docs(framing): open the GPU probe SIGINT teardown lane

`ctrl_c_on_launcher_group_does_not_reach_spawned_daemon` fails in gate
stage `sweep-crdt` with "child did not exit within 5s". It is
PRE-EXISTING on main --- 72da24a fails it in a clean worktree with its
own target dir --- so while it reds, no branch can present a green
sixteen-stage gate, main included. §5b is held behind this lane.

Framing revision 1, and it proposes NO FIX, because the mechanism is not
known. What it does instead is fix the shape of the problem so the next
attempt is not another guess:

  - Ground truth, cited: neither binary handles signals. `run_gpu`
    (src/main.rs:324) blocks in `command.status()` with no handler, and
    grepping all of pmacs-gpu/src for signal machinery returns nothing.
    The probe polls at 50ms. Two processes with default SIGINT
    disposition should both die at once --- this deepens the puzzle
    rather than explaining it, and the framing says so.
  - Ruled out by measurement, with the method for each: load, tmpfs
    (tested by experiment, not argument), leaked daemons, inotify,
    --workspace feature unification, and any specific preceding test.
  - The reduction paradox stated as the problem's real shape: 5/5 in
    the full sweep, 0/N in every reduction, including all 37 preceding
    targets plus the suite.
  - One retracted claim kept as a warning, because it was mine: the
    "mechanism located" report described a healthy teardown. The
    sampler behind it caught 394 launchers with a 5s maximum lifetime
    while the failing instance must live 8s or more.

The first step is diagnostic only: an instrument keyed on the FAILING
instance --- launchers outliving ~6s --- capturing /proc/<pid>/status
signal masks, since SigIgn survives fork and exec while handlers do not.

Acceptance criteria are written now so the fix cannot quietly become
"make the test pass": a demonstrated mechanism with a mutation-tested
witness, sweep-crdt green three consecutive times, the reduction paradox
explained or recorded as unexplained, and no deadline raised or test
skipped.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016bqGA6s9tTUFzYpbeW3tai
This commit is contained in:
Levi Neuwirth 2026-08-18 14:21:58 +02:00
parent 72da24adea
commit f1992a65d9
No known key found for this signature in database
2 changed files with 209 additions and 0 deletions

View File

@ -270,6 +270,55 @@ hazard in a shape that looks committed. **A documented error message
that never appears is worse than no documentation**, because the reader
waits for a signal that is not coming.
## GPU launcher / probe SIGINT teardown — FRAMING ONLY, awaiting approval
**Written with the branch's FIRST commit**, per the standing correction
from #171 and #215.
- **Branch `gpu-probe-sigint-teardown`**, base `githubsucks/main` @
**`72da24a`**, worktree
`/home/jeans/Repos/personal/pmacs-probe-sigint`. Recover with
`git fetch githubsucks && git checkout gpu-probe-sigint-teardown`.
- **No PR. Framing revision 1 at `docs/gpu-probe-sigint-framing.md`;
NO IMPLEMENTATION and no fix proposed** — the mechanism is not known
yet, and the framing says so rather than guessing.
- **Why it exists.** `ctrl_c_on_launcher_group_does_not_reach_spawned_daemon`
fails in gate stage `sweep-crdt` with "child did not exit within 5s".
**Pre-existing on `main`** — `72da24a` fails it in a clean worktree
with its own target dir, 119 binaries green and one red. While it
reds, **no branch can present a green sixteen-stage gate, `main`
included.**
- **`panel-mapping-generation` (§5b) is HELD BEHIND THIS LANE** by
explicit instruction. That lane is code-complete at `5174f73` with
its own fifteen stages green; its sixteenth stage is this defect.
- **Reproduction is 5/5 in the full sweep and 0/N in every reduction.**
All 37 preceding targets plus the suite run green, and other
packages' targets run after the failure. That paradox is the shape of
the problem, not an aside.
- **Ruled out by measurement — do not re-run:** machine load; tmpfs
starving RAM (settled by experiment, not argument — `/tmp` 21G→1.2G,
available 27G→45G, still red); leaked daemons; inotify;
`--workspace` feature unification; any specific preceding test.
- **Ground truth that DEEPENS rather than explains it.** Neither binary
handles signals: `run_gpu` (`src/main.rs:324`) blocks in
`command.status()` with no handler, and grepping the whole of
`pmacs-gpu/src` for signal machinery returns nothing. The probe polls
at 50ms (`pmacs-gpu/src/main.rs:1065`). Two processes with default
`SIGINT` disposition should both die at once.
- **One retracted claim, kept as a warning.** An earlier "mechanism
located" report — launcher in `do_wait`, probe child in
`futex_do_wait` — was retracted on its own evidence: the sampler
caught 394 launchers with a maximum lifetime of 5s, while the failing
instance must live 8s or more. It had described a HEALTHY teardown.
The first step is therefore an instrument recording **only**
launchers outliving ~6s, with `/proc/<pid>/status` signal masks —
`SigIgn` survives `fork` AND `exec`, handlers do not.
- **Gates:** `./scripts/gate --protocol --acceptance
gpu_invocation_acceptance` at minimum; acceptance criterion A2
requires `sweep-crdt` green **three consecutive times** before the
fix is believed, since 1/1 is not evidence for a defect that hid from
every reduction.
## `scripts/gate` TMPDIR isolation — PR #240 OPEN
**Written with the branch's first commit**, per the standing correction

