Fixes issue 547 (a fix-stuck warning stays open on the operator's channel after its issue is resolved, and the sweep raises new ones for resolved issues).
What was wrong
stillStuck and the sweep treated only wontfix as the end of a stuck fix. A resolved issue whose fix the tracker never heard as delivered (its own walk failed or stopped, and a later merge carried it; the delivery side is issue 480) stayed stuck for good: clearStuck never cleared notify.issue-tracker.fix-stuck.N, and the sweep said new warnings for issues already resolved (356 was closed at 09:59:52Z and warned at 10:03:17Z).
What changes (modules/mesh-issues)
delivery.go: stillStuck returns false for any closed issue (closedStatus: resolved or wontfix). clearStuck clears with words that say why: "resolved: issue N was proven live by its worker", "issue N was closed as wontfix: …", or the delivered wording as before (stuckEnded). Sweep runs clearStuck for every issue and then skips every closed one before saying a fix stuck.
The clear reaches the channel the same way as before: operator-channel.notify with clear, at the issue's own moved (resolving) and, when that fails, at every five-minute sweep until it is taken. The 16 open warnings all carry stuck-open true, so the first sweep after delivery clears them.
README: the stuck paragraph says a closed issue ends the warning and is never said stuck.
Tests (delivery_test.go)
New TestAResolvedIssueEndsItsStuckWarningAndIsNeverSaidStuck: for resolved and wontfix, an issue at fixing with a failed fix is said stuck, is moved closed, and its warning is cleared at the move (resolved says resolved: …); the next sweep says nothing new. A warning left open for an already resolved issue (the 16 live ones' state) is cleared at the first sweep after the channel failed at the move. An issue resolved with a failed fix never said stuck is never said stuck. Fails on main (not cleared at the move: []).
The three existing stuck tests named the fix by resolving the issue, which this fix no longer warns for; they now name it with the issue at fixing (fixedAtFixing), as agents do (ADR 0281 §3, ADR 0284). One sets delivered-not-resolved-after to 24h so only the stuck warning is counted.
go vet, go test -race ./... in the module pass. MESH_GATE=<controller built from mesh-controller main dd2fd28> MESH_CHECK_MODULES=mesh-issues sh merge-check.sh passes (words, agent-home, paths, manifests: 69 as kept, the module's tests).
Limits I chose
ADR 0281 §4 says the warning is cleared when every fix is delivered; this adds "or the issue is closed", as the issue's expected behaviour states. The record may want a progressive insight saying so.
No manifest change and no state migration.
Reviewers' checklist
The new test fails without the fix; checks green.
A resolved or wontfix issue is cleared at its move and at every sweep while the channel fails, and never said stuck.
An issue at fixing with a stuck fix is still said once and cleared when delivered.
No change in modules/messenger or modules/mesh-delivery.
By fix-547@g14. Not merged.
Fixes issue 547 (a fix-stuck warning stays open on the operator's channel after its issue is resolved, and the sweep raises new ones for resolved issues).
## What was wrong
`stillStuck` and the sweep treated only `wontfix` as the end of a stuck fix. A resolved issue whose fix the tracker never heard as delivered (its own walk failed or stopped, and a later merge carried it; the delivery side is issue 480) stayed stuck for good: `clearStuck` never cleared `notify.issue-tracker.fix-stuck.N`, and the sweep said new warnings for issues already resolved (356 was closed at 09:59:52Z and warned at 10:03:17Z).
## What changes (modules/mesh-issues)
- `delivery.go`: `stillStuck` returns false for any closed issue (`closedStatus`: resolved or wontfix). `clearStuck` clears with words that say why: "resolved: issue N was proven live by its worker", "issue N was closed as wontfix: …", or the delivered wording as before (`stuckEnded`). `Sweep` runs `clearStuck` for every issue and then skips every closed one before saying a fix stuck.
- The clear reaches the channel the same way as before: `operator-channel.notify` with `clear`, at the issue's own `moved` (resolving) and, when that fails, at every five-minute sweep until it is taken. The 16 open warnings all carry `stuck-open` true, so the first sweep after delivery clears them.
