Seat.Paths (compiled, never stored, like Receives): node-hostname names /etc/hosts and /etc/hostname; node-uplink names /etc/resolv.conf stepping back for node-resolver; node-resolver names /etc/resolv.conf; every other seat names none.
CanHold refuses a holder that does not declare every path its seat names (as a resource or as a file the mesh renders), naming the seat and the missing paths. That covers registration (CatalogueProblems) and handover (seat --to/--add).
At registration, a module that declares a seat's path and claims none of the seats naming it is refused. A module claiming any one of them may declare it, so the uplink and the resolver can both declare /etc/resolv.conf. Access (reading) stays allowed.
A module's own seat may name paths (SeatDeclaration.Paths), under the same rules: node seats only, literal, clean and absolute. A path another seat already names is refused, and the second declaration loses. A holder of such a seat must declare its paths.
Composition: checkResources applies the same rule to what the controller generates for a machine, so a computed module cannot write a seat's path.
The step-back is read from the seat definition (stepsBack in node_resolver.go), replacing the uplink/resolver special case. It now drops a resource on the path as well as a rendered file.
seats shows a PATHS column, and below the table who owns each seat's path on each machine, with the step-back applied (path-owners in --json).
Why
Which module owns /etc/hosts or /etc/resolv.conf could be found only by reading every manifest. A module not holding node-hostname could declare /etc/hosts on a machine where nobody held it, and nothing refused it.
How it was tested
The new internal/catalogue/seat_paths_test.go covers the rows of ADR 0287 "How it is checked":
the compiled starting set and its rules: node only, no placeholder, one owner per path, no cycle;
CanHold with the path declared as a resource, declared as a rendered file, under the operator's home, and not declared at all;
registration refusing /etc/hosts from a module that does not claim node-hostname, and accepting the uplink and the resolver together;
a module's own seat refused at mesh scope, with ${, or when it names /etc/hosts;
composition refusing a generated resource on /etc/hosts;
the step-back read from the seat;
PathOwners;
CatalogueProblems over the catalogue, with no ADR 0287 problem: run against the captured copy and against mesh-catalog main via MESH_CHECK_BESIDE.
The existing uplink/resolver composition tests pass with the special case removed. The seats listing is tested in cmd/mesh-controller/seats_test.go. With missingPaths and unclaimedSeatPaths turned off, the four refusal tests fail. sh merge-check.sh passes locally; the store tests were skipped because no postgres was available.
Fixes issue 412
## What
Builds hq ADR 0287 in `internal/catalogue`:
- `Seat.Paths` (compiled, never stored, like `Receives`): `node-hostname` names `/etc/hosts` and `/etc/hostname`; `node-uplink` names `/etc/resolv.conf` stepping back for `node-resolver`; `node-resolver` names `/etc/resolv.conf`; every other seat names none.
- `CanHold` refuses a holder that does not declare every path its seat names (as a resource or as a file the mesh renders), naming the seat and the missing paths. That covers registration (`CatalogueProblems`) and handover (`seat --to/--add`).
- At registration, a module that declares a seat's path and claims none of the seats naming it is refused. A module claiming any one of them may declare it, so the uplink and the resolver can both declare `/etc/resolv.conf`. Access (reading) stays allowed.
- A module's own seat may name `paths` (`SeatDeclaration.Paths`), under the same rules: node seats only, literal, clean and absolute. A path another seat already names is refused, and the second declaration loses. A holder of such a seat must declare its paths.
- Composition: `checkResources` applies the same rule to what the controller generates for a machine, so a computed module cannot write a seat's path.
- The step-back is read from the seat definition (`stepsBack` in `node_resolver.go`), replacing the uplink/resolver special case. It now drops a resource on the path as well as a rendered file.
- `seats` shows a PATHS column, and below the table who owns each seat's path on each machine, with the step-back applied (`path-owners` in `--json`).
## Why
Which module owns `/etc/hosts` or `/etc/resolv.conf` could be found only by reading every manifest. A module not holding `node-hostname` could declare `/etc/hosts` on a machine where nobody held it, and nothing refused it.
## How it was tested
The new `internal/catalogue/seat_paths_test.go` covers the rows of ADR 0287 "How it is checked":
- the compiled starting set and its rules: node only, no placeholder, one owner per path, no cycle;
- `CanHold` with the path declared as a resource, declared as a rendered file, under the operator's home, and not declared at all;
- registration refusing `/etc/hosts` from a module that does not claim `node-hostname`, and accepting the uplink and the resolver together;
- a module's own seat refused at mesh scope, with `${`, or when it names `/etc/hosts`;
- composition refusing a generated resource on `/etc/hosts`;
- the step-back read from the seat;
- `PathOwners`;
- `CatalogueProblems` over the catalogue, with no ADR 0287 problem: run against the captured copy and against mesh-catalog main via `MESH_CHECK_BESIDE`.
The existing uplink/resolver composition tests pass with the special case removed. The `seats` listing is tested in `cmd/mesh-controller/seats_test.go`. With `missingPaths` and `unclaimedSeatPaths` turned off, the four refusal tests fail. `sh merge-check.sh` passes locally; the store tests were skipped because no postgres was available.
Fixes issue 412
mesh/merge-gate pass: builds build-agent, mesh-controller → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without …
Which module owns /etc/hosts or /etc/resolv.conf on a machine was answered only by reading every
manifest, and a module not holding node-hostname could declare /etc/hosts where nobody held it.
hq ADR 0287 makes the paths part of the seat's definition: node-hostname names /etc/hosts and
/etc/hostname, node-uplink names /etc/resolv.conf stepping back for node-resolver, node-resolver
names /etc/resolv.conf.
