Please find below a brief summary of the progress Collabora have
made last week on Pipewire and Weston/Wayland.
PipeWire
George:
- Continued refactoring, polishing and documenting WirePlumber
APIs:
- refactored WpSession and WpEndpoint to be WpProxy subclasses
for efficiency and simplicity
- refactored the object manager and the whole code around
registry handling
so that locally created proxies are associated with global
IDs, improving memory efficiency
and reducing protocol round-trips
Julian:
- Implemented wireplumber logic to create and link the new
pipewire audio-fade-source control
node with the new control port of pipewire audioconvert. This
allows wireplumber to send control
properties to a stream and so be able to do fade in/out audio
effects.
- Implemented wireplumber policy logic to send fade in/out
properties when creating/destroying
endpoint links.
- Fixed wireplumber config policy unit tests to work with new
fade in/out policy logic.
Weston / Wayland
Marius:
- whilst testing output transform (as the default AGL images
comes with a now 90 rotation)
noticed another qtwayland problem, somehow similary to the
previous qtwayland in which some of the
panels weren't displayed. This time, the client acks a
configure event but attempts to commit a buffer
with invalid width/height. While digging a bit the qtwayland
code I've discovered that it uses the v6 unstable
XDG shell and I can't force to use the stable one though the
release notes of qtwayland says otherwise (the
main interface used to bind to is wrongly used in the code).
- posted a series of patches for agl-compositor to update to
libweston7 (for the next branch on AGL).
This took more than expected due to the fact that the agl-poky
configuration removes the x11 distro
feature keyword, combined with the fact that the weston recipe
from yocto will now build weston and libweston
differently based on what distro features have been passed.
agl-compositor has, besides a wayland and DRM
backend, a x11 backend, which I believe could be really useful
for local development. Meson is
a bit clunky on this side as well such that it makes a bit
things tricky to determine if the libweston7
has or not the x11 backend headers installed.
- As there are now two issues related to qtwayland I've
re-tested the output transform and panel display
issues using the next version of qtwayland from the next
branch, specifically 5.13, and I can confirm
that those problems are fixed under the newer qtwayland
version.
(This obviously tests the update to libweston7 as well).
Kind regards,
Nick