There was a question in IRC about whether pa_simple_read() blocks or
not. It's already documented on the simple API overview page, but it's
good to say it also in the function reference. As a bonus, I added
some additional details to the documentation too.
The function was redundant, because all it did was call adapter_free()
for each adapter in the hashmap, and that can be delegated to
pa_hashmap when freeing or emptying it.
This fixes a bug where calling time_restart can leave the current event
in the cache, even though the restart scheduled the event in the future.
This would cause the event to get executed more frequently than it should.
Previously module-bluez5-discover and module-bluez4-discover were being
tracked using their pa_module pointer. But during daemon shutdown these
modules are unloaded before module-bluetooth-discover, leaving stale
pointers in module-bluetooth-discover's userdata. To avoid this problem
this commit makes module-bluetooth-discover keep track of
module-bluez5-discover and module-bluez4-discovery by their indexes.
PA_SAMPLE_24NE generated in pa_sndfile_read_sample_spec is not
handled in pa_sndfile_readf and writef function. paplay/parecord
used to get aborted for 24bit depth wav files
module-alsa-{sink,source}.c call pa_alsa_{sink,source}_new with
mapping set to NULL. Guard against this, like the rest of the
function does.
module-alsa-card does not use NULL, so this went unnoticed so far.
When the creation of u->thread fails, then pa_thread_mq_done() in
pa__done() will crash, because pa_thread_mq_init() was never called.
Allocating the thread_mq object separately, instead of embedding it
in the userdata struct, allows pa__done() to call pa_thread_mq_done()
only when necessary.
This is a cleaner solution, because it also removes paths that are
being removed because they are subsets of other paths.
Otherwise, the lingering paths could cause jack detection related
assertion failures.
BugLink: https://bugs.freedesktop.org/show_bug.cgi?id=69676
Reported-and-tested-by: Kalev Lember <kalevlember@gmail.com>
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
The author of this module, Tanu Kaskinen, has said that this module
"is not suitable for general use". Also, it is still causing crashes
on card removal (see bug 69871).
Qpaeq, and possibly other tools, use this module - but they can load
the module manually if they still wish to use it.
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
Create a wrapper module called module-bluetooth-discover to avoid
breaking backward-compatibility of default.pa. This wrapper may
eventually be dropped altoghether with BlueZ 4 support.
For quite some time now the device driver module doesn't work well
without the discovery module, so for the BlueZ 5 support we'll prevent
the device driver module to be loaded if the discovery module is not
loaded.
Create the thread function, the render and push functions for A2DP, the
process message function for communication between the I/O thread and
the main thread, and other helper functions related to them.
Get the remote device information stored in pa_bluetooth_discovery. This
also creates the mandatory parameter 'path' for module-bluez5-device,
which is used to inform the object path of the remote device in BlueZ on
the module load.
bluetoothd always send the GetManagedObjects() reply messages with the
objects array argument following an in-depth order starting from the
root. That means parents will always be known at the time their children
objects are parsed, if clients parse the objects in the same order they
appear in the array, as we do in PulseAudio.
This commit tries to protect PulseAudio in the case bluetoothd changes
that behavior for some reason. It hasn't been tested, since this
situation never occurs.