CanHold refuses a holder that does not declare every path (registration and handover), registration
refuses a module declaring a seat's path while claiming none of the seats naming it, a module's own
seat names paths under the same rules (node only, literal, not named by another seat), composition
refuses a generated resource on a seat's path, the uplink's step-back is read from the seat instead
of a special case, and `seats` shows each seat's paths and who owns each on each machine.
build-agent keeps data (ADR 0233): what it holds is backed up before it moves
route-proxy provides route: its consumers are sent again after it
route-proxy keeps data (ADR 0233): what it holds is backed up before it moves
grants step: if this changes what the bus's user list says (a new tool, subject or user), novox is sent that list alone, every build there kept, before build-agent, route-proxy is judged on its first machine
Machines
mesh-controller on novox: passed — healthy 3 times over 4m7s
mesh-controller on the rest running it: sent
build-agent on ace: passed — healthy 1 times over 1m39s
build-agent on the rest running it: sent
Transitions
2026-10-12 15:09 (new) → proposed (announced): novox/mesh-controller#267's head announced
2026-10-12 15:16 proposed → checked (checked): the verdict names this commit
2026-10-12 15:16 checked → ready (accepted): the gate passed or warned, and the repository's own check did not fail
2026-10-12 19:22 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 19:22 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 19:31 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-controller@24347ea5a02b` — **delivered** since 2026-10-12T19:31:22Z
**Delivery plan** — builds build-agent, mesh-controller, route-proxy → ace, g14, novox, shanks; no bus step
- tier 0: mesh-controller
- tier 1: build-agent
- tier 2: route-proxy
- ace: receives build-agent, route-proxy
- g14: receives build-agent
- novox: receives mesh-controller, build-agent, route-proxy
- shanks: receives build-agent
- build-agent keeps data (ADR 0233): what it holds is backed up before it moves
- route-proxy provides route: its consumers are sent again after it
- route-proxy keeps data (ADR 0233): what it holds is backed up before it moves
- grants step: if this changes what the bus's user list says (a new tool, subject or user), novox is sent that list alone, every build there kept, before build-agent, route-proxy is judged on its first machine
**Machines**
- mesh-controller on novox: passed — healthy 3 times over 4m7s
- mesh-controller on the rest running it: sent
- build-agent on ace: passed — healthy 1 times over 1m39s
- build-agent on the rest running it: sent
**Transitions**
- 2026-10-12 15:09 (new) → proposed (announced): novox/mesh-controller#267's head announced
- 2026-10-12 15:16 proposed → checked (checked): the verdict names this commit
- 2026-10-12 15:16 checked → ready (accepted): the gate passed or warned, and the repository's own check did not fail
- 2026-10-12 19:22 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 19:22 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 19:31 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: build-agent, mesh-controller)
mesh/repo-check: PASS — its merge-check.sh passed
Delivery plan — builds build-agent, mesh-controller → ace, g14, novox, shanks; no bus step
tier 0: mesh-controller
tier 1: build-agent
ace: receives build-agent
g14: receives build-agent
novox: receives mesh-controller, build-agent
shanks: receives build-agent
build-agent keeps data (ADR 0233): what it holds is backed up before it moves
grants step: if this changes what the bus's user list says (a new tool, subject or user), novox is sent that list alone, every build there kept, before build-agent is judged on its first machine
read by no module's build: cmd/mesh-controller/seats_test.go, internal/catalogue/seat_paths_test.go, internal/catalogue/uplink_verbs_test.go
Run by the build seat on novox as build-1791814810148047906 (builds --log build-1791814810148047906).
--- the gate: build-agent, mesh-controller
mesh-controller: ok, 58 tool(s); WARNING: merge-window ask(s) for a setting and do(es) not say whether it is trusted, so it counts as trusted: set at the terminal alone; say "trusted" false where nothing trusts it (novox/hq issue 339); WARNING: controller stay(s) up and say(s) not how it is ready — judged by liveness alone, refused from 2026-11-18 (ADR 0240 rule 8)
long-running resources without health: 1
files asking for a setting without saying whether it is trusted: 1
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 1 module(s) changed against 4 machine(s), as the snapshot of 2026-10-12T14:13:26Z 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 2026-10-12T14:13:26Z
modules the change moves:
- mesh-controller (changed)
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 2 module(s) in 2 tier(s): build-agent, mesh-controller
notes:
- cmd/mesh-controller/seats_test.go, internal/catalogue/seat_paths_test.go and internal/catalogue/uplink_verbs_test.go read by no module's build, and rebuild nothing (issue 280)
