Repository navigation
test: guard fixture home isolation from live fleet state - #1
Merged
gk-io-dev merged 7 commits intoSep 17, 2026
Merged
Conversation
…solation regressions
gk-io-dev
force-pushed
the
fm/firstmate-test-fm-home-isolation-recovery-20260917
branch
from
September 17, 2026 09:18
9a6bfac to
0b8fc3e
Compare
…ver uncommitted diff from prior timed-out agent (env -u FM_HOME approach) — reverted it since it didn't match instructed fix. Applied exact one-line change: added FM_HOME="$root" to the node invocation in test_opencode_plugin_delivers_exact_nudge_once (tests/fm-sessionstart-nudge.test.sh). Ran the test file directly, exit=0, no failures. Did not touch ci-1 or ci-fix-agent-timeout per instructions
…ausing the "Behavior portable serial 5" (afk family) failure: propose/confirm captured the ISO timestamp and epoch via two separate `date` calls that could straddle a second boundary, producing an entered/entered_epoch (or confirmed/confirmed_epoch) pair that disagreed by 1s. tests/fm-afk-return.test.sh's test_missing_epoch_record_stays_required_after_disappearing asserts the return brief's rendered away-window start (derived from entered_epoch) matches the stored ISO "entered" field, so the race caused an intermittent mismatch exactly as seen in the CI log ("away 2026-09-17T11:52:52Z" vs expected "...51Z"). Fix: capture epoch first, then derive the ISO string from that same epoch via a new helper fm_afk_contract_iso_of_epoch, so the pair can never straddle a boundary. Verified the previously-failing test passes deterministically (8/8 isolated runs) and the rest of fm-afk-return.test.sh and fm-afk-contract.test.sh remain green (one unrelated, pre-existing, sandbox-only fd-redirect test fails identically on unmodified baseline too, so left alone). "Behavior portable serial 2" fails due to tests/fm-backend.test.sh's symlinked-prefix Treehouse-lock test, which is the same pre-existing defect documented and explicitly declared out-of-scope across three prior rounds of this branch's history (confirmed reproducing identically on fork main with this PR's change fully reverted). Per the user's repeated explicit instruction, this was left untouched. That check may continue to report failure for this unrelated, already-flagged reason
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What Changed
tests/lib.sh: added an ambient-home poison guard that runs on first source — clears any inheritedFM_ROOT_OVERRIDE/FM_STATE_OVERRIDE/FM_DATA_OVERRIDE/FM_CONFIG_OVERRIDE/FM_PROJECTS_OVERRIDE/FM_PENDING_REPLY_DIR_OVERRIDEand pinsFM_HOMEto a fresh per-process scratch directory (with a.fm-ambient-guardmarker, no fleet state), so a fixture that forgets to set its ownFM_HOMEcan no longer fall through to the real checkout/live fleet home.tests/fm-ambient-home-guard.test.sh: regression coverage with a sentinel MAIN/MATE secondmate fixture proving (1)tests/lib.shpinsFM_HOMEaway from$ROOT, (2) explicitly bypassing the guard reproduces the original incident (a fabricated child ledger line landing on a live parent channel), (3) a forgottenFM_HOMEoverride now lands in the guard directory instead of the sentinel, and (4) inherited directory overrides are cleared and can't reach the sentinel either.tests/fm-watcher-lock.test.sh: threefm-guard.shinvocations intest_guard_warningsnow setFM_HOMEexplicitly alongsideFM_ROOT_OVERRIDE, since they can no longer rely on an unsetFM_HOMEfalling back to root.CONTRIBUTING.md: added a pointer to the ambient-home guard comment intests/lib.shas the source of truth for fixture home initialization and override ordering.🤖 Generated with Claude Code
Risk Assessment
✅ Low: Test-only change adding an ambient-home guard in tests/lib.sh plus regression test; fallback logic matches actual bin/fm-inactive-reconcile.sh behavior, existing suites already clear these overrides in the test runner, and watcher-lock test updates are consistent with the new pinned FM_HOME.
Testing
All 4 new ambient-home-guard tests and all 34 watcher-lock tests pass live against the real bash test harness; confirmed the core regression test fails without the fix (not vacuous). No product/runtime surface beyond this bash test tooling, and no findings.
bin/fm-test-run.sh tests/fm-ambient-home-guard.test.sh→ok - tests/lib.sh pins FM_HOME to an empty, non-root directoryok - unsetting FM_HOME reproduces the incident against a private fixture (never against $ROOT)— fabricated PR ledger line landed on fixture's mate.status when FM_HOME unsetok - a forgotten FM_HOME override lands in the pinned guard directory, not the sentinel— sentinel fixture untouched (fingerprint unchanged, no 'leak-child' trace)ok - fixture initialization clears inherited directory overrides and protects live statenot ok - tests/lib.sh must export a non-empty FM_HOME, exit=1. Restored fixed tests/lib.sh afterward; worktree clean.bin/fm-test-run.sh tests/fm-watcher-lock.test.sh→ all 34 casesok, exit=0Pipeline
Updates from git push no-mistakes
⏭️ **intent** - skipped
✅ No issues found.
✅ **Rebase** - passed
✅ No issues found.
✅ **Review** - passed
✅ No issues found.
✅ **Test** - passed
✅ No issues found.
bin/fm-test-run.sh tests/fm-ambient-home-guard.test.sh→ok - tests/lib.sh pins FM_HOME to an empty, non-root directoryok - unsetting FM_HOME reproduces the incident against a private fixture (never against $ROOT)— fabricated PR ledger line landed on fixture's mate.status when FM_HOME unsetok - a forgotten FM_HOME override lands in the pinned guard directory, not the sentinel— sentinel fixture untouched (fingerprint unchanged, no 'leak-child' trace)ok - fixture initialization clears inherited directory overrides and protects live statenot ok - tests/lib.sh must export a non-empty FM_HOME, exit=1. Restored fixed tests/lib.sh afterward; worktree clean.bin/fm-test-run.sh tests/fm-watcher-lock.test.sh→ all 34 casesok, exit=0bin/fm-test-run.sh tests/fm-ambient-home-guard.test.shbin/fm-test-run.sh tests/fm-watcher-lock.test.shmanual: swapped tests/lib.sh to base-commit content to prove the new guard test fails without the fix, then restored✅ **Document** - passed
✅ No issues found.
✅ **Lint** - passed
✅ No issues found.
✅ **Push** - passed
✅ No issues found.