Age | Commit message (Collapse) | Author | Files | Lines |
|
Reasoning:
- the profile-graphical-html5 belongs in the main layer
- meta-pipewire is enabled by default
For reference - the new location is:
https://gerrit.automotivelinux.org/gerrit/gitweb?p=AGL/meta-agl.git;a=summary
Bug-AGL: SPEC-3453
Signed-off-by: Jan-Simon Möller <jsmoeller@linuxfoundation.org>
Change-Id: I0d60f37f1303be97b0eaf2274b217473d4a1ec5d
|
|
Bug-AGL: SPEC-3098
Change-Id: Id304bf6d1adcf7cd739b705a23772c25ce5258e8
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|
|
* Refreshed & combined some patches
Bug-AGL: SPEC-2837
Change-Id: Ibd5849eff26bb51eb68580eac144aa07e646d127
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|
|
* Refresh patches
* Remove endpoint API patches, which have been merged
* Remove log timestamp patch, which has been merged
* Added a patch to module-access to avoid all those
false-positive security errors in the journal
* Update URLs to point to gitlab.freedesktop.org now
that the project has moved
* Switch from gitsm to git, since the project no longer
makes use of submodules
Bug-AGL: SPEC-2837
Change-Id: I53ef9548e48827b00595162c0a30e12b302eefd9
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|
|
Update patches:
* remove merged patches
* remove the alsa algorithm correction that was needed for qemu,
as it seems to cause severe underruns on other platforms now;
I will recheck qemu to see if this is still an issue
* add a workaround needed to get the audioconvert plugin running
in merge+split mode (needed to get all the channels out instead
of just the front left channel that we were getting before)
Update version to 0.2.91 to reflect the fact that we are now tracking
the master branch instead of the work branch.
Enable the spa audiomixer plugin which is now required to mix audio on ports
Bug-AGL: SPEC-2837
Change-Id: I2558aa5487b9c9918e077bf450230c143abf7e6c
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|
|
Sometimes we get an error from sendmsg when too many objects
exist in the graph and when this happens, clients start to
silently fail to do things in a weird way. These patches fix
that by handling the situation where sendmsg returns EAGAIN
and trying again when the socket is unblocked. Previously,
data would silently be dropped, which is what caused the weird
behaviors.
See also https://github.com/PipeWire/pipewire/issues/111
Bug-AGL: SPEC-2837
Change-Id: Ie30083545629114f10a28e628f54d85e22d13058
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|
|
This refreshes all the patches, removing all those that made it upstream,
redoing the endpoint extension (now called session-manager extension),
and adding some more last moment fixes.
In addition, the configuration files for wireplumber & pipewire
are being updated to load the new modules, as the module set
has changed in both daemons.
Finally, the pipewire recipe is adding PACKAGECONFIG options
for jack and vulkan, so that we can actually disable them.
Pipewire upstream builds them by default and we don't want that.
Bug-AGL: SPEC-2837
Change-Id: Id42119c027558466f0a0aa71813ff15f33dfcb56
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|
|
Bug-AGL: SPEC-2792
Change-Id: Ida682a405c4cc5d2f84a98cb71f89a7bb41ad489
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
(cherry picked from commit c0eaa62e6fafd0b4cc057c0089ff110426bfde25)
|
|
In pipewire we only enable that if the bluez5 feature is enabled
because otherwise the module is not compiled and will error out
at runtime.
In wireplumber we always enable it since it does not depend on bluez.