--- its merge-check.sh (go toolchain)
ok github.com/novox/mesh-controller/cmd/mesh-builder 1.081s
ok github.com/novox/mesh-controller/cmd/mesh-controller 332.121s
ok github.com/novox/mesh-controller/examples/objectstore-provisioner 1.019s
ok github.com/novox/mesh-controller/examples/postgres-provisioner 1.027s
ok github.com/novox/mesh-controller/examples/redis-provisioner 1.015s
ok github.com/novox/mesh-controller/examples/route-proxy 1.143s
ok github.com/novox/mesh-controller/internal/artifacts 1.060s
? github.com/novox/mesh-controller/internal/beside [no test files]
ok github.com/novox/mesh-controller/internal/broker 10.204s
ok github.com/novox/mesh-controller/internal/builder 67.444s
ok github.com/novox/mesh-controller/internal/catalogue 8.845s
ok github.com/novox/mesh-controller/internal/conditions 2.640s
ok github.com/novox/mesh-controller/internal/envfile 1.039s
ok github.com/novox/mesh-controller/internal/facts 1.039s
ok github.com/novox/mesh-controller/internal/identity 11.701s
ok github.com/novox/mesh-controller/internal/inventory 138.468s
ok github.com/novox/mesh-controller/internal/lease 17.700s
ok github.com/novox/mesh-controller/internal/licences 10.345s
ok github.com/novox/mesh-controller/internal/licences/adapters 1.015s
ok github.com/novox/mesh-controller/internal/link 72.152s
ok github.com/novox/mesh-controller/internal/lint 2.536s
ok github.com/novox/mesh-controller/internal/outward 1.042s
ok github.com/novox/mesh-controller/internal/overlay 1.037s
ok github.com/novox/mesh-controller/internal/secrets 1.504s
ok github.com/novox/mesh-controller/internal/store 7.751s
ok github.com/novox/mesh-controller/internal/testbus 1.193s
ok github.com/novox/mesh-controller/internal/token 1.016s
store tests, each run against the store:
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeWalksTheOthersAgain (0.49s)
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeEndsTheMergesThatContainIt (0.47s)
./cmd/mesh-controller --- PASS: TestAStopWalksTheOthersAgainInTheOpenBatch (0.46s)
./cmd/mesh-controller --- PASS: TestAStopNamesWhatStaysOnTheMachinesAndWhatStillCarriesIt (0.52s)
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeLeavesItOutOfABatch (0.46s)
./cmd/mesh-controller --- PASS: TestSavingPlansAndMovingMergesIsOneAct (0.48s)
./cmd/mesh-controller --- PASS: TestNodeShowAndStatusReadTheLastSendFromTheStore (0.53s)
./cmd/mesh-controller --- PASS: TestTheConditionsVerbReadsTheLooksTheBucketNoLongerHolds (0.33s)
./cmd/mesh-controller --- PASS: TestAWalkLetGoAndStoppedByVerbsIsSaidStartedAndEnded (0.56s)
./cmd/mesh-controller --- PASS: TestAWalkStoppedOrClosedByAPersonIsSaidEndedStopped (0.65s)
./cmd/mesh-controller --- PASS: TestAWalkRetriedByAPersonIsSaidStartedAgain (0.59s)
./cmd/mesh-controller --- PASS: TestABacklogReleasedByAPersonIsSaidStarted (0.48s)
./cmd/mesh-controller --- PASS: TestAWalkEndedByCancelOrClearIsSaidEnded (0.65s)
./cmd/mesh-controller --- PASS: TestAWalkRestartedByARebuildIsSaidStarted (0.51s)
./cmd/mesh-controller --- PASS: TestAWalkStoppedByHandAndRetriedEndsFailedWhenItFailsAgain (0.52s)
./cmd/mesh-controller --- PASS: TestReplay237 (0.52s)
./cmd/mesh-controller --- PASS: TestAnUnassignThatLeavesItsNodeUnresolvedSaysSoOnceInItsAnswer (0.49s)
./cmd/mesh-controller --- PASS: TestAnUnassignThatLeavesItsNodeResolvedSaysNothingOfIt (0.48s)
./cmd/mesh-controller --- PASS: TestAnAssignThatLeavesItsNodeUnresolvedSaysOnlyThePlansRefusal (0.50s)
./cmd/mesh-controller --- PASS: TestADirectoryOwnersKeyIsRefusedThroughAVerb (0.57s)
./internal/inventory --- PASS: TestTheLastSendSaysWhatTheMachineAnswered (0.28s)
./internal/inventory --- PASS: TestAReportIsTheLastSendsAnswerOnlyWhenItIsAboutThatSend (0.29s)
./internal/inventory --- PASS: TestAnIssuedValueIsSealedToEveryHolderOnceAndNotAgain (0.32s)
./internal/inventory --- PASS: TestAnIssuedValueIsSealedAgainAfterAKeyChangeOrAnAcceptance (0.30s)
./internal/inventory --- PASS: TestAConditionsLooksAreKeptBesideTheBucket (0.27s)
./internal/inventory --- PASS: TestAWalksStartIsKeptAndARefusedSaveLeavesItUnsaid (0.29s)
**Merge check** at `a3bf5da9`
- `mesh/merge-gate`: **PASS** — every machine composes with the change as it did without (4 of 4 compose) (modules: build-agent, mesh-controller)
- `mesh/repo-check`: **PASS** — its merge-check.sh passed
**Delivery plan** — builds build-agent, mesh-controller → ace, g14, novox, shanks; no bus step
- tier 0: mesh-controller
- tier 1: build-agent
- ace: receives build-agent
- g14: receives build-agent
- novox: receives mesh-controller, build-agent
- shanks: receives build-agent
- build-agent keeps data (ADR 0233): what it holds is backed up before it moves
- grants step: if this changes what the bus's user list says (a new tool, subject or user), novox is sent that list alone, every build there kept, before build-agent is judged on its first machine
- read by no module's build: cmd/mesh-controller/seats_test.go, internal/catalogue/seat_paths_test.go, internal/catalogue/uplink_verbs_test.go
Run by the build seat on novox as `build-1791814810148047906` (`builds --log build-1791814810148047906`).
