This seems to be an omission from when we transferred the default device
properties from the config file on to the Lua script. Both auto-profile
and auto-port are meant to be disabled, so that we can apply our own
management logic.
Fixes: #734
When node activation fails, it won't be created and its object-removed
signal can be called with some properties missing. This breaks node
name deduplication, causing error
wplua: [string "alsa.lua"]:182: attempt to concatenate a nil value (local 'node_name')
It occurs e.g. when switching ALSA device profile, while some
application has opened the device with ALSA and is keeping it busy.
Fix by handling the activation failure, and tolerating missing property
in object-removed.
Cancel the async calls that get the name of the application owning the
service, when WpReserveDevice is finalized or we are going to make
another call.
Fixes UAF accessing self when the async callback runs.
This patch adds a 500ms timeout callback to switch to HSP/HFP when a stream
starts capturing BT audio. This avoids quickly switching from A2DP to HSP/HFP
back and forth if an application just wants to probe the BT source for a short
period of time.
See #634
Some jobs do not make sense to be run on certain cases and we can save
CI cycles that way.
Most build rules can be run only on merge requests and custom branches.
On master, it only makes sense to build fedora_with_docs for the
purpose of deploying updated documentation on gitlab pages.
The analysis steps only make sense to run when the relevant files have
changed.
The actual deduplication only checked the first device. Extend it to the
full list, as intended.
Fixes: 848b6326 (monitor-utils: Rework camera deduplication code)
The previous code made some invalid assumptions and was rather
complicated. Rework and simplify it.
The approach work as follows:
1. Once a new device gets registered, store its data in a list of pending
devices.
2. Start a timer. On timeout, we check all pending devices and depulicate
according to our heuristic. The timer gets reset whenever a device is
added in order to avoid race conditions.
3. On timeout, add the pending devices to categories. For now: UVC cameras
from the V4L2 plugin, libcamera cameras and other cameras from the V4L2
plugin.
4. Then process the different categories in order of preference and store
their V4L2 device IDs in a list for each plugin.
5. Before creating a camera, check that the V4L2 device(s) it uses are not
yet used by a already existing camera from the other plugin.
While on it, drop support for Pipewire versions that don't report V4L2
device IDs at all.
Closes https://gitlab.freedesktop.org/pipewire/wireplumber/-/issues/689
Closes https://gitlab.freedesktop.org/pipewire/wireplumber/-/issues/708
`wp_dbus_connection_disable()` creates a new GCancellable
object at the end, which is never freed if the GObject
is then destroyed. To fix this, override `finalize()` and
clear everything there as well.
Direct leak of 64 byte(s) in 1 object(s) allocated from:
0 0x70e688efd1aa in calloc /usr/src/debug/gcc/gcc/libsanitizer/asan/asan_malloc_linux.cpp:77
1 0x70e6874b3e62 in g_malloc0 (/usr/lib/libglib-2.0.so.0+0x63e62) (BuildId: 7b781c8d1a6e2161838c5d8f3bd797797c132753)
2 0x70e6875dea75 in g_type_create_instance (/usr/lib/libgobject-2.0.so.0+0x3ea75) (BuildId: 5af5e0f7d0a900ecb6083fbd71e22e5522d872e2)
3 0x70e6875c3804 (/usr/lib/libgobject-2.0.so.0+0x23804) (BuildId: 5af5e0f7d0a900ecb6083fbd71e22e5522d872e2)
4 0x70e6875c4e7e in g_object_new_with_properties (/usr/lib/libgobject-2.0.so.0+0x24e7e) (BuildId: 5af5e0f7d0a900ecb6083fbd71e22e5522d872e2)
5 0x70e6875c5ed1 in g_object_new (/usr/lib/libgobject-2.0.so.0+0x25ed1) (BuildId: 5af5e0f7d0a900ecb6083fbd71e22e5522d872e2)
6 0x70e684d2a8a6 in wp_dbus_connection_disable ../subprojects/wireplumber/modules/module-dbus-connection.c:173
7 0x70e688a833cc in wp_plugin_deactivate ../subprojects/wireplumber/lib/wp/plugin.c:144
8 0x70e688a7126c in wp_object_deactivate ../subprojects/wireplumber/lib/wp/object.c:542
9 0x70e688a6e74e in wp_object_dispose ../subprojects/wireplumber/lib/wp/object.c:191
10 0x70e6875c0f6c in g_object_unref (/usr/lib/libgobject-2.0.so.0+0x20f6c) (BuildId: 5af5e0f7d0a900ecb6083fbd71e22e5522d872e2)
11 0x70e6841f7d6d in wp_portal_permissionstore_plugin_disable ../subprojects/wireplumber/modules/module-portal-permissionstore.c:207
12 0x70e688a833cc in wp_plugin_deactivate ../subprojects/wireplumber/lib/wp/plugin.c:144
[...]
