You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[BUG] Platform assumptions: cgroup v1/hybrid and non-systemd session managers #156
Two latent platform assumptions, same class as the recent #153 (desktop Exec quoting vs SDDM/plasmalogin) and #154 (SELinux label read as a snap id). Neither is fatal everywhere, but both degrade behaviour on a broad set of systems. Filing as a checklist.
1. cgroup v1 / hybrid (matching process signals)
Where:src/embedder/meta_window_state/process_info.rs → get_cgroup_from_pid, consumed by the matcher's process-grouping / launch-provenance signals.
What: it only recognises the cgroup v2 unified line (0::<path>); any other layout returns null.
Impact: on cgroup v1 / hybrid systems (RHEL 8, Ubuntu <= 20.04, some container setups) ProcessInfo::cgroup is null, so the app-*.scope / same-cgroup signals disappear and sequential/parallel launches fall back — the same "new tile instead of the placeholder" symptom, from a different cause.
Task:
fall back to the v1 name=systemd: line and normalise both to the same path form
unit test the v1, v2 and hybrid fixtures
2. Non-systemd session managers
Where:extra/assets/veshell-session.
What: it only knows systemctl --user and dinitctl; otherwise it prints "No systemd or dinit detected" and exits.
add launchers for the common managers, or make the fallback run veshell --session directly (keeping the stale-state recovery where the manager allows it)
document the manual veshell --session path for the rest
Context
Same class as the two just fixed: #153 (SDDM/plasmalogin hand the raw Exec to exec $@) and #154 (/proc/<pid>/attr/current assumed to be an AppArmor snap label, which on SELinux is every process).
Activity