```
--- the gate: build-agent, mesh-controller
mesh-controller: ok, 58 tool(s); WARNING: merge-window ask(s) for a setting and do(es) not say whether it is trusted, so it counts as trusted: set at the terminal alone; say "trusted" false where nothing trusts it (novox/hq issue 339); WARNING: controller stay(s) up and say(s) not how it is ready — judged by liveness alone, refused from 2026-11-18 (ADR 0240 rule 8)
long-running resources without health: 1
files asking for a setting without saying whether it is trusted: 1
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 1 module(s) changed against 4 machine(s), as the snapshot of 2026-10-12T14:13:26Z 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 2026-10-12T14:13:26Z
modules the change moves:
- mesh-controller (changed)
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 2 module(s) in 2 tier(s): build-agent, mesh-controller
notes:
- cmd/mesh-controller/seats_test.go, internal/catalogue/seat_paths_test.go and internal/catalogue/uplink_verbs_test.go read by no module's build, and rebuild nothing (issue 280)
--- its merge-check.sh (go toolchain)
ok github.com/novox/mesh-controller/cmd/mesh-builder 1.081s
ok github.com/novox/mesh-controller/cmd/mesh-controller 332.121s
ok github.com/novox/mesh-controller/examples/objectstore-provisioner 1.019s
ok github.com/novox/mesh-controller/examples/postgres-provisioner 1.027s
ok github.com/novox/mesh-controller/examples/redis-provisioner 1.015s
ok github.com/novox/mesh-controller/examples/route-proxy 1.143s
ok github.com/novox/mesh-controller/internal/artifacts 1.060s
? github.com/novox/mesh-controller/internal/beside [no test files]
ok github.com/novox/mesh-controller/internal/broker 10.204s
ok github.com/novox/mesh-controller/internal/builder 67.444s
ok github.com/novox/mesh-controller/internal/catalogue 8.845s
ok github.com/novox/mesh-controller/internal/conditions 2.640s
ok github.com/novox/mesh-controller/internal/envfile 1.039s
ok github.com/novox/mesh-controller/internal/facts 1.039s
ok github.com/novox/mesh-controller/internal/identity 11.701s
ok github.com/novox/mesh-controller/internal/inventory 138.468s
ok github.com/novox/mesh-controller/internal/lease 17.700s
ok github.com/novox/mesh-controller/internal/licences 10.345s
ok github.com/novox/mesh-controller/internal/licences/adapters 1.015s
ok github.com/novox/mesh-controller/internal/link 72.152s
ok github.com/novox/mesh-controller/internal/lint 2.536s
ok github.com/novox/mesh-controller/internal/outward 1.042s
ok github.com/novox/mesh-controller/internal/overlay 1.037s
ok github.com/novox/mesh-controller/internal/secrets 1.504s
ok github.com/novox/mesh-controller/internal/store 7.751s
ok github.com/novox/mesh-controller/internal/testbus 1.193s
ok github.com/novox/mesh-controller/internal/token 1.016s
store tests, each run against the store:
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeWalksTheOthersAgain (0.49s)
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeEndsTheMergesThatContainIt (0.47s)
./cmd/mesh-controller --- PASS: TestAStopWalksTheOthersAgainInTheOpenBatch (0.46s)
./cmd/mesh-controller --- PASS: TestAStopNamesWhatStaysOnTheMachinesAndWhatStillCarriesIt (0.52s)
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeLeavesItOutOfABatch (0.46s)
./cmd/mesh-controller --- PASS: TestSavingPlansAndMovingMergesIsOneAct (0.48s)
./cmd/mesh-controller --- PASS: TestNodeShowAndStatusReadTheLastSendFromTheStore (0.53s)
./cmd/mesh-controller --- PASS: TestTheConditionsVerbReadsTheLooksTheBucketNoLongerHolds (0.33s)
./cmd/mesh-controller --- PASS: TestAWalkLetGoAndStoppedByVerbsIsSaidStartedAndEnded (0.56s)
./cmd/mesh-controller --- PASS: TestAWalkStoppedOrClosedByAPersonIsSaidEndedStopped (0.65s)
./cmd/mesh-controller --- PASS: TestAWalkRetriedByAPersonIsSaidStartedAgain (0.59s)
./cmd/mesh-controller --- PASS: TestABacklogReleasedByAPersonIsSaidStarted (0.48s)
./cmd/mesh-controller --- PASS: TestAWalkEndedByCancelOrClearIsSaidEnded (0.65s)
./cmd/mesh-controller --- PASS: TestAWalkRestartedByARebuildIsSaidStarted (0.51s)
./cmd/mesh-controller --- PASS: TestAWalkStoppedByHandAndRetriedEndsFailedWhenItFailsAgain (0.52s)
./cmd/mesh-controller --- PASS: TestReplay237 (0.52s)
./cmd/mesh-controller --- PASS: TestAnUnassignThatLeavesItsNodeUnresolvedSaysSoOnceInItsAnswer (0.49s)
./cmd/mesh-controller --- PASS: TestAnUnassignThatLeavesItsNodeResolvedSaysNothingOfIt (0.48s)
./cmd/mesh-controller --- PASS: TestAnAssignThatLeavesItsNodeUnresolvedSaysOnlyThePlansRefusal (0.50s)
./cmd/mesh-controller --- PASS: TestADirectoryOwnersKeyIsRefusedThroughAVerb (0.57s)
./internal/inventory --- PASS: TestTheLastSendSaysWhatTheMachineAnswered (0.28s)
./internal/inventory --- PASS: TestAReportIsTheLastSendsAnswerOnlyWhenItIsAboutThatSend (0.29s)
./internal/inventory --- PASS: TestAnIssuedValueIsSealedToEveryHolderOnceAndNotAgain (0.32s)
./internal/inventory --- PASS: TestAnIssuedValueIsSealedAgainAfterAKeyChangeOrAnAcceptance (0.30s)
./internal/inventory --- PASS: TestAConditionsLooksAreKeptBesideTheBucket (0.27s)
./internal/inventory --- PASS: TestAWalksStartIsKeptAndARefusedSaveLeavesItUnsaid (0.29s)
```
Independent review for issue 412 / hq ADR 0287. Two small defects need fixing before merge. Neither one touches today's mesh, but both open gaps in the very rules this PR adds.
