread?
Thanks,
--
Jay L. T. Cornwall
http://www.jcornwall.me.uk/