Handle only devices associated with Bluetooth loopback nodes.
Make sure the node.link-group iteration cannot get stuck if there is a
loop in the link graph.
Revamp the profiles section, making use of the inherits feature
and add commonly used profiles for systemwide & embedded use cases,
as well as profiles for a split-instance configuration where the
policy hooks run in a separate instance from the alsa, bluetooth
and camera monitors (which run in 3 separate instances respectively)
Also add an example on how to configure the loaded profile using
a config file instead of the CLI switch.
Fixes: #608
This allows to inherit all the profile definitions of another profile
before the current profile's definitions are parsed, allowing for
more complex structures to be present in the default wireplumber.conf
without too much copy-paste
In some cases, requires/wants dependencies are not enough. As we saw in
!617, the m-standard-event-source module needs to be loaded after all
the hooks, otherwise there may be missed events that the hook was
supposed to "catch", but they were delivered before the hook was actually
loaded. In a similar fashion, we have in purpose put all the "monitor"
components at the every end of the array because if we load them earlier,
they will create devices and nodes before all the hooks are in place to
react.
While in standard configuration we can work this around, in extended
user configurations with custom components, it is impossible to do this
without overriding the entire components array.
To fix this properly, introduce before/after dependencies. They work in
a similar fashion as they work with event hooks. They do not implicitly
"pull" any components to be loaded, but they affect the ordering if the
mentioned components are indeed present.
Note that for backwards compatibility reasons and unlike systemd units,
the "requires"/"wants" targets imply an "after" dependency on them.
Fixes: #600
This allows changing the metadata name in the configuration file
and get all the Lua scripts looking at this new settings metadata
without requiring any other code changes. It can be useful in
multi-instance setups where we'd like the instances to be isolated
and load their own settings instead of relying on each other for that.
This patch removes the manually created object managers in the BT profile
autoswitch logic, and instead uses the object managers from the event source,
simplifying the logic a bit.
The intention here is to fix a Bluetooth bug where changing between different
profiles momentarily causes streams to be linked to an internal speaker.
Unfortunately, this case is not handled by the event dispatcher because some
signals are emitted using idle callbacks, and therefore we have to rely on
extra logic in rescan.lua to solve the problem.
This reverts commit 8012fbd5cf.
There seems to be some underlying bug that sometimes causes a crash
when the Bluetooth monitor uses a loopback for autoswitching between
profiles. Reverting temporarily until the cause is fully determined.
See #682
The previous assumption that any target with "device.intended-roles"
should be managed by the role-based policy is wrong, as for example
Bluetooth SCO nodes always have "device.intended-roles = Communication"
and some ALSA devices managed by ACP also do. This is meant to be used
as a hint for the desktop policy (it's been there in PulseAudio as well)
and does not necessarily mean that a role-priority-based policy should
be applied on all links to those devices.
Instead, use a new property to explicitly mark all the targets that
are meant to be managed by the role-based policy and respect that in
all places where we check for a potential role-based policy link.
Fixes: #682
This way of definining dependencies ensures that if we remove one
of the find-* hooks from the config, the rest of them will continue
to work in the expected order. Previously, removing one of them
would break the entire chain.