Blocking
1. The step-back now drops a resource in the check but not in the declaration. stepsBack (node_resolver.go) now also leaves out a resource on the seat's path. Only checkResources sees that result through steppedBack. The declaration itself (declaration.go:1124) calls stepsBack only for FactsFrom; the module's own resources are copied from the unstepped m.Resources.
I checked this with a scratch test: systemd-networkd with /etc/resolv.conf declared as a resource, plus a node-resolver holder rendering it. On this head, Resolve accepts it and Declaration composes the file twice (systemd-networkd.resolvers and local-resolver.fact-resolvers). On the base (9eddf36) the same Resolve is refused.
So the check now passes a machine that composes two writers of the resolver file, where it used to refuse it. This is not live today: every uplink holder (networkmanager, systemd-networkd) renders the file as a fact.
Fix: apply the same step-back to the resource loop in Declaration, or keep the step-back to rendered files only. Add a declaration-level test with the uplink declaring the path as a resource; TestTheUplinkStepsBackByTheSeatsDefinition covers only stepsBack and checkResources.
2. A path that is not clean gets past "only a claimant declares a seat's path". declaredPaths compares a resource's path byte for byte. A module that claims nothing and declares any of these as a file resource passes both CatalogueProblems and checkResources (I checked each one through ParseManifest):
/etc//hosts
/etc/./hosts
/etc/../etc/hosts
/etc/hosts/
The machine then writes /etc/hosts. A seat's own path is required to be clean, but a declared path is not.
Fix: compare path.Clean(p) in declaredPaths, which missingPaths and unclaimedSeatPaths both use. Add one of these paths to TestOnlyAClaimantDeclaresASeatsPath.
No manifest in mesh-catalog or mesh-media-catalog main uses such a path today.
Checked and fine
Check table (item 1). Every row of ADR 0287's table has a test. I ran 13 mutations, one per rule; each was caught by at least one test:
seatPathProblem, SeatPathProblems, missingPaths, unclaimedSeatPaths and seatPathsOnMachine made to return nothing;
the module-seat conflict check, the module-seat holder paths check and declaredSeatPathProblems turned off;
the step-back of resources and the step-back of facts turned off;
UseSeats not copying Paths, and the PathOwners step-back turned off.
One small gap: the handover row is proven through CanHold directly, not through seat --to. The command calls the same CanHold, so I accept it.
Today's catalogues (item 2).module check over mesh-catalog main and over mesh-media-catalog main gives byte-identical output with the base, main and PR binaries, and no ADR 0287 problem.
I also composed checkResources with the manifests of each node's current assignment (ace, g14, novox, shanks, from mesh-controller.node). Base, main and the PR each give 0 problems on every node.
Nine assigned modules have no manifest in either catalogue (mesh-host, node-tools, the mesh-wireguard overlay generator, and six app repositories). node-tools' manifest names none of the three paths, and the overlay generator writes only /etc/wireguard/mesh0.conf.
A composition refusal from this PR depends only on a module's own declared paths against the seats' paths, not on where it is assigned. So no live node can be refused by it.
Same step-back result (item 3). g14 and shanks run networkmanager and systemd-resolved. On base and PR, the uplink loses only its resolvers fact and keeps its three resources, and systemd-resolved keeps the file. ace and novox are unchanged.
Dropping the resolver file (item 4). The path is dropped only when the node-resolver holder on that machine declares it, so the file always keeps one writer in the composition. No uplink declares it as a resource today. Finding 1 is the remaining risk: the opposite of losing the file, it is composed twice.
Module-declared seats (item 5). Paths must be literal, clean and absolute, on a node seat only. A path already named loses to the first declaration, and a module's seat never steps back. Only literal comparison is applied; see finding 2 for the declared side.
Minor: a module seat may name / or /etc. That is harmless with literal matching, but the "role's own path" clause is not machine-checked.
seats --json (item 6). The change only adds: per-row paths (omitempty) and a top-level path-owners (omitempty). Every existing key is unchanged. The seats seat verb passes the JSON through.
Local checks (item 7). gofmt and go vet are clean. go test -race ./... passes, apart from TestATierPutsABrokenModuleBackAtOnce/gate_kept_on_node-tools in cmd/mesh-controller. That is delivery-plan code this PR does not touch, and it failed once on timing while another race run was loading the machine. It passed 3 of 3 runs alone on this head.
All 26 store tests in testdata/store-tests pass against a throwaway postgres:17 (MESH_TEST_POSTGRES).
The PR merged locally onto current main (3cfc121, 7 commits ahead of its base) builds, vets, and passes the catalogue and cmd suites.
Statuses on the head (item 8). mesh/merge-gate success ("every machine composes with the change as it did without"), mesh/repo-check success, mesh/delivery success ("ready"). The forge says it is mergeable.
Once 1 and 2 are fixed, each with a test, I expect to approve.
Verdict: request changes at a3bf5da9b1657dbb77e7ba63d84c614c56592d90
Independent review for issue 412 / hq ADR 0287. Two small defects need fixing before merge. Neither one touches today's mesh, but both open gaps in the very rules this PR adds.
