v0.2.1

Upgrading from 0.2.0

• cargo installs: rerun rlm guard enable to refresh your user unit (it adds RestartPreventExitStatus=78 75), then systemctl --user restart rlm-guard. Packaged installs get the new unit with the package.
• rlm limit now refuses to run while the config file is invalid, instead of silently using defaults and dropping your protect list. Fix the file, or pass --force.
• Guard config values now have upper bounds: calm_hold_secs and freeze_cooldown_secs at most 86400, mem_available_floor_mb and min_rss_mb at most 16 TiB. A config above them makes rlm-guard exit with status 78 until it is fixed.

Fixed

• rlm-guard never keeps its lock, journal or history in a shared directory such as /tmp. It uses the per-user state directory, then $XDG_RUNTIME_DIR/rlm; with neither, it does not freeze or cap.
• A second rlm-guard exits with status 75 and systemd does not restart it every 2 seconds.
• The cold-start wait for growth data is bounded to 3 samples.
• Startup waits at most 2 seconds for the systemd bus, then restores through cgroupfs directly.
• Guard config arithmetic cannot overflow.
• rlm guard enable: % in the binary path is escaped, paths systemd cannot run are refused with a clear error, relative PATH entries are ignored, the unit is written atomically, and a masked unit is reported with the unmask command.
• GUI: the guard switch gives up after 30 seconds, the Run page removes a leftover cgroup once its processes exit (and stops retrying after a minute of failures), and the service state cache cannot deadlock.
• Packaging: the AUR package stops the guard on removal, the rpm keeps an edited rlm-delegate.conf on upgrade, and the deb has a short synopsis.
• crates.io: README links and the hero image resolve, and rlmctl-core and rlmctl-common have readmes.

Changed

• Common terminal emulators and the tmux and screen multiplexers are on the built-in protect list.
• CI and release workflows use the Node 24 versions of their actions.