Bug-AGL: SPEC-2792
Change-Id: I53bbc387bccd953633affd4fc34949a2fd488ed7
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
(cherry picked from commit 9fc6d8e9bbc9746777c03471b643992abcb182a5)
|
|
Bug-AGL: SPEC-2674
Change-Id: Id5b099a8efed2a2f9bdae142ef69bbc09deae8e8
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|
|
pipewire
Bug-AGL: SPEC-2634
Change-Id: I10301e0c244fad60b31a4dfa6dc0dc61512a4867
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|
|
0004 adds a useful cli utility
0006 aids debugging
0007 fixes audio in qemu (SPEC-2570)
0008 & 0009 add functionality necessary for implementing multiple
streams per device
The rest is the same as before
Bug-AGL: SPEC-2570
Change-Id: If6736f0fc87491edce6981c0be44094239ddaa89
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|
|
DISABLE_RTKIT saves us from nasty error messages about not being able
to connect to dbus. This doesn't make sense anyway, real-time mode
should not be enabled on embedded.
PIPEWIRE_DEBUG/G_MESSAGES_DEBUG are put there in comments so that
people can easily uncomment to figure out what is wrong with their
pipewire setup.
Bug-AGL: SPEC-2473
Change-Id: Id07d3d21f472885434158c300a227d53020c2189
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|
|
Pipewire shares memory with its clients using open file
descriptors (memfd or shared memory file) which are created within
pipewire and therefore they have the same smack label as the
pipewire process. Clients must be able to read and write to this
memory, therefore they need rw acess to that smack label.
Since all AGL apps have only write access to the System label,
we need to use a different smack label which can be granted rw
access from the applications that need to use audio.
"System::Pipewire" is chosen here to follow the "System::Sub"
pattern that is described in the documentation.
Bug-AGL: SPEC-2554
Change-Id: I81cbf82adfde3ef4f67872bd91293370339b18d7
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|
|
Bug-AGL: SPEC-2393
Change-Id: I2e4feef547642c2b3d34403f7a1a45999cd4db11
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|
|
This allows granting pipewire additional permissions, most notably
the permission to access the alsa devices (SupplementaryGroups=audio)
and therefore fixes pipewire for the "running as non-root" setup
Bug-AGL: SPEC-2554
Change-Id: Ie9192a7f42f4929f70114ddc39a23d94f9ba84fc
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|
|
Add DISTRO_FEATURES based enabling of bluez5 support to pipewire recipe
so the Bluetooth plugin is enabled for experimentation.
Bug-AGL: SPEC-2473
Change-Id: Ic63df580c4005e924a2172d2d513dbdca94457f4
Signed-off-by: Scott Murray <scott.murray@konsulko.com>
|
|
Add a bbappend to the pipewire recipe to create the required symlink
in /etc/systemd/user/sockets.target.wants to enable per-user socket
activation. A bbappend is used as this change is likely to be AGL
specific (for now?).
Bug-AGL: SPEC-2521
Change-Id: I4b349ab8151eabfb933f2dcfcf9d293fcb3ea321
Signed-off-by: Scott Murray <scott.murray@konsulko.com>
|
|
We were missing the
LIC_FILES_CHKSUM = "file:///${COMMON_LICENSE_DIR}/MIT;md5=0835ade698e0bcf8506ecda2f7b4f302"
Bug-AGL: SPEC-2473
Change-Id: Ic7751ceb08fef1f19426e66c850740abf653fef7
Signed-off-by: Jan-Simon Möller <jsmoeller@linuxfoundation.org>
|
|
This does enhance the recipes for pipewire and wireplumber in 2 ways:
a) we add a separate package for the configuration and make it replaceable
w/o recompiling pipewire or wireplumber itself
b) this in turn makes the recipes not AGL specific and upstreamable as-is.
v2: enhanced config recipes
v3 & v4: fix typo
Bug-AGL: SPEC-2473
Change-Id: I1f6f32d7cc64d9424706ca6bc87550f5e2c6d359
Signed-off-by: Jan-Simon Möller <jsmoeller@linuxfoundation.org>
|
|
Bug-AGL: SPEC-2473
Change-Id: Ide95be79adb69437564e94071d0315ae5cfd9ae2
Signed-off-by: George Kiagiadakis <george.kiagiadakis@collabora.com>
|