mesh-knowledge: the review asks changes for a decision nothing would cite, so its acceptance never turns the module unhealthy (hq issue 504) #237

Merged
mesh-admin merged 4 commits from fix/504-propose-refuses-uncited-record into main 2026-10-12 11:44:58 +00:00
Contributor

Fixes issue 504 (the knowledge base accepts a record that its own whole-set rule cited then fails, and its health turns red).

What was wrong

The whole-set rule cited (a decision accepted here is cited by another record or names a home in hq) ran only after an acceptance (ADR 0290 §4). The review approved a decision with no home and no record linking it, the acceptance broke the rule, and module.mesh-knowledge.novox.unhealthy was raised for a fault the author could only repair afterwards (ADR 0298 on 2026-10-11; ADR 0303 and 0306–0312 today).

The decision

The rule is right; acceptance must not break it. hq held it (cycle.py) at the merge of the record's own pull request, which carried the design citing it. The knowledge base now holds it at the same point: before acceptance.

It is judged in the review, not at propose: no record can link a number only reserved (link-target refuses it), so a record that will cite a new decision can only be proposed after it. Refusing at propose would make "cited by another record" impossible for a new record.

What changes (modules/mesh-knowledge)

  • rules.go: UncitedOnAcceptance judges cited ahead over the records as JudgeSet reads them: a home, or a link from an accepted or superseded record. An imported record's correction is not judged, as JudgeSet does not judge it.
  • review.go reviewOne: before the judgement, a decision that would be uncited gets a changes review by the knowledge base itself, naming the rule and what to do (name its home, or have the citing record accepted first and propose this one again). No judgement is spent. If the record set cannot be read, the record stays proposed and unreviewed, never approved.
  • The rule run after every acceptance is unchanged and still makes the module unhealthy when it fails, e.g. after a correction of the citing record drops its link (to-be 55 §10).
  • main.go (propose's description, homes) and README.md say it.

Tests

  • TestADecisionNothingWouldCiteIsNotAccepted: no home, no citer: it stays proposed, the review is changes by the knowledge base naming the rule cited, no judgement is asked, and health stays healthy. With a home it is accepted and health stays healthy.
  • TestARecordAcceptedCitingANewDecisionLetsItBeAcceptedAndDroppingTheLinkIsLoud (replaces TestAnAcceptanceThatLeavesADecisionCitedByNothingMakesTheModuleUnhealthy): A asks changes, B citing A is accepted, A proposed again is accepted, health stays healthy. A correction of B that drops the link turns the module unhealthy naming A, and a third record citing A heals it.
  • Both fail on main with review.go reverted (memory and postgres).
  • valid() in the test kit now names a home.
  • go vet passes. MESH_GATE=<controller built from mesh-controller main d881f47> MESH_CHECK_MODULES=mesh-knowledge sh merge-check.sh passes: words, agent-home, paths, manifests (69 as kept), and the module's tests.

What this does not do

The 10 decisions already accepted uncited (ADR 0303, 0306–0312 and 2 more, as the condition lists them) stay real faults. No hq design names any of them. Each needs its author to name its home through a progressive insight (knowledge-base.propose with homes and insight), or a record that cites it. Until then the module stays unhealthy. This PR prevents new ones only.

Reviewers' checklist

  • The tests fail without the fix; the checks are green.
  • No hand workaround; no data touched. No manifest change.
  • The guard never approves: an unreadable record set leaves the record proposed.
  • The post-acceptance rule is unchanged, so a citer dropping its link later is still loud.

By fix-504@g14. Not merged.

Fixes issue 504 (the knowledge base accepts a record that its own whole-set rule `cited` then fails, and its health turns red). ## What was wrong The whole-set rule `cited` (a decision accepted here is cited by another record or names a home in hq) ran only after an acceptance (ADR 0290 §4). The review approved a decision with no home and no record linking it, the acceptance broke the rule, and `module.mesh-knowledge.novox.unhealthy` was raised for a fault the author could only repair afterwards (ADR 0298 on 2026-10-11; ADR 0303 and 0306–0312 today). ## The decision The rule is right; acceptance must not break it. hq held it (`cycle.py`) at the merge of the record's own pull request, which carried the design citing it. The knowledge base now holds it at the same point: before acceptance. It is judged **in the review, not at propose**: no record can link a number only reserved (`link-target` refuses it), so a record that will cite a new decision can only be proposed after it. Refusing at propose would make "cited by another record" impossible for a new record. ## What changes (modules/mesh-knowledge) - `rules.go`: `UncitedOnAcceptance` judges `cited` ahead over the records as `JudgeSet` reads them: a home, or a link from an accepted or superseded record. An imported record's correction is not judged, as `JudgeSet` does not judge it. - `review.go` `reviewOne`: before the judgement, a decision that would be uncited gets a `changes` review by the knowledge base itself, naming the rule and what to do (name its home, or have the citing record accepted first and propose this one again). No judgement is spent. If the record set cannot be read, the record stays proposed and unreviewed, never approved. - The rule run after every acceptance is unchanged and still makes the module unhealthy when it fails, e.g. after a correction of the citing record drops its link (to-be 55 §10). - `main.go` (propose's description, `homes`) and `README.md` say it. ## Tests - `TestADecisionNothingWouldCiteIsNotAccepted`: no home, no citer: it stays proposed, the review is `changes` by the knowledge base naming `the rule cited`, no judgement is asked, and health stays healthy. With a home it is accepted and health stays healthy. - `TestARecordAcceptedCitingANewDecisionLetsItBeAcceptedAndDroppingTheLinkIsLoud` (replaces `TestAnAcceptanceThatLeavesADecisionCitedByNothingMakesTheModuleUnhealthy`): A asks changes, B citing A is accepted, A proposed again is accepted, health stays healthy. A correction of B that drops the link turns the module unhealthy naming A, and a third record citing A heals it. - Both fail on main with review.go reverted (memory and postgres). - `valid()` in the test kit now names a home. - `go vet` passes. `MESH_GATE=<controller built from mesh-controller main d881f47> MESH_CHECK_MODULES=mesh-knowledge sh merge-check.sh` passes: words, agent-home, paths, manifests (69 as kept), and the module's tests. ## What this does not do The 10 decisions already accepted uncited (ADR 0303, 0306–0312 and 2 more, as the condition lists them) stay real faults. No hq design names any of them. Each needs its author to name its home through a progressive insight (`knowledge-base.propose` with `homes` and `insight`), or a record that cites it. Until then the module stays unhealthy. This PR prevents new ones only. ## Reviewers' checklist - [ ] The tests fail without the fix; the checks are green. - [ ] No hand workaround; no data touched. No manifest change. - [ ] The guard never approves: an unreadable record set leaves the record proposed. - [ ] The post-acceptance rule is unchanged, so a citer dropping its link later is still loud. By fix-504@g14. Not merged.
Author
Contributor