## Blocking
**1. The step-back now drops a resource in the check but not in the declaration.**
`stepsBack` (node_resolver.go) now also leaves out a resource on the seat's path. Only `checkResources` sees that result through `steppedBack`. The declaration itself (declaration.go:1124) calls `stepsBack` only for `FactsFrom`; the module's own resources are copied from the unstepped `m.Resources`.
I checked this with a scratch test: systemd-networkd with `/etc/resolv.conf` declared as a *resource*, plus a node-resolver holder rendering it. On this head, `Resolve` accepts it and `Declaration` composes the file twice (`systemd-networkd.resolvers` and `local-resolver.fact-resolvers`). On the base (9eddf36) the same `Resolve` is refused.
So the check now passes a machine that composes two writers of the resolver file, where it used to refuse it. This is not live today: every uplink holder (networkmanager, systemd-networkd) renders the file as a fact.
Fix: apply the same step-back to the resource loop in `Declaration`, or keep the step-back to rendered files only. Add a declaration-level test with the uplink declaring the path as a resource; `TestTheUplinkStepsBackByTheSeatsDefinition` covers only `stepsBack` and `checkResources`.
**2. A path that is not clean gets past "only a claimant declares a seat's path".**
`declaredPaths` compares a resource's path byte for byte. A module that claims nothing and declares any of these as a file resource passes both `CatalogueProblems` and `checkResources` (I checked each one through `ParseManifest`):
- `/etc//hosts`
- `/etc/./hosts`
- `/etc/../etc/hosts`
- `/etc/hosts/`
The machine then writes `/etc/hosts`. A seat's own path is required to be clean, but a declared path is not.
Fix: compare `path.Clean(p)` in `declaredPaths`, which `missingPaths` and `unclaimedSeatPaths` both use. Add one of these paths to `TestOnlyAClaimantDeclaresASeatsPath`.
No manifest in mesh-catalog or mesh-media-catalog main uses such a path today.
## Checked and fine
- **Check table (item 1).** Every row of ADR 0287's table has a test. I ran 13 mutations, one per rule; each was caught by at least one test:
- `seatPathProblem`, `SeatPathProblems`, `missingPaths`, `unclaimedSeatPaths` and `seatPathsOnMachine` made to return nothing;
- the module-seat conflict check, the module-seat holder paths check and `declaredSeatPathProblems` turned off;
- the step-back of resources and the step-back of facts turned off;
- `UseSeats` not copying `Paths`, and the `PathOwners` step-back turned off.
One small gap: the handover row is proven through `CanHold` directly, not through `seat --to`. The command calls the same `CanHold`, so I accept it.
- **Today's catalogues (item 2).** `module check` over mesh-catalog main and over mesh-media-catalog main gives byte-identical output with the base, main and PR binaries, and no ADR 0287 problem.
I also composed `checkResources` with the manifests of each node's current assignment (ace, g14, novox, shanks, from `mesh-controller.node`). Base, main and the PR each give 0 problems on every node.
Nine assigned modules have no manifest in either catalogue (mesh-host, node-tools, the mesh-wireguard overlay generator, and six app repositories). node-tools' manifest names none of the three paths, and the overlay generator writes only `/etc/wireguard/mesh0.conf`.
A composition refusal from this PR depends only on a module's own declared paths against the seats' paths, not on where it is assigned. So no live node can be refused by it.
- **Same step-back result (item 3).** g14 and shanks run networkmanager and systemd-resolved. On base and PR, the uplink loses only its `resolvers` fact and keeps its three resources, and systemd-resolved keeps the file. ace and novox are unchanged.
- **Dropping the resolver file (item 4).** The path is dropped only when the node-resolver holder on that machine declares it, so the file always keeps one writer in the composition. No uplink declares it as a resource today. Finding 1 is the remaining risk: the opposite of losing the file, it is composed twice.
- **Module-declared seats (item 5).** Paths must be literal, clean and absolute, on a node seat only. A path already named loses to the first declaration, and a module's seat never steps back. Only literal comparison is applied; see finding 2 for the declared side.
Minor: a module seat may name `/` or `/etc`. That is harmless with literal matching, but the "role's own path" clause is not machine-checked.
- **`seats --json` (item 6).** The change only adds: per-row `paths` (omitempty) and a top-level `path-owners` (omitempty). Every existing key is unchanged. The `seats` seat verb passes the JSON through.
- **Local checks (item 7).** gofmt and `go vet` are clean.
`go test -race ./...` passes, apart from `TestATierPutsABrokenModuleBackAtOnce/gate_kept_on_node-tools` in cmd/mesh-controller. That is delivery-plan code this PR does not touch, and it failed once on timing while another race run was loading the machine. It passed 3 of 3 runs alone on this head.
All 26 store tests in testdata/store-tests pass against a throwaway postgres:17 (`MESH_TEST_POSTGRES`).
The PR merged locally onto current main (3cfc121, 7 commits ahead of its base) builds, vets, and passes the catalogue and cmd suites.
- **Statuses on the head (item 8).** mesh/merge-gate success ("every machine composes with the change as it did without"), mesh/repo-check success, mesh/delivery success ("ready"). The forge says it is mergeable.
Once 1 and 2 are fixed, each with a test, I expect to approve.
mesh/merge-gate head 24347ea5a02b merged into trunk 84deb530ee5a — pass: builds build-agent, mesh-controller → ace, g14, novox, shanks; no bus step; every …
A resource on the resolver file was left out by the check and still composed by the declaration, so an
uplink declaring it as a resource beside the resolver passed with two writers; it is refused again, as
before. And /etc//hosts or /etc/../etc/hosts is /etc/hosts to the machine, so it is compared cleaned.
