Let's add PipeWire and Socket into the description, the current messages
are not overly obvious:
Listening on Multimedia System.
Closing Multimedia System.
Where pipewire and friends are started via socket activation,
a PA client connection to the pulse socket triggers
pipewire-pulse.service. That in turn triggers pipewire.service through
Wants/After and once that is started up, pipewire-pulse actually
starts up.
At the same time, pipewire-media-session or wireplumber are started
through WantedBy/after in the respective service files.
Depending on which leg the race condition got out of bed with today,
pipewire-pulse may be finished before the session manager has set up the
graph and the PA client doesn't see any devices.
Fix this by adding a dependency on the session manager in
pipewire-pulse, installed via an Alias so media-session and wireplumber
can install the same alias. Wants is a light dependency, so for the case
where it doesn't exist we fall back to the current behavior anyway.
This doesn't remove the race condition since systemd may deem the
session manager to have started before the graph is set up, but it
should reduce the occurances.
The disadvantage here: only one Alias will be installed by systemd, so
first-come, first-serve in the case of both media-session and
wireplumber being installed.
See #1553
Without this, a journal entry merely looks like this:
Stopping Multimedia Service ...
Which is obviously terrible from a branding perspective but also makes
it harder for users to figure out what process was responsible for this
entry.
Some distributions set --auto_features=enabled which messes with the
internal logic of the build system when features are used for other
purposes than pure dependency control. The only solution is to either
avoid the value auto or change the type of the option to non-feature.
This commit does the later by replacing -Dmedia-session, -Dwireplumber
and -Dsession-manager with the new -Dsession-managers array and
-Ddefault-session-manager combo options.
Fixes#1333Fixes#1336
You can only run either pulseaudio *or* the pipewire-pulse daemon at one
time. So make the pipewire-pulse units conflict with their pulseaudio
counterparts to avoid both running at the same time
Use the default backlog size, SOMAXCONN (128), which is the same as for
the unix socket in module-protocol-native. If the backlog queue gets
full incoming connections will be closed before they can be accepted.
This causes connect() to fail with EAGAIN on the client side and since
the server socket is closed it will not be possible to use the client
socket further.
/run/pipewire is initially created by pipewire.socket when it creates
the socket file, but it is owned by root with mode 0755.
In the .service unit, RuntimeDirectory=pipewire re-creates /run/pipewire
for user pipewire (owned by pipewire:pipewire, 0755) when this unit starts.
This is necessary for the pipewire daemon to be able to write the .lock file.
By default, when the service unit stops, systemd deletes this runtime
directory and therefore it also deletes the socket that is owned by
pipewire.socket. When this happens, the service is no longer
socket-activateable even though pipewire.socket is active.
RuntimeDirectoryPreserve=yes prevents the service unit from removing
this runtime directory and socket activation still works as expected
Add a new build option 'system-user-unit=dir', which can be used to
override the pkg-config default value.
By doing this, it becomes more practical to make a local unprivileged
build of the latest PipeWire:
$ meson --prefix=~/.local build
$ meson configure build -Dpipewire-alsa=false
$ meson configure build -Dsystemd-user-unit-dir=~/.config/systemd/user
$ meson configure build -Dudevrulesdir="`mktemp -d`"
$ ninja -C build install
For a local build enabling ALSA plugins is not plausible.
JACK needs a simple addition to the ~/.profile:
export LD_LIBRARY_PATH=$HOME/.local/lib/x86_64-linux-gnu/pipewire-0.3/jack
Finally, the genuine PulseAudio must be masked and local services
enabled:
systemctl --user enable pipewire.socket
systemctl --user enable pipewire-pulse.socket
systemctl --user mask pulseaudio.service
A sanity check [*] shows that nothing has leaked out of the home
directory (even though at least on my Debian 10 system PolKit still
asks for authorization during the meson build).
Not perfect, but lowers a barrier to run the development version a lot,
as that does not anymore to trash your system.
[*] find / -name "*pipewire*" \! -path "/home/*" \! -path "/var/lib/flatpak/*" 2> /dev/null
Signed-off-by: Jarkko Sakkinen <jarkko@suppilovahvero.lan>
Long story short, this is something pulseaudio has since 2017 [1]. If
pipewire-pulse is aiming to be a drop-in replacement, the same behaviour
is probably expected.
I noticed that without this line I hit some issues with systemd socket
activation. In fact, it does not work properly at all. It probably has
something to do with the fact that pipewire-pulse is started for GDM
session and is somehow propagated down to my own user session.
[1] 24928d6b6f
Add a small application that loads the pulse server and makes it
listen on the pulseaudio socket.
Add some systemd activation files to start the service.
Add system service and socket in addition to the user service and
socket. The system service is started as pipewire user and group and
clients need to be in the pipewire group to be able to connect.
On NixOS all packages are installed into their own immutable prefix.
Because of this get_pkgconfig_variable will return a
path from within systemd's prefix and we cannot write to it.
By using define_variable we can replace the respective directory
to be from the paths from meson.
Don't recurse into the directory of systemd when it was not found.
There are still some issues because libudev is actually provided
by systemd these days and we still need that for building the v4l2 and
alsa monitors.
Remove spec file
Fix systemd service file
Send error on link errors
Init socket to -1 so that we don't close anything on error
Wait for all proxy objects in device monitor