Delivery novox/mesh-catalog@56a9a1b082cb — delivered since 2026-10-12T11:50:38Z

Delivery plan — builds mesh-knowledge → novox; no bus step

  • tier 0: mesh-knowledge
  • novox: receives mesh-knowledge

Machines

  • radarr on ace: passed — healthy 1 times over 49s
  • mesh-knowledge on novox: passed — healthy 1 times over 1m30s
  • radarr on the rest running it: sent
  • mesh-knowledge on the rest running it: sent

Transitions

  • 2026-10-12 11:27 (new) → proposed (announced): novox/mesh-catalog#237's head announced
  • 2026-10-12 11:35 proposed → checked (checked): the verdict names this commit
  • 2026-10-12 11:35 checked → ready (accepted): the gate passed or warned, and the repository's own check did not fail
  • 2026-10-12 11:47 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 11:47 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 11:50 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@56a9a1b082cb` — **delivered** since 2026-10-12T11:50:38Z **Delivery plan** — builds mesh-knowledge → novox; no bus step - tier 0: mesh-knowledge - novox: receives mesh-knowledge **Machines** - radarr on ace: passed — healthy 1 times over 49s - mesh-knowledge on novox: passed — healthy 1 times over 1m30s - radarr on the rest running it: sent - mesh-knowledge on the rest running it: sent **Transitions** - 2026-10-12 11:27 (new) → proposed (announced): novox/mesh-catalog#237's head announced - 2026-10-12 11:35 proposed → checked (checked): the verdict names this commit - 2026-10-12 11:35 checked → ready (accepted): the gate passed or warned, and the repository's own check did not fail - 2026-10-12 11:47 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 11:47 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 11:50 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`.
Author
Contributor

Merge check at e799bb69

  • mesh/merge-gate: PASS — every machine composes with the change as it did without (4 of 4 compose) (modules: mesh-knowledge)
  • mesh/repo-check: FAIL — THE CHANGE ALTERS ITS OWN CHECK (merge-check.sh, merge-check-typescript.sh): main's version judged it; the change's judges the pull requests after it merges; its merge-check.sh failed: merge-check.sh: cd: line 86: can't cd to checks/deps: No such file or directory

Delivery plan — builds mesh-knowledge → novox; no bus step

  • tier 0: mesh-knowledge
  • novox: receives mesh-knowledge
  • read by no module's build: modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_test.go

Run by the build seat on ace as build-1791802348217273420 (builds --log build-1791802348217273420).

--- the gate: mesh-knowledge
note: mesh-knowledge uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here
mesh-knowledge: ok, 2 tool(s), invokes seat:issue-tracker.reserve, claude-code.claude_code_judge, claude-code.claude_code_judged, records.records_status, records.records_list, records.records_read, seat:operator-channel.notify, ready: keep by unit every 30s; watch by its tool knowledge_health every 1m0s
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 2026-10-12T10:50:44Z 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 2026-10-12T10:50:44Z

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-knowledge

notes:
  - modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go and modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_test.go read by no module's build, and rebuild nothing (issue 280)
--- THE CHANGE ALTERS ITS OWN CHECK (merge-check.sh, merge-check-typescript.sh): main's version judged it; the change's judges the pull requests after it merges
--- 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 42893 strings of 687 files
ok  	git.novox.be/novox/mesh-catalog/checks/agenthome	0.013s
agent-home: no archive and no file written into under the agent's home
ok  	git.novox.be/novox/mesh-catalog/checks/paths	0.007s
paths: no module names another's state on its machine (3 known, listed to go)
long-running resources without health: 69, as kept
merge-check.sh: cd: line 86: can't cd to checks/deps: No such file or directory
**Merge check** at `e799bb69` - `mesh/merge-gate`: **PASS** — every machine composes with the change as it did without (4 of 4 compose) (modules: mesh-knowledge) - `mesh/repo-check`: **FAIL** — THE CHANGE ALTERS ITS OWN CHECK (merge-check.sh, merge-check-typescript.sh): main's version judged it; the change's judges the pull requests after it merges; its merge-check.sh failed: merge-check.sh: cd: line 86: can't cd to checks/deps: No such file or directory **Delivery plan** — builds mesh-knowledge → novox; no bus step - tier 0: mesh-knowledge - novox: receives mesh-knowledge - read by no module's build: modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_test.go Run by the build seat on ace as `build-1791802348217273420` (`builds --log build-1791802348217273420`). ``` --- the gate: mesh-knowledge note: mesh-knowledge uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here mesh-knowledge: ok, 2 tool(s), invokes seat:issue-tracker.reserve, claude-code.claude_code_judge, claude-code.claude_code_judged, records.records_status, records.records_list, records.records_read, seat:operator-channel.notify, ready: keep by unit every 30s; watch by its tool knowledge_health every 1m0s 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 2026-10-12T10:50:44Z 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 2026-10-12T10:50:44Z 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-knowledge notes: - modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go and modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_test.go read by no module's build, and rebuild nothing (issue 280) --- THE CHANGE ALTERS ITS OWN CHECK (merge-check.sh, merge-check-typescript.sh): main's version judged it; the change's judges the pull requests after it merges --- 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 42893 strings of 687 files ok git.novox.be/novox/mesh-catalog/checks/agenthome 0.013s agent-home: no archive and no file written into under the agent's home ok git.novox.be/novox/mesh-catalog/checks/paths 0.007s paths: no module names another's state on its machine (3 known, listed to go) long-running resources without health: 69, as kept merge-check.sh: cd: line 86: can't cd to checks/deps: No such file or directory ```
jschoubben force-pushed fix/504-propose-refuses-uncited-record from e799bb6934 to 374bf5d6c5 2026-10-12 10:58:27 +00:00 Compare
Author
Contributor

Merge check at 374bf5d6

  • mesh/merge-gate: PASS — every machine composes with the change as it did without (4 of 4 compose) (modules: mesh-knowledge)
  • mesh/repo-check: FAIL — its merge-check.sh failed: --- FAIL: TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid (0.89s); and TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid/memory

Delivery plan — builds mesh-knowledge → novox; no bus step

  • tier 0: mesh-knowledge
  • novox: receives mesh-knowledge
  • read by no module's build: modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_test.go

Run by the build seat on shanks as build-1791802730082556099 (builds --log build-1791802730082556099).

