Simon Ser @emersion@octodon.social
soju 0.1.0 is released!
In today's episode of "life sucks thanks to Firefox": Firefox assumes the compositor will release shm buffers immediately. Of course this cannot be relied upon, this is completely compositor implementation-defined. KWin doesn't do it for instance.
The Firefox "fix" is to detect KWin via XDG_CURRENT_DESKTOP and blit buffers in this case. Brilliant.
Thoughts on writing a wayland window manager with wlroots -- inclem.net
http://inclem.net/2021/04/17/wayland/writing_a_wayland_compositor_with_wlroots/
The Pixman software renderer has been merged into wlroots!
https://github.com/swaywm/wlroots/pull/2661
There's still a lot of optimization work to be done, and it doesn't yet work with the DRM backend, but it's a good first step :)
First image of glxgears running under KWinFT with its WIP wlroots backend
> Read old wlroots issues
> Find a 3 year old comment of mine
> It asks a question about GBM that doesn't make any sense
What a fool I was.
Status update, April 2021
This is cool because (1) less tricky code to maintain in wlroots + single code-path (2) better support for our non-systemd users (3) we can share all of this goodness with other Wayland compositors
wlroots has dropped all of its session code in favor of libseat!
https://github.com/swaywm/wlroots/pull/2839
Next up is Weston, it got initial support for libseat a few days ago.
Hrm, C++ has made its way into Mesa for some reason. Now compilation is super slow.
river stream is live
FreeBSD 13.0 now available on builds.sr.ht
It directly deals with VkImage (individual frames), seems to be zero-copy, and can integrate well with the existing extensions for DMA-BUFs, format modifiers and explicit synchronization
The new Vulkan video stuff looks cool
https://www.khronos.org/blog/an-introduction-to-vulkan-video
wlroots 0.13.0 and Sway 1.6 are released!
Contributing to libva is… painful. Been trying to convince them over and over again that they should handle format modifiers like all other existing APIs, because people actually took some effort designing these APIs.
But they keep insisting: "we have no clue how modifiers work but they should work like we think they work".
Hrm, Mesa CI has so many false-positives. It's basically a game of submitting MRs to disable tests and trying again until passes.
Oh nice, a new Mesa Vulkan driver is in the works for Mali GPUs.