- README: the stuck paragraph says a closed issue ends the warning and is never said stuck.
## Tests (delivery_test.go)
- New `TestAResolvedIssueEndsItsStuckWarningAndIsNeverSaidStuck`: for resolved and wontfix, an issue at fixing with a failed fix is said stuck, is moved closed, and its warning is cleared at the move (resolved says `resolved: …`); the next sweep says nothing new. A warning left open for an already resolved issue (the 16 live ones' state) is cleared at the first sweep after the channel failed at the move. An issue resolved with a failed fix never said stuck is never said stuck. Fails on main (`not cleared at the move: []`).
- The three existing stuck tests named the fix by resolving the issue, which this fix no longer warns for; they now name it with the issue at fixing (`fixedAtFixing`), as agents do (ADR 0281 §3, ADR 0284). One sets `delivered-not-resolved-after` to 24h so only the stuck warning is counted.
- `go vet`, `go test -race ./...` in the module pass. `MESH_GATE=<controller built from mesh-controller main dd2fd28> MESH_CHECK_MODULES=mesh-issues sh merge-check.sh` passes (words, agent-home, paths, manifests: 69 as kept, the module's tests).
## Limits I chose
- ADR 0281 §4 says the warning is cleared when every fix is delivered; this adds "or the issue is closed", as the issue's expected behaviour states. The record may want a progressive insight saying so.
- No manifest change and no state migration.
## Reviewers' checklist
- [ ] The new test fails without the fix; checks green.
- [ ] A resolved or wontfix issue is cleared at its move and at every sweep while the channel fails, and never said stuck.
- [ ] An issue at fixing with a stuck fix is still said once and cleared when delivered.
- [ ] No change in modules/messenger or modules/mesh-delivery.
By fix-547@g14. Not merged.
A resolved issue's worker proved the acceptance live, so a fix whose own walk failed
and that a later merge carried is not waited for. Only wontfix ended the warning, so
16 warnings stayed open on the operator's channel and the sweep raised new ones for
resolved issues.
Deliverynovox/mesh-catalog@f067fab9db3b — delivered since 2026-10-12T13:29:20Z
Delivery plan — builds mesh-issues → novox; no bus step
tier 0: mesh-issues
novox: receives mesh-issues
Machines
build-agent on ace: passed — carried by plan-1791811240112982222
mesh-issues on novox: passed — healthy 1 times over 21s
mesh-sdk on every machine running it: sent
mesh-issues on the rest running it: sent
Transitions
2026-10-12 11:32 (new) → proposed (announced): novox/mesh-catalog#239's head announced
2026-10-12 11:36 proposed → checked (checked): the verdict names this commit
2026-10-12 11:36 checked → ready (accepted): the gate passed or warned, and the repository's own check did not fail
2026-10-12 13:27 ready → published (merged): on the trunk its modules follow: merged there, and its walk opened — or nothing for a walk to move
2026-10-12 13:27 published → delivering (go): its walk started: let go by this owner in its group's order, by a person, or on the controller's own path — not a failed walk that carried a later commit for it, after which it waits for a walk of its own commit
2026-10-12 13:29 delivering → delivered (done): its walk is done, carrying its own commit: every machine of its deploy plan passed, or was left as its policy says
The commit's note under refs/notes/mesh-plan keeps every transition: git log --notes=mesh-plan.
