mirror of
https://github.com/AstronautCheetah/frame-tweaks.git
synced 2026-10-17 18:00:18 +02:00
AstronautCheetah
aeb23572d8
dev/run-tests.py: - Ctrl+C killed the parts running, then every part still queued started, one after another, silently (the pool waited for them on the way out): none starts now, those running end. - Putting the headset on moved each test process's main thread only (sched_setaffinity sets one thread): a Qt test's other threads, and what they started, stayed on every core. Every thread is pinned, every second while it's worn; a sensor read that fails keeps what it said last (it was "off": every core), and --pin pins when unsure. - A part that exited 0 without its results counted as passed; output that isn't UTF-8, or a signal Python has no name for, crashed the report. - FRAME_TWEAKS_TESTS: another tests folder (for its own tests). dev/check.sh: a FRAME_TWEAKS_CHECKS it doesn't know stops the tweak check, saying so (hidden, it ran on every core with the headset on); a file moved out of packs/ counts (--no-renames). dev/test-tweaks.sh: a status.sh that broke (127, a missing command) failed only in the window; .venv and caches aren't copied (a contributor's PyQt6, 280 MB, into /tmp each run). dev/fresh-install.sh: --from checkout with VERSION X.Y.Z-rc.N made a tag get.sh didn't take; the bwrap info file, whose pid is killed on the way out, was in a folder the sandbox can write; the control groups are read-only (it could freeze or kill your services), /dev/shm and /var/tmp its own, and no X11 with Wayland there (an X program can type into the other X windows). release.yml: a test release's number comes after the highest, not the count (one deleted gave a number that exists). dev/safe-push.sh says where its replay is when the push is refused.