Dont-reconnect streams should not be moved to a new target, even though
other streams would move.
Pulseaudio DONT_MOVE works like this, and also pipewire-media-session
worked like this.
When device nodes go away, we should remove links and peer_id references
concerning them, so that we don't leave behind stale links and stale id
entries in the peer_id tables.
Also fix bug in si_flags[out_id] handling in unhandleLinkable.
Policies need to know if there are session items that are pending
activation. Linkables are not activated in the same order as nodes
appear, which causes problems for e.g. resolving target nodes, if some
of the linkables are pending.
Register linkables in create-items before they are activated. When
activation completes, remove those that did not activate successfully.
Policies can filter out inactive items by tracking active-features
flags.
If there are existing linkables that are not ready, suspend policy-node
processing, and continue it only after all linkables are ready.
This makes the configuration more consistent. The storing of the default routes
logic needs to be implemented eventually in a C module-default-routes, similar
to module-default-profile.
Persistent profiles will never change. This can be useful if we want to keep
a specific device profile even if a new one with higher priority becomes
available. For example, keep the 'off' profile instead of switching to best
profile when monitor screens is resumed.
See #138
You can now override:
* Whether to store/restore properties like volume
* Whether to store/restore the target.node
* Things like media.name/application.name/etc which affect which
entry is going to be used in the state file
Related to #169
If a stream audio node does not have the 'media.role' property set, the policy
will asign the 'Default' media role name to it. In addition to this, if the
'Default' endpoint can not be found, the policy will link the stream audio node
with the lowest priority endpoint.
Video nodes will always be handled by policy-node.lua, and they will never be
handled by policy-endpoint-client.lua
This allows users to assign a particular endpoint for streams that don't have
the 'media.role' property set.
Only the configured headset profile needs to be persistent, the switched
status and saved profile should not persist over wireplumber restart.
When devices appear initially, they should not appear as switched,
but recheck switch status.
Monitor whether streams are running or suspended. When they are
suspended, switch to normal mode.
This is required e.g. for programs such as Teams, which keep recording
streams open but inactive for their whole runtime.
Profile restore also needs to be done with a timeout, to avoid switching
rapidly when streams are moved between sources.
Use Lua tables properly as maps.
Headset profile saving does not work properly currently (only called on
metadata change; if headset-profile == current-profile, it doesn't get
saved etc.)
Change it to follow the logic:
- Each device has a "normal" and "switched" mode. The latter becomes
active if a "communication" input stream exists and the current
default sink is bluez sink. The switched mode is exited only when
there are no "communication" input streams.
- On transition normal->switched, record current profile as the "saved"
profile. Then switch to "headset" profile. If no headset profile
was saved previously, pick profile with input direction & highest
priority.
- On transition switched->normal, save currently active profile as the
"headset" profile. Then switch back to "saved" profile.
Monitor streams can be identified based on the stream.monitor property.
They should not be identified based on application name, because that
may be localized, and the previous code does not work properly on
non-English locale.
This also improves the status output to display the configured default node
names using the new get-default-configured-node-name API, and refactors
set-default to also use the new set-default-configured-node-name API from
the default-nodes-api module.
Currently, spa-bluez5 devices start in non-off profile, which causes
m-device-activation logic to set the highest priority profile instead of
the saved one.
Fix this by having spa-bluez5 start devices in off profile, similarly as
spa-alsa does.
We now only use module-rt. Explicitly set the nice.level to -11 instead
of leaving the default (do nothing). Adjust the rt limits to their new
defaults in module-rt (unlimited).
We now only use module-rt. Explicitly set the nice.level to -11 instead
of leaving the default (do nothing). Adjust the rt limits to their
defaults in module-rt.
The "saved headset profile" is taken as the profile that was active when
the headset mode deactivates. So in case the user switches profile during
the time when the Communication input stream is active, the selected
profile is remembered next time.
When a stream gets added, we might have a situation where the device
does not yet exist to actually link on the basis of intended role. For
example this can be observed if no mic sources are present and then the
mic appears after a bluetooth profile switch to HFP.
To fix this, iterate over streams when a device connects in addition to
iterating over devices when a stream connects.
The Role traversal of config data to take notice of the
media class as well so that source and sink roles dont
overstep on each other and pick up unexpected endpoints.
This is expected to be printed if a stream opens and closes
very quickly, like in the case of pipewire-alsa opening the device
for querying formats and then closing it immediately.
If the default nodes module is loaded, and the default node is not found, it
means that there is no valid target for that session item. We only need to
manually find a default target if the default nodes module is not loaded.
Fixes the warning:
[string "policy-node.lua"]:117: attempt to index a nil value (field '?')
stack traceback:
[string "policy-node.lua"]:117: in function <[string "policy-node.lua"]:109>