<!-- mesh-delivery:view -->
**Delivery** `novox/mesh-catalog@f067fab9db3b` — **delivered** since 2026-10-12T13:29:20Z
**Delivery plan** — builds mesh-issues → novox; no bus step
- tier 0: mesh-issues
- novox: receives mesh-issues
**Machines**
- build-agent on ace: passed — carried by plan-1791811240112982222
- mesh-issues on novox: passed — healthy 1 times over 21s
- mesh-sdk on every machine running it: sent
- mesh-issues on the rest running it: sent
**Transitions**
- 2026-10-12 11:32 (new) → proposed (announced): novox/mesh-catalog#239's head announced
- 2026-10-12 11:36 proposed → checked (checked): the verdict names this commit
- 2026-10-12 11:36 checked → ready (accepted): the gate passed or warned, and the repository's own check did not fail
- 2026-10-12 13:27 ready → published (merged): on the trunk its modules follow: merged there, and its walk opened — or nothing for a walk to move
- 2026-10-12 13:27 published → delivering (go): its walk started: let go by this owner in its group's order, by a person, or on the controller's own path — not a failed walk that carried a later commit for it, after which it waits for a walk of its own commit
- 2026-10-12 13:29 delivering → delivered (done): its walk is done, carrying its own commit: every machine of its deploy plan passed, or was left as its policy says
The commit's note under `refs/notes/mesh-plan` keeps every transition: `git log --notes=mesh-plan`.
mesh/merge-gate: PASS — every machine composes with the change as it did without (4 of 4 compose) (modules: mesh-issues)
mesh/repo-check: PASS — its merge-check.sh passed, each part in its toolchain: merge-check.sh (go), merge-check-typescript.sh (typescript)
Delivery plan — builds mesh-issues → novox; no bus step
tier 0: mesh-issues
novox: receives mesh-issues
read by no module's build: modules/mesh-issues/README.md, modules/mesh-issues/cmd/mesh-issues/delivery_test.go
Run by the build seat on shanks as build-1791804743233946753 (builds --log build-1791804743233946753).
--- the gate: mesh-issues
mesh-issues: ok, invokes claude-code.claude_code_judge, seat:mesh-delivery.checks, seat:operator-channel.notify, records.records_status, records.records_list, records.records_read, mesh-knowledge.knowledge_highest, keeps state issues, counters, grooming, reservations, keeps log changes (1024 MiB)
long-running resources without health: 0
files asking for a setting without saying whether it is trusted: 0
1 manifest(s) checked. Judged against the seats this binary carries; a claim on one of the mesh's own seats is judged fully at registration, and a seat declared by a module not given here reads as unknown. Identities judged on a 6-character machine name, against the bounds of the providers given here
judging 0 module(s) changed against 4 machine(s), as the snapshot of 2026-10-12T11:32:09Z says they run
composing every machine as the mesh is
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
composing every machine with the change
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
merge gate: PASS — every machine composes with the change as it did without (4 of 4 compose)
judged by controller development build against the facts of 2026-10-12T11:32:09Z
machines:
- a machine (f35): composes
- holds mesh-dns-resolver (idt): composes
- holds git, holds mesh-artifact-store, holds mesh-broker, holds mesh-catalog, holds mesh-delivery, holds mesh-dns-resolver, holds mesh-store, holds mesh-vault, holds npm-package-registry, the control node, the hub (ygcyr): composes
- a machine (yvbpsm): composes
a merge would rebuild 1 module(s) in 1 tier(s): mesh-issues
notes:
- modules/mesh-issues/README.md and modules/mesh-issues/cmd/mesh-issues/delivery_test.go read by no module's build, and rebuild nothing (issue 280)
--- its merge-check.sh (go toolchain)
ok git.novox.be/novox/mesh-catalog/checks/words 0.003s
words: 12 retired words and 70 daemons' lines, none in the 43132 strings of 690 files
ok git.novox.be/novox/mesh-catalog/checks/agenthome 0.012s
agent-home: no archive and no file written into under the agent's home
ok git.novox.be/novox/mesh-catalog/checks/paths 0.006s
paths: no module names another's state on its machine (3 known, listed to go)
long-running resources without health: 69, as kept
ok git.novox.be/novox/mesh-catalog/checks/deps 0.233s
testing modules/mesh-issues
ok mesh-issues/cmd/mesh-issues 12.539s
--- its merge-check-typescript.sh (typescript toolchain)
**Merge check** at `f067fab9`
- `mesh/merge-gate`: **PASS** — every machine composes with the change as it did without (4 of 4 compose) (modules: mesh-issues)
- `mesh/repo-check`: **PASS** — its merge-check.sh passed, each part in its toolchain: merge-check.sh (go), merge-check-typescript.sh (typescript)
**Delivery plan** — builds mesh-issues → novox; no bus step
- tier 0: mesh-issues
- novox: receives mesh-issues
- read by no module's build: modules/mesh-issues/README.md, modules/mesh-issues/cmd/mesh-issues/delivery_test.go
Run by the build seat on shanks as `build-1791804743233946753` (`builds --log build-1791804743233946753`).