--- the gate: mesh-knowledge
note: mesh-knowledge uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here
mesh-knowledge: ok, 2 tool(s), invokes seat:issue-tracker.reserve, claude-code.claude_code_judge, claude-code.claude_code_judged, records.records_status, records.records_list, records.records_read, seat:operator-channel.notify, ready: keep by unit every 30s; watch by its tool knowledge_health every 1m0s
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 2026-10-12T10:57:02Z 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 2026-10-12T10:57:02Z

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-knowledge

notes:
  - modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go and modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_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	0.018s
words: 12 retired words and 70 daemons' lines, none in the 42893 strings of 687 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	0.219s
testing modules/mesh-knowledge
==================
WARNING: DATA RACE
Write at 0x000002024fc8 by goroutine 3673:
  mesh-knowledge/cmd/mesh-knowledge.inZone.func1()
      /var/lib/build-agent/workspace/check/mesh-catalog/modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go:164 +0x30
  testing.(*common).Cleanup.func1()
      /usr/local/go/src/testing/testing.go:1317 +0x168
  testing.(*common).runCleanup()
      /usr/local/go/src/testing/testing.go:1667 +0x225
  testing.tRunner.func2()
      /usr/local/go/src/testing/testing.go:2030 +0x4c
  runtime.deferreturn()
      /usr/local/go/src/runtime/panic.go:668 +0x5d
  testing.(*T).Run.gowrap1()
      /usr/local/go/src/testing/testing.go:2101 +0x38

Previous read at 0x000002024fc8 by goroutine 3685:
  time.unixTime()
      /usr/local/go/src/time/time.go:1364 +0x1ae
  time.Unix()
      /usr/local/go/src/time/time.go:1665 +0x1a2
  os.fillFileStatFromSys()
      /usr/local/go/src/os/stat_linux.go:16 +0x177
  os.(*File).Stat()
      /usr/local/go/src/os/stat_unix.go:24 +0x133
  os.statOrZero()
      /usr/local/go/src/os/file.go:875 +0x26
  os.ReadFile()
      /usr/local/go/src/os/file.go:871 +0xe4
  github.com/nats-io/nats-server/v2/server/pse.periodic()
      /var/lib/build-agent/workspace/go-modules/github.com/nats-io/nats-server/v2@v2.11.17/server/pse/pse_linux.go:52 +0x48

Goroutine 3673 (running) created at:
  testing.(*T).Run()
      /usr/local/go/src/testing/testing.go:2101 +0xb12
  mesh-knowledge/cmd/mesh-knowledge.eachStore()
      /var/lib/build-agent/workspace/check/mesh-catalog/modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go:148 +0x9c
  mesh-knowledge/cmd/mesh-knowledge.TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid()
      /var/lib/build-agent/workspace/check/mesh-catalog/modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go:17 +0x2d
  testing.tRunner()
      /usr/local/go/src/testing/testing.go:2036 +0x21c
  testing.(*T).Run.gowrap1()
      /usr/local/go/src/testing/testing.go:2101 +0x38

Goroutine 3685 (finished) created at:
  time.goFunc()
      /usr/local/go/src/time/sleep.go:215 +0x44
==================
--- FAIL: TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid (0.89s)
    --- FAIL: TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid/memory (0.39s)
        testing.go:1712: race detected during execution of test
FAIL
FAIL	mesh-knowledge/cmd/mesh-knowledge	123.534s
FAIL
--- what failed, picked from the whole of its output (the whole is in the build's log)
==================
WARNING: DATA RACE
Write at 0x000002024fc8 by goroutine 3673:
  mesh-knowledge/cmd/mesh-knowledge.inZone.func1()
      /var/lib/build-agent/workspace/check/mesh-catalog/modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go:164 +0x30
  testing.(*common).Cleanup.func1()
      /usr/local/go/src/testing/testing.go:1317 +0x168
  testing.(*common).runCleanup()
      /usr/local/go/src/testing/testing.go:1667 +0x225
  testing.tRunner.func2()
      /usr/local/go/src/testing/testing.go:2030 +0x4c
  runtime.deferreturn()
      /usr/local/go/src/runtime/panic.go:668 +0x5d
  testing.(*T).Run.gowrap1()
      /usr/local/go/src/testing/testing.go:2101 +0x38

Previous read at 0x000002024fc8 by goroutine 3685:
  time.unixTime()
      /usr/local/go/src/time/time.go:1364 +0x1ae
  time.Unix()
      /usr/local/go/src/time/time.go:1665 +0x1a2
  os.fillFileStatFromSys()
      /usr/local/go/src/os/stat_linux.go:16 +0x177
  os.(*File).Stat()
      /usr/local/go/src/os/stat_unix.go:24 +0x133
  os.statOrZero()
      /usr/local/go/src/os/file.go:875 +0x26
  os.ReadFile()
      /usr/local/go/src/os/file.go:871 +0xe4
  github.com/nats-io/nats-server/v2/server/pse.periodic()
      /var/lib/build-agent/workspace/go-modules/github.com/nats-io/nats-server/v2@v2.11.17/server/pse/pse_linux.go:52 +0x48

Goroutine 3673 (running) created at:
  testing.(*T).Run()
      /usr/local/go/src/testing/testing.go:2101 +0xb12
  mesh-knowledge/cmd/mesh-knowledge.eachStore()
      /var/lib/build-agent/workspace/check/mesh-catalog/modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go:148 +0x9c
  mesh-knowledge/cmd/mesh-knowledge.TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid()
      /var/lib/build-agent/workspace/check/mesh-catalog/modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go:17 +0x2d
  testing.tRunner()
      /usr/local/go/src/testing/testing.go:2036 +0x21c
  testing.(*T).Run.gowrap1()
      /usr/local/go/src/testing/testing.go:2101 +0x38

