Thanks, the four points split into test-harness bugs and one test premise that I over-claimed. The series is fixed locally; details below.
1. SKIP was swallowed cleanup.t now emits a TAP "# SKIP" line when the helper logs SKIP, instead of reporting a plain "ok". The helper also treats a missing tracefs, a non-writable tracefs, or a missing CONFIG_OSNOISE_TRACER as SKIP rather than a failure. 2. SIGINT race and leaked process The repeated-interrupt case now waits for the tracefs instance to appear before sending SIGINT, and verifies that it disappears after each run. The SIGINT case also reaps the background rtla when the instance never appeared. The helper is run under the suite timeout, so a stuck wait cannot hang forever. I also added SIGTERM to set_signals()/unset_signals() in common.c. Before this, rtla only handled SIGINT and SIGALRM, so SIGTERM left the instance behind. The test now checks that path instead of recording it as a known issue. 3. Missing exit-code checks / action case The self-timeout case now requires exit 0, and the action case requires exit 2 (the mismatched-action result). The action case uses -S, the stop-total threshold, and skips if that threshold was not crossed. An idle machine may not generate a sample above 1 us; that is a property of the workload, not a cleanup result. 4. Unused TIMEOUT / reset_osnoise / dead code run_cleanup_cases() now calls reset_osnoise and uses $TIMEOUT. The unused check_instances_absent() helper is gone. The series is based on 2f2f7eeef8b0a1a9b9ca998231419873ccae8d4e and is commit 455d304ece2a938d0148a908c83f1774d4dbf298. I ran bash -n and checkpatch.pl locally. The host has no writable tracefs, so the suite reports SKIP here rather than a real pass. The patch is attached. Thanks, Chaosheng
