From c1cc98ad5a7e612b274ac5c32d9da7bd6d3337bb Mon Sep 17 00:00:00 2001 From: Levi Neuwirth Date: Tue, 18 Aug 2026 14:19:45 +0200 Subject: [PATCH] docs(lane): retract the "mechanism located" claim --- it was unsupported a92ef7f said the mechanism was located: the launcher blocked in `do_wait` on a GPU probe child stuck in `futex_do_wait`. Checking the instrument against the claim shows it does not support it. The sampler caught 394 distinct launchers across a reproducing sweep and the longest-lived was 5s TOTAL. For this test to fail, a launcher must outlive its SIGINT by 5s, so its lifetime would be 8s or more. The failing instance was never captured. What I described is a healthy launcher from one of the suite's other tests --- the normal teardown shape, reported as the defect. This is the same error as the `available`-memory reading earlier in this lane: a measurement that looked conclusive, reported before checking that it discriminated. Retracted here rather than left to be found. Two facts do survive and constrain the next attempt: `pmacs-gpu` installs no signal handling at all, so the probe should die on SIGINT's default action; and its main loop is a 50ms `recv_timeout`, so it never blocks indefinitely. The next instrument must key on the failing instance --- launchers outliving ~6s, or a PID recorded by the test itself --- rather than sampling every launcher and hoping. Co-Authored-By: Claude Opus 5 (1M context) Claude-Session: https://claude.ai/code/session_016bqGA6s9tTUFzYpbeW3tai --- docs/active-work.md | 30 +++++++++++++++++++++--------- 1 file changed, 21 insertions(+), 9 deletions(-) diff --git a/docs/active-work.md b/docs/active-work.md index a684772..62843ca 100644 --- a/docs/active-work.md +++ b/docs/active-work.md @@ -478,21 +478,33 @@ from #171 and #215. `ctrl_c_on_launcher_group_does_not_reach_spawned_daemon`. **This branch is not implicated**, and no branch can pass this gate stage on this machine until the underlying defect is fixed. - - **MECHANISM LOCATED, by sampling the process table twice a second - through a reproducing sweep.** After the test SIGINTs the - launcher's process group: + - **MECHANISM STILL NOT ESTABLISHED. The "located" claim below was + RETRACTED on its own evidence** — read this before trusting it. + A process-table sampler ran through a reproducing sweep and caught + 394 distinct launchers, but the **longest-lived was 5s total**. For + this test to fail, a launcher must outlive its SIGINT by 5s, so its + total lifetime would be 8s or more. **The failing instance was + therefore never captured**, and the `do_wait`/`futex_do_wait` pair + below is a healthy launcher from one of the suite's other tests, + not the one that blew the deadline. The observation is kept only + as a description of the NORMAL teardown shape. + - Provisional and unconfirmed, from that sampler: - `pmacs --gpu --socket …` (the launcher) sits in **`do_wait`** for the full 5s — it is waiting on a child, not ignoring the signal; - `pmacs-gpu --headless-managed-probe …` (its child, same process group, so it received the SIGINT too) sits in **`futex_do_wait`** and never exits. - So the deadline is missed because the GPU probe child does not tear - down under SIGINT, and the launcher blocks on it. **The next step - is the probe's shutdown path**, not the test's timeout — raising - the 5s would only hide it. Why the child hangs *only* in a complete - sweep is still open; every prior GPU suite in the run has already - exercised the adapter by then, which is where to look first. + That is what a HEALTHY teardown looks like; it is not evidence + about the failing one. Two facts do survive and constrain the next + attempt: `pmacs-gpu` installs **no signal handling whatsoever** + (grep for SIGINT/signal_hook/sigaction across `pmacs-gpu/src` + returns nothing), so the probe should die on SIGINT's default + action; and the probe's main loop is a 50ms `recv_timeout`, so it + is never blocked indefinitely. **The next instrument must key on + the failing instance specifically** — sample only launchers whose + lifetime exceeds ~6s, or have the test itself record the PID and + timing — rather than sampling every launcher and hoping. - **Five further explanations were tested and REFUTED.** Recorded so nobody re-runs them: - *Machine load*: refuted. Red on a quiet machine (load 2.77 at