Using the subscription events caused an assertion crash sometimes when a sink
was removed and a new sink was created (i.e. card profile change) and a stream
was moved from the removed sink to the new sink. The stream dbus object's
subscription callback got a change event before the core dbus object's
subscription callback got the sink remove/creation events. The stream's
subscription callback then queried the core for the object path of the new
sink, and since the core was not yet aware of the new sink, an assertion was
hit in pa_dbusiface_device_get_path().
Now that the core uses synchronous hooks to keep the sink and source lists up
to date, this particular problem can't occur anymore.
This is required to when playing on a52: device, rewind is broken
in those plugins.
Credits to Michael Rans <mcarans@yahoo.co.uk> for finding this
workaround, and Tanu Kaskinen <tanuk@iki.fi> for providing
valuable feedback.
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@intel.com>
Previously we used libdbus's memory as keys in listening_signals, which caused
that the memory of the hashmap keys got overwritten, which led to that signals
weren't sent properly.
Positive base volume can happen, if the alsa volume range has been limited. For
example, in an embedded environment it may be known that the sound device is
capable of louder output than what the speakers can handle, so setting the max
volume below 0 dB makes sense.
The syntactically correct error meant that the timestamp was always
marked as found and only the first header was checked.
In the case where the timestamp was the first header, things
would have worked as expected.
Thanks to pino for reporting via bug refs #818
Rewinding the ring buffer completely causes audible issues with DMAs.
Previous solution didn't work with tsched=0, and used tsched_watermark
for guardband, which isn't linked to hardware and could become really high
if underflows occurred.
Added separate parameter that can be tuned to hardware limitations and size
of DMA bursts.
We need to use pa_memblockq_pop_missing() for all request handling,
including the initial request, because otherwise the counters will be
stay off during the entire runtime.
This should fix:
https://bugzilla.redhat.com/show_bug.cgi?id=559467
This should make it unlikely that we loop on SIGHUP indefinitely.
Also, this makes it possible for callbacks not to process all events and
still not busy loop.
We need to resume audio devices even for streams that are created in
corked stat, so that the latency ranges of the audio device are known
during the initial latency negotiation. If we don't the latency
negotiation will be based on placeholder data and changed later on which
clients do not expect.
This should fix issues with Skype.
https://bugzilla.redhat.com/show_bug.cgi?id=554929