```
--- the gate: mesh-issues
mesh-issues: ok, invokes claude-code.claude_code_judge, seat:mesh-delivery.checks, seat:operator-channel.notify, records.records_status, records.records_list, records.records_read, mesh-knowledge.knowledge_highest, keeps state issues, counters, grooming, reservations, keeps log changes (1024 MiB)
long-running resources without health: 0
files asking for a setting without saying whether it is trusted: 0
1 manifest(s) checked. Judged against the seats this binary carries; a claim on one of the mesh's own seats is judged fully at registration, and a seat declared by a module not given here reads as unknown. Identities judged on a 6-character machine name, against the bounds of the providers given here
judging 0 module(s) changed against 4 machine(s), as the snapshot of 2026-10-12T11:32:09Z says they run
composing every machine as the mesh is
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
composing every machine with the change
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
the bus's user list leaves out 5 user(s) the mesh has minted no credential for: controller, node.f35, node.idt, node.ygcyr, node.yvbpsm. Each is a user that cannot connect until one is issued
merge gate: PASS — every machine composes with the change as it did without (4 of 4 compose)
judged by controller development build against the facts of 2026-10-12T11:32:09Z
machines:
- a machine (f35): composes
- holds mesh-dns-resolver (idt): composes
- holds git, holds mesh-artifact-store, holds mesh-broker, holds mesh-catalog, holds mesh-delivery, holds mesh-dns-resolver, holds mesh-store, holds mesh-vault, holds npm-package-registry, the control node, the hub (ygcyr): composes
- a machine (yvbpsm): composes
a merge would rebuild 1 module(s) in 1 tier(s): mesh-issues
notes:
- modules/mesh-issues/README.md and modules/mesh-issues/cmd/mesh-issues/delivery_test.go read by no module's build, and rebuild nothing (issue 280)
--- its merge-check.sh (go toolchain)
ok git.novox.be/novox/mesh-catalog/checks/words 0.003s
words: 12 retired words and 70 daemons' lines, none in the 43132 strings of 690 files
ok git.novox.be/novox/mesh-catalog/checks/agenthome 0.012s
agent-home: no archive and no file written into under the agent's home
ok git.novox.be/novox/mesh-catalog/checks/paths 0.006s
paths: no module names another's state on its machine (3 known, listed to go)
long-running resources without health: 69, as kept
ok git.novox.be/novox/mesh-catalog/checks/deps 0.233s
testing modules/mesh-issues
ok mesh-issues/cmd/mesh-issues 12.539s
--- its merge-check-typescript.sh (typescript toolchain)
```
Independent review, fresh clone, built and tested at f067fab9db.
Checks read: mesh/merge-gate — success (4 of 4 compose). mesh/repo-check — success (merge-check.sh go + typescript). mesh/delivery — success, ready.
What I verified directly (modules/mesh-issues, go1.27.1):
go test ./... passes; go test -race ./... passes (11.9s), no races.
gofmt -l . and go vet ./... are both clean.
Diff is confined to modules/mesh-issues (README, delivery.go, delivery_test.go) — no manifest change, no other module touched, matching the PR's own claim.
Confirmed TestAResolvedIssueEndsItsStuckWarningAndIsNeverSaidStuck fails without the fix: I swapped delivery.go back to main's version (keeping the new test and closedStatus, which already existed on main in tracking.go) and reran — it fails exactly as described (not cleared at the move: [], then said 0, cleared []). With the fix restored it passes.