Goroutine 3685 (finished) created at:
  time.goFunc()
      /usr/local/go/src/time/sleep.go:215 +0x44
==================
--- FAIL: TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid (0.89s)
    --- FAIL: TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid/memory (0.39s)
        testing.go:1712: race detected during execution of test
FAIL	mesh-knowledge/cmd/mesh-knowledge	123.534s
**Merge check** at `374bf5d6` - `mesh/merge-gate`: **PASS** — every machine composes with the change as it did without (4 of 4 compose) (modules: mesh-knowledge) - `mesh/repo-check`: **FAIL** — its merge-check.sh failed: --- FAIL: TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid (0.89s); and TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid/memory **Delivery plan** — builds mesh-knowledge → novox; no bus step - tier 0: mesh-knowledge - novox: receives mesh-knowledge - read by no module's build: modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_test.go Run by the build seat on shanks as `build-1791802730082556099` (`builds --log build-1791802730082556099`). ``` --- the gate: mesh-knowledge note: mesh-knowledge uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here mesh-knowledge: ok, 2 tool(s), invokes seat:issue-tracker.reserve, claude-code.claude_code_judge, claude-code.claude_code_judged, records.records_status, records.records_list, records.records_read, seat:operator-channel.notify, ready: keep by unit every 30s; watch by its tool knowledge_health every 1m0s 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 2026-10-12T10:57:02Z 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 2026-10-12T10:57:02Z 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-knowledge notes: - modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go and modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_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 0.018s words: 12 retired words and 70 daemons' lines, none in the 42893 strings of 687 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 0.219s testing modules/mesh-knowledge ================== WARNING: DATA RACE Write at 0x000002024fc8 by goroutine 3673: mesh-knowledge/cmd/mesh-knowledge.inZone.func1() /var/lib/build-agent/workspace/check/mesh-catalog/modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go:164 +0x30 testing.(*common).Cleanup.func1() /usr/local/go/src/testing/testing.go:1317 +0x168 testing.(*common).runCleanup() /usr/local/go/src/testing/testing.go:1667 +0x225 testing.tRunner.func2() /usr/local/go/src/testing/testing.go:2030 +0x4c runtime.deferreturn() /usr/local/go/src/runtime/panic.go:668 +0x5d testing.(*T).Run.gowrap1() /usr/local/go/src/testing/testing.go:2101 +0x38 Previous read at 0x000002024fc8 by goroutine 3685: time.unixTime() /usr/local/go/src/time/time.go:1364 +0x1ae time.Unix() /usr/local/go/src/time/time.go:1665 +0x1a2 os.fillFileStatFromSys() /usr/local/go/src/os/stat_linux.go:16 +0x177 os.(*File).Stat() /usr/local/go/src/os/stat_unix.go:24 +0x133 os.statOrZero() /usr/local/go/src/os/file.go:875 +0x26 os.ReadFile() /usr/local/go/src/os/file.go:871 +0xe4 github.com/nats-io/nats-server/v2/server/pse.periodic() /var/lib/build-agent/workspace/go-modules/github.com/nats-io/nats-server/v2@v2.11.17/server/pse/pse_linux.go:52 +0x48 Goroutine 3673 (running) created at: testing.(*T).Run() /usr/local/go/src/testing/testing.go:2101 +0xb12 mesh-knowledge/cmd/mesh-knowledge.eachStore() /var/lib/build-agent/workspace/check/mesh-catalog/modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go:148 +0x9c mesh-knowledge/cmd/mesh-knowledge.TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid() /var/lib/build-agent/workspace/check/mesh-catalog/modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go:17 +0x2d testing.tRunner() /usr/local/go/src/testing/testing.go:2036 +0x21c testing.(*T).Run.gowrap1() /usr/local/go/src/testing/testing.go:2101 +0x38 Goroutine 3685 (finished) created at: time.goFunc() /usr/local/go/src/time/sleep.go:215 +0x44 ================== --- FAIL: TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid (0.89s) --- FAIL: TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid/memory (0.39s) testing.go:1712: race detected during execution of test FAIL FAIL mesh-knowledge/cmd/mesh-knowledge 123.534s FAIL --- what failed, picked from the whole of its output (the whole is in the build's log) ================== WARNING: DATA RACE Write at 0x000002024fc8 by goroutine 3673: mesh-knowledge/cmd/mesh-knowledge.inZone.func1() /var/lib/build-agent/workspace/check/mesh-catalog/modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go:164 +0x30 testing.(*common).Cleanup.func1() /usr/local/go/src/testing/testing.go:1317 +0x168 testing.(*common).runCleanup() /usr/local/go/src/testing/testing.go:1667 +0x225 testing.tRunner.func2() /usr/local/go/src/testing/testing.go:2030 +0x4c runtime.deferreturn() /usr/local/go/src/runtime/panic.go:668 +0x5d testing.(*T).Run.gowrap1() /usr/local/go/src/testing/testing.go:2101 +0x38 Previous read at 0x000002024fc8 by goroutine 3685: time.unixTime() /usr/local/go/src/time/time.go:1364 +0x1ae time.Unix() /usr/local/go/src/time/time.go:1665 +0x1a2 os.fillFileStatFromSys() /usr/local/go/src/os/stat_linux.go:16 +0x177 os.(*File).Stat() /usr/local/go/src/os/stat_unix.go:24 +0x133 os.statOrZero() /usr/local/go/src/os/file.go:875 +0x26 os.ReadFile() /usr/local/go/src/os/file.go:871 +0xe4 github.com/nats-io/nats-server/v2/server/pse.periodic() /var/lib/build-agent/workspace/go-modules/github.com/nats-io/nats-server/v2@v2.11.17/server/pse/pse_linux.go:52 +0x48 Goroutine 3673 (running) created at: testing.(*T).Run() /usr/local/go/src/testing/testing.go:2101 +0xb12 mesh-knowledge/cmd/mesh-knowledge.eachStore() /var/lib/build-agent/workspace/check/mesh-catalog/modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go:148 +0x9c mesh-knowledge/cmd/mesh-knowledge.TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid() /var/lib/build-agent/workspace/check/mesh-catalog/modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go:17 +0x2d testing.tRunner() /usr/local/go/src/testing/testing.go:2036 +0x21c testing.(*T).Run.gowrap1() /usr/local/go/src/testing/testing.go:2101 +0x38 Goroutine 3685 (finished) created at: time.goFunc() /usr/local/go/src/time/sleep.go:215 +0x44 ================== --- FAIL: TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid (0.89s) --- FAIL: TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid/memory (0.39s) testing.go:1712: race detected during execution of test FAIL mesh-knowledge/cmd/mesh-knowledge 123.534s ```
Author
Contributor

Merge check at 5970a5f5

  • mesh/merge-gate: PASS — every machine composes with the change as it did without (4 of 4 compose) (modules: mesh-knowledge)
  • 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-knowledge → novox; no bus step

  • tier 0: mesh-knowledge
  • novox: receives mesh-knowledge
  • read by no module's build: modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_test.go

Run by the build seat on novox as build-1791803519155843559 (builds --log build-1791803519155843559).

--- the gate: mesh-knowledge
note: mesh-knowledge uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here
mesh-knowledge: ok, 2 tool(s), invokes seat:issue-tracker.reserve, claude-code.claude_code_judge, claude-code.claude_code_judged, records.records_status, records.records_list, records.records_read, seat:operator-channel.notify, ready: keep by unit every 30s; watch by its tool knowledge_health every 1m0s
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 2026-10-12T11:06:02Z 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 2026-10-12T11:06:02Z

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-knowledge

notes:
  - modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go and modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_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 42893 strings of 687 files
ok  	git.novox.be/novox/mesh-catalog/checks/agenthome	0.014s
agent-home: no archive and no file written into under the agent's home
ok  	git.novox.be/novox/mesh-catalog/checks/paths	0.007s
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	0.322s
testing modules/mesh-knowledge
ok  	mesh-knowledge/cmd/mesh-knowledge	132.502s
--- its merge-check-typescript.sh (typescript toolchain)
**Merge check** at `5970a5f5` - `mesh/merge-gate`: **PASS** — every machine composes with the change as it did without (4 of 4 compose) (modules: mesh-knowledge) - `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-knowledge → novox; no bus step - tier 0: mesh-knowledge - novox: receives mesh-knowledge - read by no module's build: modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_test.go Run by the build seat on novox as `build-1791803519155843559` (`builds --log build-1791803519155843559`). ``` --- the gate: mesh-knowledge note: mesh-knowledge uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here mesh-knowledge: ok, 2 tool(s), invokes seat:issue-tracker.reserve, claude-code.claude_code_judge, claude-code.claude_code_judged, records.records_status, records.records_list, records.records_read, seat:operator-channel.notify, ready: keep by unit every 30s; watch by its tool knowledge_health every 1m0s 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 2026-10-12T11:06:02Z 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 2026-10-12T11:06:02Z 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-knowledge notes: - modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go and modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_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 42893 strings of 687 files ok git.novox.be/novox/mesh-catalog/checks/agenthome 0.014s agent-home: no archive and no file written into under the agent's home ok git.novox.be/novox/mesh-catalog/checks/paths 0.007s 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 0.322s testing modules/mesh-knowledge ok mesh-knowledge/cmd/mesh-knowledge 132.502s --- its merge-check-typescript.sh (typescript toolchain) ```
mesh-admin left a comment
Author
Contributor

Verdict: approve at 5970a5f

Independent review for issue 504, at head 5970a5f58c.

