LAB: release_ready.sh not_measured on main: autopilot_found cannot parse ap_policy_applies' subshell (autopilot.sh:76) #4968

Activity

1. noahgift

   added

   owner:aprender-78owning sessionowning session

   on Oct 9, 2026
2. noahgift commented on Oct 9, 2026

   

   noahgift

   on Oct 9, 2026

   ContributorAuthor

   More actions

   Fix committed on branch fix/4968-release-ready-subshell-form (70d0d5740a), not pushed yet. It waits for a PR slot.

   • autopilot_found reads a sourced script (. scripts/X.sh, source scripts/X.sh) as a call form and lists it as source scripts/X.sh. . "$R"/scripts/x.sh is still refused.
   • RR-T21 lists scripts/lib/release_policy.sh under the tag step.
   • New selftest rows: unlisted-dot-source-subshell (the release: standing CRUX-smoke release policy, read by autopilot, readiness and bump #4910 subshell form must parse) next to nm-unparsed-dot-quoted-root (must still be refused). Mutants M45-M47 are all killed.
   • --selftest 74/74; --mutants killed=47 total=47 errors=0.

   With the surface readable, release_ready.sh --root . at main 845004a is rc 1, no longer rc 2. These findings were hidden behind not_measured, and this LAB check's red now names them:

   ORPHANED RR-T12: no surface states tag:autopilot/readiness/bash scripts/release/release_readiness.sh --root any more
   ORPHANED RR-T20: no surface states tag:autopilot/readiness/cargo metadata --no-deps --format-version any more
   ORPHANED RR-P26: scripts/release/publish_strict.sh does not contain "no green clean-room run id recorded"
   UNLISTED tag:autopilot/deep/bash scripts/release/deep_nodefault_verdict.sh  (scripts/release/autopilot.sh:142)
   UNLISTED tag:autopilot/commit/bash scripts/release/release_readiness.sh --root  (scripts/release/autopilot.sh:280)
   UNLISTED publish:strict/no-clean-room-run-id-recorded-for-t  (scripts/release/publish_strict.sh:33)
   UNLISTED publish:strict/cannot-write-crg  (scripts/release/publish_strict.sh:34)
   UNLISTED publish:strict/cannot-load-fn-from-scripts-cascade-publish-sh  (scripts/release/publish_strict.sh:37)
   UNLISTED publish:strict/cannot-load-the-clean-room-gate  (scripts/release/publish_strict.sh:40)
   UNLISTED publish:strict/clean-room-run-crid-does-not-prove-t  (scripts/release/publish_strict.sh:41)
   CONTRADICTION publish.executor: cascade-drain.sh @ scripts/release/autopilot.sh:535 | publish_strict.sh @ docs/specifications/APR-RELEASE-001-train-and-build-kaizen.md:246
   release-ready: unlisted=7 orphaned=3 contradictions=1 listed=120 stated=124
   

   These are list drift since the readiness step moved to commit, deep_nodefault_verdict and the publish_strict refusals were added, and the executor changed. They are next on this ticket.
3. noahgift commented on Oct 9, 2026

   

   noahgift

   on Oct 9, 2026

   ContributorAuthor

   More actions

   Branch fix/4968-release-ready-subshell-form, three commits, not pushed yet (waiting for a PR slot):

   commit	what
   70d0d5740a	a sourced script (( . scripts/lib/x.sh || exit 2) is a call form. --root . rc 2 -> 1
   166aa97e92	a step opened as if run_step X && <test>; then is read as step X. The readiness block was being taken for commit-time code, so RR-T12 and RR-T20 read ORPHANED. Any other if run_step opener is now not_measured
   ed24619868	list drift: RR-P26 re-anchored on the #4869 message, RR-P38..41 (the one-door refusals), RR-T22 (the deep-nodefault verdict, #4874)

   Measured in the worktree: --root . gives unlisted=0, orphaned=0, listed=125, stated=125. --selftest passes 76/76, including the subshell-form row and the refused call forms next to it. --mutants killed 50/50 with 0 errors. pv validate reports 0 errors. This meets the done-when.

   One finding is left, rc 1: CONTRADICTION publish.executor. It is a real contradiction, not a parser artefact:

   • The spec's T-4 row names publish_strict.sh as the publish executor, and 0.70.2 T10: BLD-002 R5b one door to cargo publish: publish_strict.sh refuses a missing, red or other-commit clean-room run #4687 makes it the one door to cargo publish.
   • The autopilot's cascade step runs scripts/cascade-drain.sh -> scripts/cascade-publish.sh. git log -S publish_strict.sh -- scripts/release/autopilot.sh is empty: no commit since the tag path moved into the repo (PMAT-3459 (part 2): port the tag path into the repo — autopilot.sh + 2, 14 declared deltas #3599) has called publish_strict.sh from the autopilot.
   • The cascade path runs its own clean_room_gate and check_publish_preflight.sh, but clean_room_gate judges the newest completed run on the tag, not the run id the train recorded. The cascade path does not run publish_strict.sh's other refusals: HEAD == tag on a detached checkout, a recorded B2-gpu run id, a committed dry-run receipt, and publish order == universe.

   Which side moves is a release-path design decision under #4687, so it is not decided here. The contradiction-executor selftest row keeps it falsifiable. This check stays red on that one finding until #4687 settles the executor.
4. noahgift

   added this to the 0.71.1 milestone on Oct 10, 2026
5. noahgift

   added a parent issue on Oct 14, 2026

   • Release CI: one job, one reserved runner — a release PR must not fan out into 20 queued jobs #4433
6. noahgift

   modified the milestones: 0.71.1, backlog on Oct 14, 2026
7. noahgift commented on Oct 14, 2026

   

   noahgift

   on Oct 14, 2026

   ContributorAuthor

   More actions

   Folded into #4433 as a checklist row (TRIAGE-100, 2026-10-14): the open-ticket cap is 100, so this ticket is now tracked on that master and closed as not planned. Nothing is lost: reopen it when its work resumes, and an open PR for it reopens it.

   Agent: aprender-clerk
8. noahgift

   closed this as not plannedon Oct 14, 2026
9. noahgift

   removed a parent issue on Oct 14, 2026

   • Release CI: one job, one reserved runner — a release PR must not fan out into 20 queued jobs #4433