The 16-live-warning scenario (a resolved issue whose StuckOpen was left true) is covered by the test's last section and passes: the first sweep after delivery/resolution clears it.
fixing/open issues with a stuck, undelivered fix still warn — unchanged by this PR, and the three existing tests now exercise that path via the new fixedAtFixing helper instead of resolving first, which is the right update since resolving no longer leaves a fix waiting.
Reopening: wrote and ran a scratch test (not part of the PR, removed after) covering the realistic case — an issue resolved, reopened to fixing with a new fix named — and the sweep warns again once that new fix goes stuck. That works correctly under this PR.
One latent gap, pre-existing and not introduced by this PR: if an issue is reopened and the same already-StuckSaid pull request is named again (no new fix), the per-fix StuckSaid flag in Sweep's loop (unchanged by this diff) is never reset, so that exact fix is never said stuck a second time. This path isn't part of issue 547's report and the diff doesn't touch that logic — flagging it as a separate, non-blocking issue to open rather than a defect in this PR.
On ADR 0281 §4 (knowledge-base.get 281, section Decision): §4's accepted text ties the warning's clearing strictly to delivery ("cleared there when every fix of the issue is delivered"), and §3/clause 5 deliberately keeps resolved and delivered as two separate, independently-tracked facts ("Delivered is not resolved. resolved still means the acceptance was proven live."). Before this PR, the code's behaviour matched that word for resolved issues — only wontfix ended the wait early, and even that exception isn't written into §4. Making resolved also end the wait without a delivered signal is a deliberate policy choice (trust the human-proven resolution over the independent delivery signal), not a stale fact being corrected — the ADR's words were accurate about resolved issues when this PR found them. So I judge this a change of decision, not a progressive insight: it needs its own (small) decision record amending ADR 0281 §4 (and ideally naming the undocumented wontfix precedent too), proposed through the knowledge base, not an in-place correction of ADR 0281. The PR itself flags this as open ("the record may want a progressive insight saying so") and makes no record change, which is the right thing to leave to review rather than do unilaterally — I'd ask that a decision record be filed as a prompt follow-up, but it doesn't block this code fix: the behaviour matches issue 547's plainly stated expectation, is well tested, and the triage already judged "needs: none" for a new decision before coding.
Summary: the fix is correct, narrowly scoped, race-clean, and proven against a failing-without-the-fix test. Approving the code; open a follow-up decision record for the ADR 0281 §4 wording (closed vs. delivered) and, separately, an issue for the same-PR reopen gap in Sweep's per-fix StuckSaid latch.
Verdict: approve at f067fab
Independent review, fresh clone, built and tested at f067fab9db3b624f871bf301a474b11b2051599f.
Checks read: mesh/merge-gate — success (4 of 4 compose). mesh/repo-check — success (merge-check.sh go + typescript). mesh/delivery — success, ready.
What I verified directly (modules/mesh-issues, go1.27.1):
- `go test ./...` passes; `go test -race ./...` passes (11.9s), no races.
- `gofmt -l .` and `go vet ./...` are both clean.
- Diff is confined to modules/mesh-issues (README, delivery.go, delivery_test.go) — no manifest change, no other module touched, matching the PR's own claim.
- Confirmed `TestAResolvedIssueEndsItsStuckWarningAndIsNeverSaidStuck` fails without the fix: I swapped delivery.go back to main's version (keeping the new test and `closedStatus`, which already existed on main in tracking.go) and reran — it fails exactly as described (`not cleared at the move: []`, then `said 0, cleared []`). With the fix restored it passes.
- The 16-live-warning scenario (a resolved issue whose StuckOpen was left true) is covered by the test's last section and passes: the first sweep after delivery/resolution clears it.
- `fixing`/`open` issues with a stuck, undelivered fix still warn — unchanged by this PR, and the three existing tests now exercise that path via the new `fixedAtFixing` helper instead of resolving first, which is the right update since resolving no longer leaves a fix waiting.