A seat names no root or top-level directory.
build-agent keeps data (ADR 0233): what it holds is backed up before it moves
route-proxy provides route: its consumers are sent again after it
route-proxy keeps data (ADR 0233): what it holds is backed up before it moves
grants step: if this changes what the bus's user list says (a new tool, subject or user), novox is sent that list alone, every build there kept, before build-agent, route-proxy is judged on its first machine
Run by the build seat on g14 as build-1791817762661447750 (builds --log build-1791817762661447750).
--- the gate: build-agent, mesh-controller, route-proxy
mesh-controller: ok, 58 tool(s); WARNING: merge-window ask(s) for a setting and do(es) not say whether it is trusted, so it counts as trusted: set at the terminal alone; say "trusted" false where nothing trusts it (novox/hq issue 339); WARNING: controller stay(s) up and say(s) not how it is ready — judged by liveness alone, refused from 2026-11-18 (ADR 0240 rule 8)
long-running resources without health: 1
files asking for a setting without saying whether it is trusted: 1
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 1 module(s) changed against 4 machine(s), as the snapshot of 2026-10-12T14:54:57Z 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 2026-10-12T14:54:57Z
modules the change moves:
- mesh-controller (changed)
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 3 module(s) in 3 tier(s): build-agent, mesh-controller, route-proxy
--- its merge-check.sh (go toolchain)
ok github.com/novox/mesh-controller/cmd/mesh-builder 1.051s
ok github.com/novox/mesh-controller/cmd/mesh-controller 271.949s
ok github.com/novox/mesh-controller/examples/objectstore-provisioner 1.015s
ok github.com/novox/mesh-controller/examples/postgres-provisioner 1.028s
ok github.com/novox/mesh-controller/examples/redis-provisioner 1.016s
ok github.com/novox/mesh-controller/examples/route-proxy 1.223s
ok github.com/novox/mesh-controller/internal/artifacts 1.060s
? github.com/novox/mesh-controller/internal/beside [no test files]
ok github.com/novox/mesh-controller/internal/broker 4.803s
ok github.com/novox/mesh-controller/internal/builder 33.636s
ok github.com/novox/mesh-controller/internal/catalogue 8.646s
ok github.com/novox/mesh-controller/internal/conditions 2.645s
ok github.com/novox/mesh-controller/internal/envfile 1.042s
ok github.com/novox/mesh-controller/internal/facts 1.034s
ok github.com/novox/mesh-controller/internal/identity 5.269s
ok github.com/novox/mesh-controller/internal/inventory 90.845s
ok github.com/novox/mesh-controller/internal/lease 17.710s
ok github.com/novox/mesh-controller/internal/licences 4.808s
ok github.com/novox/mesh-controller/internal/licences/adapters 1.016s
ok github.com/novox/mesh-controller/internal/link 59.823s
ok github.com/novox/mesh-controller/internal/lint 2.714s
ok github.com/novox/mesh-controller/internal/outward 1.057s
ok github.com/novox/mesh-controller/internal/overlay 1.041s
ok github.com/novox/mesh-controller/internal/secrets 1.616s
ok github.com/novox/mesh-controller/internal/store 3.591s
ok github.com/novox/mesh-controller/internal/testbus 1.238s
ok github.com/novox/mesh-controller/internal/token 1.028s
store tests, each run against the store:
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeWalksTheOthersAgain (0.42s)
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeEndsTheMergesThatContainIt (0.42s)
./cmd/mesh-controller --- PASS: TestAStopWalksTheOthersAgainInTheOpenBatch (0.43s)
./cmd/mesh-controller --- PASS: TestAStopNamesWhatStaysOnTheMachinesAndWhatStillCarriesIt (0.40s)
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeLeavesItOutOfABatch (0.42s)
./cmd/mesh-controller --- PASS: TestSavingPlansAndMovingMergesIsOneAct (0.39s)
./cmd/mesh-controller --- PASS: TestNodeShowAndStatusReadTheLastSendFromTheStore (0.43s)
./cmd/mesh-controller --- PASS: TestTheConditionsVerbReadsTheLooksTheBucketNoLongerHolds (0.29s)
./cmd/mesh-controller --- PASS: TestAWalkLetGoAndStoppedByVerbsIsSaidStartedAndEnded (0.45s)
./cmd/mesh-controller --- PASS: TestAWalkStoppedOrClosedByAPersonIsSaidEndedStopped (0.45s)
./cmd/mesh-controller --- PASS: TestAWalkRetriedByAPersonIsSaidStartedAgain (0.44s)
./cmd/mesh-controller --- PASS: TestABacklogReleasedByAPersonIsSaidStarted (0.39s)
./cmd/mesh-controller --- PASS: TestAWalkEndedByCancelOrClearIsSaidEnded (0.44s)
./cmd/mesh-controller --- PASS: TestAWalkRestartedByARebuildIsSaidStarted (0.39s)
./cmd/mesh-controller --- PASS: TestAWalkStoppedByHandAndRetriedEndsFailedWhenItFailsAgain (0.43s)
./cmd/mesh-controller --- PASS: TestReplay237 (0.43s)
./cmd/mesh-controller --- PASS: TestAnUnassignThatLeavesItsNodeUnresolvedSaysSoOnceInItsAnswer (0.40s)
./cmd/mesh-controller --- PASS: TestAnUnassignThatLeavesItsNodeResolvedSaysNothingOfIt (0.40s)
./cmd/mesh-controller --- PASS: TestAnAssignThatLeavesItsNodeUnresolvedSaysOnlyThePlansRefusal (0.43s)
