fix(teardown): allow teardown when work is on any remote (incl. fork), not just local main - #31
Merged
kunchenguid merged 2 commits intoJun 22, 2026
Conversation
Owner
|
Now that #30 and #37 have merged, this conflicts with main - #37's secondmate work substantially rewrote The fix is still relevant: current Could you rebase onto the latest |
…, not just local main A local-only-registered project whose task pushes its work to a fork (upstream-contribution PRs) was falsely refused by fm-teardown.sh: the local-only branch short-circuited to a strict local-main check before the 'on a remote' check ran, so fork-pushed work that IS safely on a remote looked unlanded. This bit us 3x this session (no-mistakes x2, treehouse orphan). The fix unifies the safety check across modes: HEAD --not --remotes (empty == on any remote, fork included) is the primary gate for every mode. The local-main check now runs only as a fallback for local-only projects whose work is genuinely on no remote at all. The real safety property - refuse truly unpushed work - is preserved, and --force is unchanged. Re-applied onto the kunchenguid#37 teardown rewrite, which already added FM_ROOT_OVERRIDE / FM_STATE_OVERRIDE support (so the new tests can drive the script in isolation); this change is now purely the local-only fork-remote eligibility fix. Motivating incidents: - no-mistakes teardown false-refused twice during this session's 8-PR upstream-contribution batch - a treehouse orphan stuck behind the same false refuse Tests (tests/fm-teardown.test.sh): - local-only + HEAD on a fork remote-tracking branch -> ALLOW (the fix) - local-only + truly unpushed work (no remote) -> REFUSE (safety) - local-only + work merged into local main -> ALLOW (no regression) - no-mistakes + HEAD on origin -> ALLOW (no regression) - no-mistakes + truly unpushed -> REFUSE (no regression) - local-only + unpushed + --force -> ALLOW (escape hatch) AI disclosure: Human-reviewed.
e-jung
force-pushed
the
fix/teardown-local-only-fork-remote
branch
from
June 22, 2026 22:43
79a8d44 to
c534b41
Compare
Owner
|
Thanks @e-jung, merged. The rebase onto the rewritten fm-teardown.sh structure was clean, and the dedicated tests/fm-teardown.test.sh - asserting the truly-unpushed REFUSE path in every mode, not just the fork-allow happy path - made the safety guard easy to verify. The fork-pushed-work false-refuse was a real recurring bite this session, so this is a genuinely useful fix. Appreciated. |
leo1oel
referenced
this pull request
in leo1oel/nemo
Jun 23, 2026
Bring the upstream secondmate feature (PRs #37/#42/#43/#31/kunchenguid#45) to herdr-backend, rewriting every treehouse/tmux mechanism into the herdr equivalents this branch already uses. A secondmate is a crewmate whose workspace is an isolated firstmate home (its own FM_HOME) and whose brief is a charter; it runs the same spawn/brief/status/watch/teardown/recovery lifecycle and is idle by default. treehouse -> herdr mapping: - Home leasing -> a herdr worktree of the firstmate repo. fm-home-seed.sh `-` runs `herdr worktree create --cwd $FM_ROOT --branch secondmate-<id>` and records the worktree path + open_workspace_id; herdr never recycles a worktree, so persistence is automatic and all lease/return bookkeeping is dropped. The workspace id is stored beside the home marker (.fm-secondmate-home.workspace) so retirement can find it. - Retirement / seed rollback -> `herdr worktree remove --workspace <ws> --force`. An explicit <home> path stays a plain directory home (a git clone, no herdr worktree); spawn opens a workspace for it on the fly via `herdr workspace create` and teardown closes it with `herdr workspace close` then deletes the dir. - firstmate_home_has_treehouse_slot -> "does the home record a workspace marker"; if yes remove via herdr, else plain rm. New: bin/fm-home-seed.sh (charter fill, project cloning, no-mistakes init, registry edits, transactional rollback, validate subcommand all ported as-is), bin/fm-backlog-handoff.sh (no treehouse/tmux; near-verbatim). Edits (re-implemented on the diverged herdr-backend files, not applied as upstream hunks): fm-spawn.sh (--secondmate path, FM_HOME indirection, registry/home resolution, launch in home, meta records home=/home_workspace=/projects=), fm-teardown.sh (kind=secondmate retirement via herdr, child-work refusal + --force discard, plus #31: allow teardown when work is on any remote incl. a fork), fm-brief.sh (--secondmate charter scaffold), fm-watch.sh (skip stale-pane wakes for kind=secondmate), fm-project-mode.sh (honor FM_HOME/FM_DATA_OVERRIDE so a secondmate resolves its own projects.md). kunchenguid#45: dropped the ':' from the afk skill description. Skipped fm-bootstrap.sh / fm-harness.sh (removed on this branch). Docs: AGENTS.md + README.md gain herdr-flavored secondmate sections (concept, secondmates.md routing table, fm-home-seed usage, idle-by-default contract, backlog handoff, kind=secondmate recovery, watch idle) with no treehouse/tmux wording. Tests: new herdr-stubbed tests/fm-secondmate.test.sh drives the real scripts against a stub `herdr` over the full lifecycle (seed `-`, register, validate dup ids/homes/overlap, spawn, backlog handoff, retire); tests/fm-teardown.test.sh ports the #31 any-remote/fork matrix to herdr. shellcheck bin/*.sh tests/*.sh clean; 5/5 test scripts pass.
