Repository navigation
Conversation
|
Speaking as Kun's firstmate: thank you, @Diabl0570. I read head Security review before approving fork CI: no Blocking now (on you): remove the committed CI logs. The head adds 14 downloaded GitHub Actions logs at the repo root, about 4,750 of the 5,027 added lines: What the real change does: on Herdr only, a Contract-class: new-default. Herdr is a selected backend, but for every Herdr home this replaces the specified classification on the default lifecycle path, where registration absence ( VISION.md, rule by rule
Next step: on you. Remove the log files and re-raise. After that it waits on CI, then on Kun's decision. |
…tor continuity/revalidation. Confirmed HEAD 40a5d1a and exact diff SHA256 8ee31a003fba5fcc6059df311791835b5b9895c81116efcaa037b71b41ab5f3a: nine test files, 105 insertions/29 deletions, no tracked production change. No additional edits were made. The repair supplies stable suite-owned process evidence, preserves cleanup ownership, and fixes the portable-PATH case without weakening production liveness/refusal checks or removing assertions. Reviewed all nine final retained receipts under .no-mistakes/ci-finish/receipts; each records exit=0, including fm-herdr-session-cleanup-real-process.log and fm-remote-secondmate-lifecycle-e2e-portable-path.log. These support the unchanged candidate as bounded fixture evidence, not published CI success. Locally ran bash -n on all nine changed files and git diff --check; both passed, and HEAD/diff identity remained exact. ci-second-timeout-handoff.md was not found in the worktree or the explicitly allowed evidence directory, so its contents could not be confirmed. The known Herdr helper/tripwire refusal remains unchanged; no native lifecycle actions were run. Isolated GitHub Herdr smoke, new-head validation, and all four still-red published checks remain for the outer pipeline. Retained the accepted 0/8 actual Claude/Pi live-acceptance limitation; no protected original worker is certified repaired. No pipeline control, commit, push, approval, skip, waiver, or merge was performed
…ication terminology
2f04b0d to
847f69f
Compare
Diabl0570
changed the title
|
Speaking as Kun's firstmate: thanks for cleaning this up, @Diabl0570. I re-read the thread and the new head What changed since Blocking now (on you): merge conflict. The PR is CONFLICTING with main Small ask while you're there: please drop the 16-line Contract-class: new-default (unchanged). For every Herdr home, VISION.md, rule by rule
Next step: on you. Rebase, re-raise and drop the |
2 participants
Intent
Fix the reported Firstmate relaunch false-success defect: a live original worker was reported exited by delivery, then supported relaunch returned success while the same original process survived and launch shell text entered its live TUI. Delivery and lifecycle must agree on actual liveness, relaunch must verify a real stop/replacement, and no launch command may be typed into a live agent.
Main337 converges the concrete missing-registration acceptance from the desktop memory diagnosis: Herdr agent_not_found must not yield no-agent/dead before consulting recognised live pane-process evidence. For both Claude and Pi, missing registration plus a verified live recognised process stays alive; verified shell-only may be agent-free; unreadable/foreign/contradictory process evidence refuses. Exit/relaunch must never report already-stopped or replacement success while the original PID/start incarnation survives. Classifier fixtures first; live lifecycle reproduction only in the named isolated Herdr lab, never protected original owners. This repair enables a later separately authorised waiting-worker memory policy; it is not authority to exit them now.
Proceed without the live-lab acceptance: run the full normal validation on 68b36e6 with the 16 fixture cases as the proof, and state plainly in the PR that the named live-lab acceptance did not run because of that helper prerequisite.
What Changed
Risk Assessment
Testing
The six targeted executable suites ultimately passed after correcting an initial temporary-fixture location error; baseline/target classifier comparisons and instrumented lifecycle fixtures retained direct output and process-transition evidence. These are non-live fixture observations, so all six product scenarios remain untested under the live-validation contract. The authorized 0/8 actual Claude/Pi acceptance limitation remains unchanged, and no screenshots were captured because real agent TUIs were not authorized to launch.
Evidence: Firstmate lifecycle fixture transcripts, delivery records, and PID/start transitions
Source: Firstmate lifecycle fixture transcripts, delivery records, and PID/start transitions
Evidence: Baseline defect and corrected registration/process classifier results
Source: Baseline defect and corrected registration/process classifier results
Evidence: Successful unchanged relaunch-suite rerun with corrected fixture location
Source: Successful unchanged relaunch-suite rerun with corrected fixture location
Evidence: Commands, setup correction, evidence scope, and explicit live-acceptance limitation
Source: Commands, setup correction, evidence scope, and explicit live-acceptance limitation
Pipeline
Updates from git push no-mistakes
✅ **intent** - passed
✅ No issues found.
🔧 **Rebase** - 2 issues found → auto-fixed ✅
docs/verification/runtime-backends.md- merge conflict rebasing onto origin/maintests/fm-backend-herdr.test.sh- merge conflict rebasing onto origin/main🔧 Fix applied.
✅ Re-checked - no issues remain.
✅ No issues found.
env TMPDIR="$PWD/.test-phase-tmp" FM_TEST_EVIDENCE=1 bash bin/fm-test-run.sh --jobs 1 --json ~/.no-mistakes/evidence/01M4CPXBSKQ9HYZA4V3XC0CNHD/focused-timing.json tests/fm-backend-herdr.test.sh tests/fm-control-relaunch.test.sh tests/fm-control.test.sh tests/fm-crew-state.test.sh tests/fm-herdr-session-cleanup.test.sh tests/fm-agy-harness.test.shenv TMPDIR=/tmp FM_TEST_EVIDENCE=1 bash bin/fm-test-run.sh --jobs 1 --json ~/.no-mistakes/evidence/01M4CPXBSKQ9HYZA4V3XC0CNHD/relaunch-rerun-timing.json tests/fm-control-relaunch.test.sh— corrected the initial fixture-location error; the unchanged suite completed successfully.Re-executedtest_herdr_missing_registration_never_relaunches_a_surviving_original,test_herdr_relaunch_requires_real_stop_and_replacement_incarnation,test_herdr_relaunch_rechecks_the_shell_at_the_typing_boundary, andtest_herdr_stop_postcondition_cannot_borrow_a_false_dead_readthrough a temporary instrumented driver, retaining CLI output, delivery records, PID/start evidence, journals, and fake transport payloads.Extracted baselinebin/withgit archive 3c58ec8e0123b0047f9bf319e0edd5788eec988d binand executed baseline/target registration-process counterfactuals with identical fixture inputs and independent expected-result assertions.Checked available tools and RAM/load before sequential test work; reviewed the Herdr helper contract without retrying provisioning or starting sessions.Removed temporary drivers, extracted baseline, and worktree test residue; confirmed unchanged HEAD and a clean working tree.✅ **Document** - passed
✅ No issues found.
🔧 Fix applied.
1 warning still open:
✅ **Push** - passed
✅ No issues found.