Checks. mesh/merge-gate passes (builds mesh-knowledge for novox; 4 of 4 machines compose). mesh/repo-check passes. Locally in modules/mesh-knowledge: go vet ./... is clean and go test -race -count=1 ./... passes (100 s, memory and postgres stores).

The tests fail without the fix. I reverted only review.go to main. Both new tests then fail on both stores. TestADecisionNothingWouldCiteIsNotAccepted fails with "a decision nothing cites was accepted by its review". The A/B test fails because A is already accepted when it is proposed again.

The guard matches the rule it anticipates. UncitedOnAcceptance uses the same notion of "cited" as JudgeSet's cited:

  • A record cites a decision when it is Accepted() (which includes the research statuses active, graduated and abandoned) or superseded, and links the decision.
  • A self-link does not count.
  • A home counts.
  • Records are read as served (e.Current).
  • A correction of an imported record (zero AcceptedAt) is skipped, as JudgeSet skips it. A correction of a record accepted here is held. So for the 10 decisions already uncited, a correction that names a home passes. A correction that names no home gets changes, which is the right push.

Research efforts and superseding records.

  • Research efforts are never refused, because the guard checks only KindDecision.
  • A superseding decision with no home is refused. Its predecessor does not link it, so nothing cites it. This is not a wrong refusal: JudgeSet would flag the same record right after acceptance, and hq's cycle.py demanded a home too. The finding text tells the author what to do.

The guard never approves. If Store.All fails, the record goes through the existing no-verdict path: it stays proposed, Judging.Tries goes up, and no approval is given. No judgement is spent on a record the guard refuses.

Does the A/B flow deadlock? No.

  • A gets changes. B links A while A is still proposed, which link-target allows because A is no longer only reserved. B is accepted, A is proposed again, and A is accepted.
  • B needs its own anchor. In the test that is a home (from valid()). Two new records that only cite each other would both be refused until one names a home. The escape is always there and the finding names it.
  • An hq design pull request that merges after acceptance does not deadlock either. homes is free text that the knowledge base does not check against hq. On hq's side, cycle.py only requires a to-be design to name at least one decision, and records.py only requires the form "ADR NNNN". Neither requires the decision to be accepted or to exist first. So the author can name the design's path in homes before the hq pull request lands, and the two can merge in either order.

Not blocking, for a follow-up if wanted:

  1. A decision that got changes from the guard is not reviewed again automatically once its citing record is accepted. The author has to propose it again, as the finding says. Re-queueing records whose only finding was cited after each acceptance would make the B-then-A flow need no memory.
  2. Because homes is unchecked, an author can satisfy the guard with a path that never lands in hq. The same looseness exists after acceptance, so this pull request does not add it.
  3. The real judgement of B sees A in its proposed state. It might object to citing a record that is not in force. The fake judge cannot show this. It is worth watching the first time this flow is used for real.
  4. No test covers the unreadable-record-set branch.
  5. reviewOne now reads the whole record set under the lock for every review. That is fine at today's size.

The race fix is sound. time.Local is now set once in TestMain, before any goroutine starts, and nothing else assigns it. inZone now only asserts that the zone is set.

Verdict: approve at 5970a5f Independent review for issue 504, at head 5970a5f58c0fff99fdfb5b5112b193a640b02e55. **Checks.** mesh/merge-gate passes (builds mesh-knowledge for novox; 4 of 4 machines compose). mesh/repo-check passes. Locally in modules/mesh-knowledge: `go vet ./...` is clean and `go test -race -count=1 ./...` passes (100 s, memory and postgres stores). **The tests fail without the fix.** I reverted only review.go to main. Both new tests then fail on both stores. `TestADecisionNothingWouldCiteIsNotAccepted` fails with "a decision nothing cites was accepted by its review". The A/B test fails because A is already accepted when it is proposed again. **The guard matches the rule it anticipates.** `UncitedOnAcceptance` uses the same notion of "cited" as `JudgeSet`'s `cited`: - A record cites a decision when it is `Accepted()` (which includes the research statuses active, graduated and abandoned) or superseded, and links the decision. - A self-link does not count. - A home counts. - Records are read as served (`e.Current`). - A correction of an imported record (zero `AcceptedAt`) is skipped, as `JudgeSet` skips it. A correction of a record accepted here is held. So for the 10 decisions already uncited, a correction that names a home passes. A correction that names no home gets `changes`, which is the right push. **Research efforts and superseding records.** - Research efforts are never refused, because the guard checks only `KindDecision`. - A superseding decision with no home is refused. Its predecessor does not link it, so nothing cites it. This is not a wrong refusal: `JudgeSet` would flag the same record right after acceptance, and hq's cycle.py demanded a home too. The finding text tells the author what to do. **The guard never approves.** If `Store.All` fails, the record goes through the existing no-verdict path: it stays proposed, `Judging.Tries` goes up, and no approval is given. No judgement is spent on a record the guard refuses. **Does the A/B flow deadlock? No.** - A gets `changes`. B links A while A is still proposed, which `link-target` allows because A is no longer only reserved. B is accepted, A is proposed again, and A is accepted. - B needs its own anchor. In the test that is a home (from `valid()`). Two new records that only cite each other would both be refused until one names a home. The escape is always there and the finding names it. - An hq design pull request that merges after acceptance does not deadlock either. `homes` is free text that the knowledge base does not check against hq. On hq's side, cycle.py only requires a to-be design to name at least one decision, and records.py only requires the form "ADR NNNN". Neither requires the decision to be accepted or to exist first. So the author can name the design's path in `homes` before the hq pull request lands, and the two can merge in either order. **Not blocking, for a follow-up if wanted:** 1. A decision that got `changes` from the guard is not reviewed again automatically once its citing record is accepted. The author has to propose it again, as the finding says. Re-queueing records whose only finding was `cited` after each acceptance would make the B-then-A flow need no memory. 2. Because `homes` is unchecked, an author can satisfy the guard with a path that never lands in hq. The same looseness exists after acceptance, so this pull request does not add it. 3. The real judgement of B sees A in its proposed state. It might object to citing a record that is not in force. The fake judge cannot show this. It is worth watching the first time this flow is used for real. 4. No test covers the unreadable-record-set branch. 5. `reviewOne` now reads the whole record set under the lock for every review. That is fine at today's size. **The race fix is sound.** `time.Local` is now set once in `TestMain`, before any goroutine starts, and nothing else assigns it. `inZone` now only asserts that the zone is set.
jschoubben added 3 commits 2026-10-12 11:26:59 +00:00
The whole-set rule cited ran only after acceptance: the review approved a
decision with no home and no record linking it, the rule then failed, and
the module's health turned red for a fault its author could only repair
after the fact. hq held the rule at the record's own merge. The review now
judges it ahead, before the judgement: such a decision stays proposed with
a changes finding naming the rule. It is not refused at propose, because
no record can link a number only reserved.
inZone wrote time.Local while the bus test's nats server read it from a
goroutine that outlives its test; the race detector the repository check
now runs failed TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid.
mesh-knowledge: tell a failed whole-set rule to the operator as a warning naming the records, and leave the module healthy (hq issue 504)
mesh/delivery superseded: a newer head of the same pull request
mesh/merge-gate pass: builds mesh-knowledge → novox; no bus step; every machine composes with the change as it did without (4 of 4 compose)
mesh/repo-check pass: its merge-check.sh passed, each part in its toolchain: merge-check.sh (go), merge-check-typescript.sh (typescript)
8f5e8bcb6e
The operator decided on 2026-10-12 that a rule of the record set that fails
is a fault of the records, which their authors repair, not the module
failing: as issue 508 decided for a proposal never reviewed, health says
only whether the module works, and a finding must not hold the module's
own rollouts. The findings go through the notice the knowledge base
already sends (operator-channel.notify, a warning), once per finding,
kept as told across restarts, and cleared when the rule holds again.
jschoubben force-pushed fix/504-propose-refuses-uncited-record from 5970a5f58c to 8f5e8bcb6e 2026-10-12 11:26:59 +00:00 Compare
jschoubben added 1 commit 2026-10-12 11:27:08 +00:00
mesh-knowledge README: say a failed whole-set rule is a warning to the operator and leaves the module healthy (hq issue 504)
mesh/merge-gate pass: builds mesh-knowledge → novox; no bus step; every machine composes with the change as it did without (4 of 4 compose)
mesh/repo-check pass: its merge-check.sh passed, each part in its toolchain: merge-check.sh (go), merge-check-typescript.sh (typescript)
mesh/delivery delivered
56a9a1b082
Author
Contributor

