https://bugs.kde.org/show_bug.cgi?id=520834

--- Comment #3 from adam <[email protected]> ---
(In reply to David Edmundson from comment #2)
> FWIW we don't release the -X versions, only 6.6.5 
> 
> https://gitlab.archlinux.org/archlinux/packaging/packages/kwin/-/commit/
> ed1cf16cb9f2a1723ac12f564770a4de8b6062f9
> 
> This is the only change, which won't be causing a change in how sockets are
> handled. Either a 3rd party package changed or things are just racey in your
> scripts.

Thanks for taking a look, David. That is a massive clue. 

You are completely right. Looking at that Arch GitLab commit, pkgrel -3 was
strictly a rebuild for Arch's distribution-wide `gcc-libs` deprecation.
Upstream KWin wasn't touched at all. 

Given that it was just compiled against a newer toolchain, the new binary
likely executes just slightly faster, which perfectly explains why it suddenly
exposed a race condition in my setup. My `autologin.service` explicitly uses
`DefaultDependencies=no` to bypass `basic.target` and
`systemd-user-sessions.service` to shave off boot time. 

KWin is clearly initializing before the system has fully prepared the D-Bus or
XWayland sockets that `bwrap` and `umu-run` rely on to spawn their containers. 

I will remove the `IgnorePkg` lock, update back to -3, and fix the legwork on
my end by either restoring the standard systemd dependencies or injecting a
strict socket wait condition before `kwin_wayland_wrapper` fires.

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to