augustomklee
added a commit
to augustomklee/firstmate
that referenced
this pull request
Jun 24, 2026
Port upstream batch 1: teardown fork-remote eligibility (kunchenguid#31) + afk skill desc (kunchenguid#45)
vipentti
pushed a commit
to vipentti/firstmate
that referenced
this pull request
Aug 5, 2026
…, not just local main (kunchenguid#31) * fix(teardown): allow teardown when work is on any remote (incl. fork), not just local main A local-only-registered project whose task pushes its work to a fork (upstream-contribution PRs) was falsely refused by fm-teardown.sh: the local-only branch short-circuited to a strict local-main check before the 'on a remote' check ran, so fork-pushed work that IS safely on a remote looked unlanded. This bit us 3x this session (no-mistakes x2, treehouse orphan). The fix unifies the safety check across modes: HEAD --not --remotes (empty == on any remote, fork included) is the primary gate for every mode. The local-main check now runs only as a fallback for local-only projects whose work is genuinely on no remote at all. The real safety property - refuse truly unpushed work - is preserved, and --force is unchanged. Re-applied onto the kunchenguid#37 teardown rewrite, which already added FM_ROOT_OVERRIDE / FM_STATE_OVERRIDE support (so the new tests can drive the script in isolation); this change is now purely the local-only fork-remote eligibility fix. Motivating incidents: - no-mistakes teardown false-refused twice during this session's 8-PR upstream-contribution batch - a treehouse orphan stuck behind the same false refuse Tests (tests/fm-teardown.test.sh): - local-only + HEAD on a fork remote-tracking branch -> ALLOW (the fix) - local-only + truly unpushed work (no remote) -> REFUSE (safety) - local-only + work merged into local main -> ALLOW (no regression) - no-mistakes + HEAD on origin -> ALLOW (no regression) - no-mistakes + truly unpushed -> REFUSE (no regression) - local-only + unpushed + --force -> ALLOW (escape hatch) AI disclosure: Human-reviewed. * no-mistakes(test): Wire fm-teardown.test.sh into CI and README --------- Co-authored-by: e-jung <8334081+e-jung@users.noreply.github.com>
This was referenced Aug 7, 2026
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
2 participants
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
Fixes
bin/fm-teardown.sh's local-only false-refuse: when a project is registered[local-only]but a task pushes its work to a fork (upstream-contribution PRs), teardown refused with "work not on main" even though the work IS safely on the fork remote. The local-only branch short-circuited to a strict local-main check before thegit log HEAD --not --remotescheck ran, so legitimate fork-pushed work looked unlanded.Motivation
This bug bit us 3× this session:
no-mistakesteardown false-refused twice during this session's 8-PR upstream-contribution batch (projects registeredlocal-onlywhose work was shipped via fork PRs to upstream).In each case the work was safely on a fork remote, but teardown demanded a local-main merge that does not exist for upstream-contribution workflows.
The fix
Unify the safety check across modes.
git log HEAD --not --remotes(empty ==HEADreachable from any remote-tracking branch) becomes the primary gate for every mode — and a fork IS a remote. The local-main check now runs only as a fallback forlocal-onlyprojects whose work is genuinely on no remote at all.Safety properties preserved:
--forceescape hatch is unchanged (still requires the captain's explicit discard word).no-mistakesanddirect-PRmode behavior is unchanged (they always used the--remotescheck).Files:
bin/fm-teardown.sh— the fix; also teaches the script to honorFM_ROOT_OVERRIDE/FM_STATE_OVERRIDElike its siblings (fm-guard.sh,fm-watch.sh), which is what makes it testable in isolation.tests/fm-teardown.test.sh— new. Covers all six cases (fork-remote allow, truly-unpushed refuse, merged-to-main allow,no-mistakesorigin allow,no-mistakesunpushed refuse,--forceoverride). Wired into.github/workflows/ci.yml..github/workflows/ci.yml— runstests/fm-teardown.test.shin the Behavior tests job alongsidetests/fm-wake-queue.test.sh.README.md— liststests/fm-teardown.test.shin the behavior-test index.AGENTS.md— rule docs: align tmux and harness guidance #3 and the local-only delivery-mode section both reflect that a fork counts as a remote in any mode.Verification
bash tests/fm-teardown.test.sh— 6/6 okbash tests/fm-wake-queue.test.sh— 26/26 ok (no regression)shellcheck bin/fm-teardown.sh tests/fm-teardown.test.sh— cleanPipeline
Updates from git push no-mistakes
AI disclosure: Human-reviewed. The fix design, the safety-property analysis, the test matrix, and the AGENTS.md wording were authored by the captain's crewmate; the CI/README wiring was applied by the no-mistakes pipeline under explicit captain approval.