Added after the operator's decision of 2026-10-12 on issue 504 (heads ca3c3e6, 8f5e8bc, 56a9a1b; a re-review is needed):

  • A failed whole-set rule no longer makes the module unhealthy. Health names the findings but answers healthy ("the database serves; N rule(s) ... fail, told to the operator as a warning ..."). Health now says only whether the module works (issue 508's principle), so a finding never holds the module's own walks.
  • The findings are told as a warning through the notice the knowledge base already sends. That is operator-channel.notify with severity warning, the same way tellUnreviewed works; no new mechanism. TellFindings (kb.go) runs in the bundle's background loop, never under the write lock. The keep process judges but tells nobody.
    • One notice per finding, keyed knowledge.finding.<rule>.<records> (for example knowledge.finding.cited.adr-0312).
    • What was told is kept in the store's told set, so a restart does not tell it again.
    • A notice that fails to send is not kept as told, and is sent again at the next step.
    • The notice is cleared once the rule holds for those records again.
  • Tests:
    • TestARecordAcceptedCitingANewDecisionLetsItBeAcceptedAndDroppingTheLinkIsLoud now shows: health stays healthy and names the record; the operator is told once with the rule and the record; a restarted knowledge base tells nothing again; the notice is cleared once a record cites it.
    • New: TestAFindingsNoticeThatFailsIsSentAgain.
    • The health assertion fails on the previous head, which answered unhealthy.
  • Race fix (ca3c3e6): the local time zone is set once in TestMain. inZone used to write time.Local while a nats server goroutine from the bus test was reading it, and the race detector failed TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid in the repository check.
  • README and the health tool's description now say this.
  • go test -race passes. MESH_GATE=<controller main d881f47> MESH_CHECK_MODULES=mesh-knowledge sh merge-check.sh passes.

Effect live once delivered: the 10 decisions already accepted uncited (ADR 0303, 0306–0312 and 2 more) become warning notices to the operator, and module.mesh-knowledge.novox.unhealthy clears.

Design amendment: hq #325 (to-be 55 §5, §10).

**Added after the operator's decision of 2026-10-12 on issue 504** (heads ca3c3e6, 8f5e8bc, 56a9a1b; a re-review is needed): - **A failed whole-set rule no longer makes the module unhealthy.** `Health` names the findings but answers healthy ("the database serves; N rule(s) ... fail, told to the operator as a warning ..."). Health now says only whether the module works (issue 508's principle), so a finding never holds the module's own walks. - **The findings are told as a warning through the notice the knowledge base already sends.** That is `operator-channel.notify` with severity warning, the same way `tellUnreviewed` works; no new mechanism. `TellFindings` (kb.go) runs in the bundle's background loop, never under the write lock. The `keep` process judges but tells nobody. - One notice per finding, keyed `knowledge.finding.<rule>.<records>` (for example `knowledge.finding.cited.adr-0312`). - What was told is kept in the store's told set, so a restart does not tell it again. - A notice that fails to send is not kept as told, and is sent again at the next step. - The notice is cleared once the rule holds for those records again. - **Tests:** - `TestARecordAcceptedCitingANewDecisionLetsItBeAcceptedAndDroppingTheLinkIsLoud` now shows: health stays healthy and names the record; the operator is told once with the rule and the record; a restarted knowledge base tells nothing again; the notice is cleared once a record cites it. - New: `TestAFindingsNoticeThatFailsIsSentAgain`. - The health assertion fails on the previous head, which answered unhealthy. - **Race fix (ca3c3e6):** the local time zone is set once in `TestMain`. `inZone` used to write `time.Local` while a nats server goroutine from the bus test was reading it, and the race detector failed `TestAReviewOwedOutlivesARestartAndItsLastTryIsSaid` in the repository check. - **README** and the health tool's description now say this. - `go test -race` passes. `MESH_GATE=<controller main d881f47> MESH_CHECK_MODULES=mesh-knowledge sh merge-check.sh` passes. **Effect live once delivered:** the 10 decisions already accepted uncited (ADR 0303, 0306–0312 and 2 more) become warning notices to the operator, and `module.mesh-knowledge.novox.unhealthy` clears. Design amendment: hq #325 (to-be 55 §5, §10).
Author
Contributor

Merge check at 8f5e8bcb

  • mesh/merge-gate: PASS — every machine composes with the change as it did without (4 of 4 compose) (modules: mesh-knowledge)
  • 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-knowledge → novox; no bus step

  • tier 0: mesh-knowledge
  • novox: receives mesh-knowledge
  • read by no module's build: modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_test.go

Run by the build seat on ace as build-1791804428567769774 (builds --log build-1791804428567769774).