- Reopening: wrote and ran a scratch test (not part of the PR, removed after) covering the realistic case — an issue resolved, reopened to `fixing` with a *new* fix named — and the sweep warns again once that new fix goes stuck. That works correctly under this PR.
- One latent gap, pre-existing and **not** introduced by this PR: if an issue is reopened and the *same already-`StuckSaid`* pull request is named again (no new fix), the per-fix `StuckSaid` flag in `Sweep`'s loop (unchanged by this diff) is never reset, so that exact fix is never said stuck a second time. This path isn't part of issue 547's report and the diff doesn't touch that logic — flagging it as a separate, non-blocking issue to open rather than a defect in this PR.
On ADR 0281 §4 (`knowledge-base.get 281`, section Decision): §4's accepted text ties the warning's clearing strictly to delivery ("cleared there when every fix of the issue is delivered"), and §3/clause 5 deliberately keeps `resolved` and `delivered` as two separate, independently-tracked facts ("Delivered is not resolved. resolved still means the acceptance was proven live."). Before this PR, the code's behaviour matched that word for resolved issues — only `wontfix` ended the wait early, and even that exception isn't written into §4. Making `resolved` also end the wait without a delivered signal is a deliberate policy choice (trust the human-proven resolution over the independent delivery signal), not a stale fact being corrected — the ADR's words were accurate about resolved issues when this PR found them. So I judge this a **change of decision, not a progressive insight**: it needs its own (small) decision record amending ADR 0281 §4 (and ideally naming the undocumented wontfix precedent too), proposed through the knowledge base, not an in-place correction of ADR 0281. The PR itself flags this as open ("the record may want a progressive insight saying so") and makes no record change, which is the right thing to leave to review rather than do unilaterally — I'd ask that a decision record be filed as a prompt follow-up, but it doesn't block this code fix: the behaviour matches issue 547's plainly stated expectation, is well tested, and the triage already judged "needs: none" for a new decision before coding.
Summary: the fix is correct, narrowly scoped, race-clean, and proven against a failing-without-the-fix test. Approving the code; open a follow-up decision record for the ADR 0281 §4 wording (closed vs. delivered) and, separately, an issue for the same-PR reopen gap in `Sweep`'s per-fix `StuckSaid` latch.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Fixes issue 547 (a fix-stuck warning stays open on the operator's channel after its issue is resolved, and the sweep raises new ones for resolved issues).
What was wrong
stillStuckand the sweep treated onlywontfixas the end of a stuck fix. A resolved issue whose fix the tracker never heard as delivered (its own walk failed or stopped, and a later merge carried it; the delivery side is issue 480) stayed stuck for good:clearStucknever clearednotify.issue-tracker.fix-stuck.N, and the sweep said new warnings for issues already resolved (356 was closed at 09:59:52Z and warned at 10:03:17Z).What changes (modules/mesh-issues)
delivery.go:stillStuckreturns false for any closed issue (closedStatus: resolved or wontfix).clearStuckclears with words that say why: "resolved: issue N was proven live by its worker", "issue N was closed as wontfix: …", or the delivered wording as before (stuckEnded).SweeprunsclearStuckfor every issue and then skips every closed one before saying a fix stuck.operator-channel.notifywithclear, at the issue's ownmoved(resolving) and, when that fails, at every five-minute sweep until it is taken. The 16 open warnings all carrystuck-opentrue, so the first sweep after delivery clears them.Tests (delivery_test.go)
TestAResolvedIssueEndsItsStuckWarningAndIsNeverSaidStuck: for resolved and wontfix, an issue at fixing with a failed fix is said stuck, is moved closed, and its warning is cleared at the move (resolved saysresolved: …); the next sweep says nothing new. A warning left open for an already resolved issue (the 16 live ones' state) is cleared at the first sweep after the channel failed at the move. An issue resolved with a failed fix never said stuck is never said stuck. Fails on main (not cleared at the move: []).fixedAtFixing), as agents do (ADR 0281 §3, ADR 0284). One setsdelivered-not-resolved-afterto 24h so only the stuck warning is counted.go vet,go test -race ./...in the module pass.MESH_GATE=<controller built from mesh-controller main dd2fd28> MESH_CHECK_MODULES=mesh-issues sh merge-check.shpasses (words, agent-home, paths, manifests: 69 as kept, the module's tests).Limits I chose
Reviewers' checklist
By fix-547@g14. Not merged.