View File

@ -0,0 +1,160 @@
# GPU launcher / probe SIGINT teardown — framing
Revision 1. Status: **awaiting approval. No implementation.**
## 1. The problem, stated as what is observed
`gpu_invocation_acceptance::crdt::ctrl_c_on_launcher_group_does_not_reach_spawned_daemon`
fails inside the `sweep-crdt` gate stage with
```
child did not exit within 5s (tests/gpu_invocation_acceptance.rs:180, called from :1115)
```
The test spawns `pmacs --gpu --socket <s>` in its own process group
(`process_group(0)`, `:1107`), waits for the probe to report
`phase=ready`, sends `SIGINT` to the **group**, and requires the
launcher to exit within five seconds.
This is **pre-existing on `main`**. It is not caused by any feature
branch: the identical `build-crdt && sweep-crdt` pair run at `72da24a`
in a clean worktree with its own target directory fails the same test —
119 test binaries green, one red.
## 2. Why this blocks more than one lane
`sweep-crdt` is stage 15 of the sixteen-stage `--protocol` gate. While it
reds, **no branch can present a green gate**, `main` included. The
`panel-mapping-generation` (§5b) lane is complete with its own fifteen
stages green and is held behind this lane by explicit instruction.
## 3. Ground truth (cited, not recalled)
- **The launcher does not handle signals.** `run_gpu`
(`src/main.rs:324`) builds a `Command` for the GPU binary and blocks
in `command.status()` (`:363`) — a plain `waitpid`. It installs no
handler, so `SIGINT`'s default action should terminate it outright.
- **`pmacs-gpu` does not handle signals either.** Grepping
`SIGINT|signal_hook|sigaction|ctrlc|set_handler|pthread_sigmask|sigprocmask`
across `pmacs-gpu/src` returns **nothing**. The probe should also die
on the default action.
- **The probe never blocks indefinitely.** `run_headless_managed_probe`
(`pmacs-gpu/src/main.rs:1065`) loops on
`event_rx.recv_timeout(Duration::from_millis(50))`, so it wakes twenty
times a second regardless of traffic.
- **The daemon *does* handle signals**, and deliberately:
`src/daemon.rs:629-641` registers `SIGTERM`/`SIGINT` through
`signal_hook::flag`. The daemon is the process the test asserts must
**survive**; it is detached from the launcher's group, which is the
property under test.
- **Only one test in the suite uses `process_group(0)`** — this one.
Taken together the ground truth **deepens** the puzzle rather than
explaining it: two processes with default `SIGINT` disposition, one of
them polling at 50 ms, should both die immediately.
## 4. What has been ruled out, each by measurement
Recorded so this lane does not re-run them.
| hypothesis | how it was refuted |
|---|---|
| machine load | red on a quiet box, load 2.77 at launch |
| tmpfs starving RAM | `/tmp` 21 G → 1.2 G, available 27 G → 45 G; still red |
| leaked test daemons | peak 58, only +8 per sweep; green runs already sat at 46–60 |
| inotify exhaustion | 47 instances in use of 1024 |
| `--workspace` feature unification | same two targets under `--workspace` are green |
| a specific preceding test | **all 37 preceding targets + the suite run green** |
| later targets / other packages | they run at log lines 4848+, after the failure at 3066 |
**The last row is the strange one and is the real shape of this
problem**: the same binaries, in the same order, with the same tests
before it, pass as a subset and fail as part of the whole 119-target
sweep. Reproduction is 5/5 in the full sweep across two trees, and 0/N
in every reduction attempted.
## 5. One retracted claim, kept as a warning
An earlier attempt reported the mechanism as "the launcher blocks in
`do_wait` on a probe child stuck in `futex_do_wait`". **That was
retracted on its own evidence.** The sampler behind it caught 394
launchers and the longest-lived was **5 s total**; the failing instance