--- the gate: mesh-knowledge
note: mesh-knowledge uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here
mesh-knowledge: ok, 2 tool(s), invokes seat:issue-tracker.reserve, claude-code.claude_code_judge, claude-code.claude_code_judged, records.records_status, records.records_list, records.records_read, seat:operator-channel.notify, ready: keep by unit every 30s; watch by its tool knowledge_health every 1m0s
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 2026-10-12T11:22:09Z 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 2026-10-12T11:22:09Z

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-knowledge

notes:
  - modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go and modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_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 43151 strings of 690 files
ok  	git.novox.be/novox/mesh-catalog/checks/agenthome	0.015s
agent-home: no archive and no file written into under the agent's home
ok  	git.novox.be/novox/mesh-catalog/checks/paths	0.007s
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	0.282s
testing modules/mesh-knowledge
ok  	mesh-knowledge/cmd/mesh-knowledge	109.743s
--- its merge-check-typescript.sh (typescript toolchain)
Unable to find image 'novox.internal:5100/mesh-tools/build@sha256:20d3ae7c4f08c886ed3f0ee50128c69c8a3a0f0ab4605b064b1f24b64fb5b9a5' locally
novox.internal:5100/mesh-tools/build@sha256:20d3ae7c4f08c886ed3f0ee50128c69c8a3a0f0ab4605b064b1f24b64fb5b9a5: Pulling from mesh-tools/build
501c3e032d32: Already exists
d456d1aa143e: Already exists
60f37e2c93ec: Already exists
e53f92c42dd6: Already exists
b571db164614: Already exists
c97f29541340: Already exists
177d10251d51: Pulling fs layer
9959e49dce84: Pulling fs layer
71d3214eca9e: Pulling fs layer
3b9efb7e21bc: Pulling fs layer
99baede48b7f: Pulling fs layer
b39b9c5e1d8c: Pulling fs layer
762a70081f46: Pulling fs layer
e3a36fed88f4: Pulling fs layer
99baede48b7f: Waiting
b39b9c5e1d8c: Waiting
762a70081f46: Waiting
e3a36fed88f4: Waiting
3b9efb7e21bc: Waiting
177d10251d51: Verifying Checksum
71d3214eca9e: Verifying Checksum
71d3214eca9e: Download complete
177d10251d51: Pull complete
3b9efb7e21bc: Verifying Checksum
3b9efb7e21bc: Download complete
99baede48b7f: Verifying Checksum
99baede48b7f: Download complete
b39b9c5e1d8c: Verifying Checksum
b39b9c5e1d8c: Download complete
762a70081f46: Verifying Checksum
762a70081f46: Download complete
9959e49dce84: Download complete
e3a36fed88f4: Verifying Checksum
e3a36fed88f4: Download complete
9959e49dce84: Pull complete
71d3214eca9e: Pull complete
3b9efb7e21bc: Pull complete
99baede48b7f: Pull complete
b39b9c5e1d8c: Pull complete
762a70081f46: Pull complete
e3a36fed88f4: Pull complete
Digest: sha256:20d3ae7c4f08c886ed3f0ee50128c69c8a3a0f0ab4605b064b1f24b64fb5b9a5
Status: Downloaded newer image for novox.internal:5100/mesh-tools/build@sha256:20d3ae7c4f08c886ed3f0ee50128c69c8a3a0f0ab4605b064b1f24b64fb5b9a5
**Merge check** at `8f5e8bcb` - `mesh/merge-gate`: **PASS** — every machine composes with the change as it did without (4 of 4 compose) (modules: mesh-knowledge) - `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-knowledge → novox; no bus step - tier 0: mesh-knowledge - novox: receives mesh-knowledge - read by no module's build: modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_test.go Run by the build seat on ace as `build-1791804428567769774` (`builds --log build-1791804428567769774`). ``` --- the gate: mesh-knowledge note: mesh-knowledge uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here mesh-knowledge: ok, 2 tool(s), invokes seat:issue-tracker.reserve, claude-code.claude_code_judge, claude-code.claude_code_judged, records.records_status, records.records_list, records.records_read, seat:operator-channel.notify, ready: keep by unit every 30s; watch by its tool knowledge_health every 1m0s 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 2026-10-12T11:22:09Z 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 2026-10-12T11:22:09Z 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-knowledge notes: - modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go and modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_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 43151 strings of 690 files ok git.novox.be/novox/mesh-catalog/checks/agenthome 0.015s agent-home: no archive and no file written into under the agent's home ok git.novox.be/novox/mesh-catalog/checks/paths 0.007s 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 0.282s testing modules/mesh-knowledge ok mesh-knowledge/cmd/mesh-knowledge 109.743s --- its merge-check-typescript.sh (typescript toolchain) Unable to find image 'novox.internal:5100/mesh-tools/build@sha256:20d3ae7c4f08c886ed3f0ee50128c69c8a3a0f0ab4605b064b1f24b64fb5b9a5' locally novox.internal:5100/mesh-tools/build@sha256:20d3ae7c4f08c886ed3f0ee50128c69c8a3a0f0ab4605b064b1f24b64fb5b9a5: Pulling from mesh-tools/build 501c3e032d32: Already exists d456d1aa143e: Already exists 60f37e2c93ec: Already exists e53f92c42dd6: Already exists b571db164614: Already exists c97f29541340: Already exists 177d10251d51: Pulling fs layer 9959e49dce84: Pulling fs layer 71d3214eca9e: Pulling fs layer 3b9efb7e21bc: Pulling fs layer 99baede48b7f: Pulling fs layer b39b9c5e1d8c: Pulling fs layer 762a70081f46: Pulling fs layer e3a36fed88f4: Pulling fs layer 99baede48b7f: Waiting b39b9c5e1d8c: Waiting 762a70081f46: Waiting e3a36fed88f4: Waiting 3b9efb7e21bc: Waiting 177d10251d51: Verifying Checksum 71d3214eca9e: Verifying Checksum 71d3214eca9e: Download complete 177d10251d51: Pull complete 3b9efb7e21bc: Verifying Checksum 3b9efb7e21bc: Download complete 99baede48b7f: Verifying Checksum 99baede48b7f: Download complete b39b9c5e1d8c: Verifying Checksum b39b9c5e1d8c: Download complete 762a70081f46: Verifying Checksum 762a70081f46: Download complete 9959e49dce84: Download complete e3a36fed88f4: Verifying Checksum e3a36fed88f4: Download complete 9959e49dce84: Pull complete 71d3214eca9e: Pull complete 3b9efb7e21bc: Pull complete 99baede48b7f: Pull complete b39b9c5e1d8c: Pull complete 762a70081f46: Pull complete e3a36fed88f4: Pull complete Digest: sha256:20d3ae7c4f08c886ed3f0ee50128c69c8a3a0f0ab4605b064b1f24b64fb5b9a5 Status: Downloaded newer image for novox.internal:5100/mesh-tools/build@sha256:20d3ae7c4f08c886ed3f0ee50128c69c8a3a0f0ab4605b064b1f24b64fb5b9a5 ```
Author
Contributor

Merge check at 56a9a1b0

  • mesh/merge-gate: PASS — every machine composes with the change as it did without (4 of 4 compose) (modules: mesh-knowledge)
  • 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-knowledge → novox; no bus step

  • tier 0: mesh-knowledge
  • novox: receives mesh-knowledge
  • read by no module's build: modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_test.go

Run by the build seat on novox as build-1791804460553230633 (builds --log build-1791804460553230633).

