44b3aef4

By: Ammar <ammar.nanjiani@gmail.com>

fix(steammachine): suppress drkonqi crash processor to stop a crash loop

Observed on first boot: over 15,000 coredumps at roughly 10 per second,
all of them the same binary,
/nix/store/...-drkonqi-6.7.4/libexec/.drkonqi-coredump-launcher-wrapped,
SIGABRT.

The drkonqi package ships drkonqi-coredump-processor@.service with
`WantedBy=systemd-coredump@.service`, so every crash pulls it in. It
expects a KDE Plasma session socket to report to, but Gaming Mode is a
gamescope session, so the socket never connects:

  socket state unexpectedly QLocalSocket::UnconnectedState
  aborting crash processing

It aborts, and its own SIGABRT is a coredump, which pulls it in again.

systemctl mask cannot fix this because NixOS owns /etc/systemd/system for
unit files and the path is already a store symlink. suppressedSystemUnits
is the declarative equivalent; nixpkgs uses it the same way for
getty@.service in kmscon.nix.

Also merges the three flat `systemd.*` assignments into one block, which
statix requires (the repo's own convention against repeated keys).

Verified on the running system: the rate fell from ~10/s to ~6/min after a
runtime-level mitigation, confirming the diagnosis, but only a config
change removes it.

Total

12m43s

Timeline