Delivery
novox/mesh-catalog@f067fab9db3b— delivered since 2026-10-12T13:29:20ZDelivery plan — builds mesh-issues → novox; no bus step
Machines
Transitions
The commit's note under
refs/notes/mesh-plankeeps every transition:git log --notes=mesh-plan.Merge check at
f067fab9mesh/merge-gate: PASS — every machine composes with the change as it did without (4 of 4 compose) (modules: mesh-issues)mesh/repo-check: PASS — its merge-check.sh passed, each part in its toolchain: merge-check.sh (go), merge-check-typescript.sh (typescript)Delivery plan — builds mesh-issues → novox; no bus step
Run by the build seat on shanks as
build-1791804743233946753(builds --log build-1791804743233946753).Verdict: approve at
f067fabIndependent review, fresh clone, built and tested at
f067fab9db.Checks read: mesh/merge-gate — success (4 of 4 compose). mesh/repo-check — success (merge-check.sh go + typescript). mesh/delivery — success, ready.
What I verified directly (modules/mesh-issues, go1.27.1):
go test ./...passes;go test -race ./...passes (11.9s), no races.gofmt -l .andgo vet ./...are both clean.TestAResolvedIssueEndsItsStuckWarningAndIsNeverSaidStuckfails without the fix: I swapped delivery.go back to main's version (keeping the new test andclosedStatus, which already existed on main in tracking.go) and reran — it fails exactly as described (not cleared at the move: [], thensaid 0, cleared []). With the fix restored it passes.fixing/openissues with a stuck, undelivered fix still warn — unchanged by this PR, and the three existing tests now exercise that path via the newfixedAtFixinghelper instead of resolving first, which is the right update since resolving no longer leaves a fix waiting.fixingwith a new fix named — and the sweep warns again once that new fix goes stuck. That works correctly under this PR.StuckSaidpull request is named again (no new fix), the per-fixStuckSaidflag inSweep's loop (unchanged by this diff) is never reset, so that exact fix is never said stuck a second time. This path isn't part of issue 547's report and the diff doesn't touch that logic — flagging it as a separate, non-blocking issue to open rather than a defect in this PR.On ADR 0281 §4 (
knowledge-base.get 281, section Decision): §4's accepted text ties the warning's clearing strictly to delivery ("cleared there when every fix of the issue is delivered"), and §3/clause 5 deliberately keepsresolvedanddeliveredas two separate, independently-tracked facts ("Delivered is not resolved. resolved still means the acceptance was proven live."). Before this PR, the code's behaviour matched that word for resolved issues — onlywontfixended the wait early, and even that exception isn't written into §4. Makingresolvedalso end the wait without a delivered signal is a deliberate policy choice (trust the human-proven resolution over the independent delivery signal), not a stale fact being corrected — the ADR's words were accurate about resolved issues when this PR found them. So I judge this a change of decision, not a progressive insight: it needs its own (small) decision record amending ADR 0281 §4 (and ideally naming the undocumented wontfix precedent too), proposed through the knowledge base, not an in-place correction of ADR 0281. The PR itself flags this as open ("the record may want a progressive insight saying so") and makes no record change, which is the right thing to leave to review rather than do unilaterally — I'd ask that a decision record be filed as a prompt follow-up, but it doesn't block this code fix: the behaviour matches issue 547's plainly stated expectation, is well tested, and the triage already judged "needs: none" for a new decision before coding.Summary: the fix is correct, narrowly scoped, race-clean, and proven against a failing-without-the-fix test. Approving the code; open a follow-up decision record for the ADR 0281 §4 wording (closed vs. delivered) and, separately, an issue for the same-PR reopen gap in
Sweep's per-fixStuckSaidlatch.