--- the gate: mesh-knowledge
note: mesh-knowledge uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here
mesh-knowledge: ok, 2 tool(s), invokes seat:issue-tracker.reserve, claude-code.claude_code_judge, claude-code.claude_code_judged, records.records_status, records.records_list, records.records_read, seat:operator-channel.notify, ready: keep by unit every 30s; watch by its tool knowledge_health every 1m0s
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 2026-10-12T11:22:09Z 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 2026-10-12T11:22:09Z

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-knowledge

notes:
  - modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go and modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_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 43151 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	0.292s
testing modules/mesh-knowledge
ok  	mesh-knowledge/cmd/mesh-knowledge	135.150s
--- its merge-check-typescript.sh (typescript toolchain)
**Merge check** at `56a9a1b0` - `mesh/merge-gate`: **PASS** — every machine composes with the change as it did without (4 of 4 compose) (modules: mesh-knowledge) - `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-knowledge → novox; no bus step - tier 0: mesh-knowledge - novox: receives mesh-knowledge - read by no module's build: modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_test.go Run by the build seat on novox as `build-1791804460553230633` (`builds --log build-1791804460553230633`). ``` --- the gate: mesh-knowledge note: mesh-knowledge uses the seat "operator-channel", which no module declares and the mesh does not define among the manifests given; registration judges it against the whole catalogue, and passing the declaring module's manifest too judges it here mesh-knowledge: ok, 2 tool(s), invokes seat:issue-tracker.reserve, claude-code.claude_code_judge, claude-code.claude_code_judged, records.records_status, records.records_list, records.records_read, seat:operator-channel.notify, ready: keep by unit every 30s; watch by its tool knowledge_health every 1m0s 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 2026-10-12T11:22:09Z 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 2026-10-12T11:22:09Z 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-knowledge notes: - modules/mesh-knowledge/README.md, modules/mesh-knowledge/cmd/mesh-knowledge/kit_test.go, modules/mesh-knowledge/cmd/mesh-knowledge/review_owed_test.go and modules/mesh-knowledge/cmd/mesh-knowledge/wholeset_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 43151 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 0.292s testing modules/mesh-knowledge ok mesh-knowledge/cmd/mesh-knowledge 135.150s --- its merge-check-typescript.sh (typescript toolchain) ```
mesh-admin left a comment
Author
Contributor

Verdict: approve at 56a9a1b

Re-review after the rebase. My earlier approval was at 5970a5f. The cited guard and the race fix are unchanged in substance from that review. This review covers what is new: the operator's decision (issue 504, note 4) that a failed whole-set rule is a warning to the operator and leaves the module healthy.

Checks. mesh/merge-gate passes (builds mesh-knowledge for novox; 4 of 4 machines compose). mesh/repo-check passes. Locally: go vet ./... is clean and go test -race -count=1 ./... passes (109 s, memory and postgres stores).

The tests fail on the old head. With kb.go and main.go taken from the old head, the new tests do not compile (TellFindings and findingPrefix are undefined). The changed health assertion (healthy while naming the finding) would also fail against the old Health, which returned unhealthy.

What I checked:

  • No repeat after a restart. Store.Told is an insert with on-conflict-do-nothing into the told table, so "told" is kept in the store. The test builds a second world on the same store, and it tells nothing again.
  • A clear reaches the operator's channel. busClearer sends operator-channel.notify with clear: true under the same key. The messenger's Notify handles clear with h.cleared(key) and uses the resolved line it stored from the first notify, so passing an empty resolved on the clear is fine. A clear that fails stays told and is retried at the next step. A notify that fails is marked untold and sent again; TestAFindingsNoticeThatFailsIsSentAgain covers this.
  • A real module fault still turns health red. Health is still unhealthy when the database is not ready (pgstore_test: "preparation step"), and when the findings or the records cannot be read. Only the whole-set findings moved to healthy.
  • Nothing is sent from the 5-minute judging process. keep() builds a KB with no Notify or Clear and only calls judgeSet, which stores findings and logs them. judgeSet never notifies, including when accept calls it under the lock. TellFindings runs only in background(), after review(), outside k.mu, and it uses only store calls that are atomic on their own.
  • Key shape. Keys look like knowledge.finding.cited.adr-0300. They are stable for a given finding.

Not blocking:

  1. On the first rollout, each decision that is already uncited (10 of them, as the condition lists them today) gets its own warning in one burst. The messenger's cap applies. It may be worth telling the operator to expect this before the rollout.
  2. A finding whose wording changes but whose key stays the same is not told again. That is fine, and it matches "once".
  3. These notices go through operator-channel.notify, so mesh-controller.conditions and doctor no longer list them. The module's health text still names them.
Verdict: approve at 56a9a1b Re-review after the rebase. My earlier approval was at 5970a5f. The `cited` guard and the race fix are unchanged in substance from that review. This review covers what is new: the operator's decision (issue 504, note 4) that a failed whole-set rule is a warning to the operator and leaves the module healthy. **Checks.** mesh/merge-gate passes (builds mesh-knowledge for novox; 4 of 4 machines compose). mesh/repo-check passes. Locally: `go vet ./...` is clean and `go test -race -count=1 ./...` passes (109 s, memory and postgres stores). **The tests fail on the old head.** With kb.go and main.go taken from the old head, the new tests do not compile (`TellFindings` and `findingPrefix` are undefined). The changed health assertion (healthy while naming the finding) would also fail against the old `Health`, which returned unhealthy. **What I checked:** - **No repeat after a restart.** `Store.Told` is an insert with on-conflict-do-nothing into the `told` table, so "told" is kept in the store. The test builds a second world on the same store, and it tells nothing again. - **A clear reaches the operator's channel.** `busClearer` sends `operator-channel.notify` with `clear: true` under the same key. The messenger's `Notify` handles `clear` with `h.cleared(key)` and uses the `resolved` line it stored from the first notify, so passing an empty `resolved` on the clear is fine. A clear that fails stays told and is retried at the next step. A notify that fails is marked untold and sent again; `TestAFindingsNoticeThatFailsIsSentAgain` covers this. - **A real module fault still turns health red.** `Health` is still unhealthy when the database is not ready (pgstore_test: "preparation step"), and when the findings or the records cannot be read. Only the whole-set findings moved to healthy. - **Nothing is sent from the 5-minute judging process.** `keep()` builds a `KB` with no `Notify` or `Clear` and only calls `judgeSet`, which stores findings and logs them. `judgeSet` never notifies, including when `accept` calls it under the lock. `TellFindings` runs only in `background()`, after `review()`, outside `k.mu`, and it uses only store calls that are atomic on their own. - **Key shape.** Keys look like `knowledge.finding.cited.adr-0300`. They are stable for a given finding. **Not blocking:** 1. On the first rollout, each decision that is already uncited (10 of them, as the condition lists them today) gets its own warning in one burst. The messenger's cap applies. It may be worth telling the operator to expect this before the rollout. 2. A finding whose wording changes but whose key stays the same is not told again. That is fine, and it matches "once". 3. These notices go through `operator-channel.notify`, so `mesh-controller.conditions` and `doctor` no longer list them. The module's health text still names them.
mesh-admin merged commit b82aa1691c into main 2026-10-12 11:44:58 +00:00
mesh-admin deleted branch fix/504-propose-refuses-uncited-record 2026-10-12 11:44:59 +00:00
Sign in to join this conversation.
No Reviewers
2 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: novox/mesh-catalog#237