must outlive its `SIGINT` by 5 s, so its lifetime would be 8 s or more.
What was described is a *healthy* teardown from another test in the
suite. The pair is worth keeping only as the normal shape.
The lesson binds this lane's first step: **an instrument that samples
every launcher and hopes to catch the failing one is not good enough.**
## 6. Bets
1. The failure is a real teardown defect reachable by a user pressing
Ctrl-C, not a test artifact — because the test asserts an ordinary
product property and the two binaries involved carry no signal
handling at all.
2. It is **not** a timing margin. A green run finishes in 0.15 s against
a 5 s deadline (33×). A margin that large does not erode; something
different happens.
3. Therefore **raising the deadline is not a fix** and is explicitly out
of scope. If the conclusion turns out to be that the deadline is
wrong, that requires its own argument and its own approval.
## 7. First step — an instrument that keys on the failing instance
No fix is proposed yet, because the mechanism is unknown. The first
commit is diagnostic only:
- **D1.** Sample only launchers whose lifetime exceeds ~6 s, so the
failing instance is the *only* thing recorded, and capture for it:
`/proc/<pid>/status` (`SigBlk`, `SigIgn`, `SigCgt`, `State`), the
per-thread `wchan` under `/proc/<pid>/task/*/wchan`, and the same for
every child. `SigIgn`/`SigBlk` answers directly whether `SIGINT` was
ignored or blocked — including whether it was **inherited**, since
`SIG_IGN` survives both `fork` and `exec` while handlers do not.
- **D2.** Record whether the launcher had already reaped its child at
the moment the deadline expired, which separates "the child will not
die" from "the launcher will not notice".
- **D3.** Run the full sweep under D1/D2 until the failure is captured
**with** its diagnostics, and only then propose a fix.
## 8. Acceptance criteria for the eventual fix
Deliberately written now, so the fix cannot quietly become "make the
test pass".
- **A1.** The named mechanism is stated and demonstrated, not inferred:
a witness that fails before the change and passes after, plus a
mutation showing the witness bites its own clause.
- **A2.** `sweep-crdt` green on `main` for **three consecutive full
runs** — 1/1 is not evidence for a defect that hid from every
reduction.
- **A3.** The reduction paradox is explained or explicitly recorded as
unexplained. If the fix makes the sweep green without accounting for
why subsets always passed, that gap is stated in the record rather
than left for the next reader to rediscover.
- **A4.** No deadline is raised, and no test is skipped, retried, or
serialised to obtain green.
- **A5.** If the mechanism proves to be in `pmacs-gpu`'s shutdown path,
a Ctrl-C on a real `pmacs --gpu` session tears down the frontend and
leaves the daemon running — the product property the test encodes.
## 9. Coherence impact (`COHERENCE.md` §20)
- **Journey steps touched: none.** This is a teardown-correctness and
gate-trustworthiness lane; no journey step changes grade.
- **Interaction islands: none added.**
- **Config registry: not touched.**
- **Background-work attribution: not touched.**
Naming these explicitly matters: a reader auditing §20 by grade movement
would otherwise conclude this lane touches nothing, when what it
restores is the ability of every other lane to prove itself.
## 10. Out of scope
- Raising or removing the 5 s deadline (bet 3).
- The ~10 daemons each sweep leaks. Real, separately recorded, and not
implicated here — green runs already ran at 46–60 leaked daemons.
- The `gpu_initial_target_acceptance` binary including this suite as a
module, which makes a reproducing sweep report the failure twice. A
tidiness question, not a correctness one.