Previously, we were always looking for the first available link on the item,
which could lead to problems if the item had more than 1 link (eg target and
monitor links). Now, we keep the Id of the last target in order to determine if
the item is linked to the proper target when rescanning.
When node.target is set to contain a node's name, it is possible to run
into the situation where you have a sink and a source with the same name
(typically the case with module-loopback) and the sink has monitor ports,
so wireplumber may link the stream to the monitor of the sink instead of the
capture ports of the source
With this policy, if a stream really wants to link to the monitor of the sink,
it has to have the "stream.capture.sink" property set or specify the
target by id.
If one link fails, the activation transition will return, but then
other links will continue to call the callback and try to access
the now invalid activation transition. With this change, the callback
is bound to the lifetime of the transition and will stop being called
after the transition returns
Fixes#76
There is a chance of race condition here in case the client knows which target
it wants to link to, but wireplumber has not yet "seen" this target because
it is getting prepared asynchronously
This may stall for up to 2 seconds clients that start streams with an invalid
node.target property, but then they will be re-routed to the default node
If the item was not already handled or if there is no target,
we don't need to search for existing links... we will not find any
Also, don't remove links or schedule rescan if the node is not eligible
for reconnection... we will delete the node below instead
* First of all, do not clear the stack... it's not necessary to do
so and it causes trouble if this function is called from within
itself, as it clears the stack of the previous context
* Then, do not stop/restart the garbage collector unless this is
the base call
Fixes#73
This is now done by the metadata implementation.
Doing this manually causes assertion failures, because then the
target.node is removed before policy-node "unhandles" the associated
session item and it tries to re-link it to the default source/sink
Now that the build directory is found automatically, if it so
happens that you have more than one build dir in your tree,
'make run' may not run the version that you expect...
./builddir/wp-uninstalled.sh can now be called without a -b argument,
it'll automatically set the full paths. This is useful when building as
a subproject.
meson has a limitation: the configure_file() object is just a file
wrapper object. Wrap this into a custom_target() whose return value has
a full_path() method. This way, pipewire can use
wireplumber_uninstalled.full_path() without having to build the
directory structure itself.
AUX channels can link to any other channel but when AUX channels are
linked, they must match exactly.
Fixes some issues with split devices with aux channels.
"unpositioned" nodes don't have a specific channel layout
(ex the Pro Audio profile); in that case, we should not try to copy this
undefined layout to the peer node, we need to configure both in their
preferred layout and do a best-effort linking of ports.
-some modules will not be available during runtime, due to
external dependencies.
-these modules can now be marked optional.
-loading of these optional modules will be attempted and if
they are not available, wp will recover and will not be
terminated.