./cmd/mesh-controller --- PASS: TestADirectoryOwnersKeyIsRefusedThroughAVerb (0.44s)
./internal/inventory --- PASS: TestTheLastSendSaysWhatTheMachineAnswered (0.23s)
./internal/inventory --- PASS: TestAReportIsTheLastSendsAnswerOnlyWhenItIsAboutThatSend (0.24s)
./internal/inventory --- PASS: TestAnIssuedValueIsSealedToEveryHolderOnceAndNotAgain (0.24s)
./internal/inventory --- PASS: TestAnIssuedValueIsSealedAgainAfterAKeyChangeOrAnAcceptance (0.24s)
./internal/inventory --- PASS: TestAConditionsLooksAreKeptBesideTheBucket (0.22s)
./internal/inventory --- PASS: TestAWalksStartIsKeptAndARefusedSaveLeavesItUnsaid (0.23s)
**Merge check** at `24347ea5`
- `mesh/merge-gate`: **PASS** — every machine composes with the change as it did without (4 of 4 compose) (modules: build-agent, mesh-controller, route-proxy)
- `mesh/repo-check`: **PASS** — its merge-check.sh passed
**Delivery plan** — builds build-agent, mesh-controller, route-proxy → ace, g14, novox, shanks; no bus step
- tier 0: mesh-controller
- tier 1: build-agent
- tier 2: route-proxy
- ace: receives build-agent, route-proxy
- g14: receives build-agent
- novox: receives mesh-controller, build-agent, route-proxy
- shanks: receives build-agent
- build-agent keeps data (ADR 0233): what it holds is backed up before it moves
- route-proxy provides route: its consumers are sent again after it
- route-proxy keeps data (ADR 0233): what it holds is backed up before it moves
- grants step: if this changes what the bus's user list says (a new tool, subject or user), novox is sent that list alone, every build there kept, before build-agent, route-proxy is judged on its first machine
Run by the build seat on g14 as `build-1791817762661447750` (`builds --log build-1791817762661447750`).
```
--- the gate: build-agent, mesh-controller, route-proxy
mesh-controller: ok, 58 tool(s); WARNING: merge-window ask(s) for a setting and do(es) not say whether it is trusted, so it counts as trusted: set at the terminal alone; say "trusted" false where nothing trusts it (novox/hq issue 339); WARNING: controller stay(s) up and say(s) not how it is ready — judged by liveness alone, refused from 2026-11-18 (ADR 0240 rule 8)
long-running resources without health: 1
files asking for a setting without saying whether it is trusted: 1
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 1 module(s) changed against 4 machine(s), as the snapshot of 2026-10-12T14:54:57Z 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 2026-10-12T14:54:57Z
modules the change moves:
- mesh-controller (changed)
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 3 module(s) in 3 tier(s): build-agent, mesh-controller, route-proxy
--- its merge-check.sh (go toolchain)
ok github.com/novox/mesh-controller/cmd/mesh-builder 1.051s
ok github.com/novox/mesh-controller/cmd/mesh-controller 271.949s
ok github.com/novox/mesh-controller/examples/objectstore-provisioner 1.015s
ok github.com/novox/mesh-controller/examples/postgres-provisioner 1.028s
ok github.com/novox/mesh-controller/examples/redis-provisioner 1.016s
ok github.com/novox/mesh-controller/examples/route-proxy 1.223s
ok github.com/novox/mesh-controller/internal/artifacts 1.060s
? github.com/novox/mesh-controller/internal/beside [no test files]
ok github.com/novox/mesh-controller/internal/broker 4.803s
ok github.com/novox/mesh-controller/internal/builder 33.636s
ok github.com/novox/mesh-controller/internal/catalogue 8.646s
ok github.com/novox/mesh-controller/internal/conditions 2.645s
ok github.com/novox/mesh-controller/internal/envfile 1.042s
ok github.com/novox/mesh-controller/internal/facts 1.034s
ok github.com/novox/mesh-controller/internal/identity 5.269s
ok github.com/novox/mesh-controller/internal/inventory 90.845s
ok github.com/novox/mesh-controller/internal/lease 17.710s
ok github.com/novox/mesh-controller/internal/licences 4.808s
ok github.com/novox/mesh-controller/internal/licences/adapters 1.016s
ok github.com/novox/mesh-controller/internal/link 59.823s
ok github.com/novox/mesh-controller/internal/lint 2.714s
ok github.com/novox/mesh-controller/internal/outward 1.057s
ok github.com/novox/mesh-controller/internal/overlay 1.041s
ok github.com/novox/mesh-controller/internal/secrets 1.616s
ok github.com/novox/mesh-controller/internal/store 3.591s
ok github.com/novox/mesh-controller/internal/testbus 1.238s
ok github.com/novox/mesh-controller/internal/token 1.028s
store tests, each run against the store:
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeWalksTheOthersAgain (0.42s)
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeEndsTheMergesThatContainIt (0.42s)
./cmd/mesh-controller --- PASS: TestAStopWalksTheOthersAgainInTheOpenBatch (0.43s)
./cmd/mesh-controller --- PASS: TestAStopNamesWhatStaysOnTheMachinesAndWhatStillCarriesIt (0.40s)
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeLeavesItOutOfABatch (0.42s)
./cmd/mesh-controller --- PASS: TestSavingPlansAndMovingMergesIsOneAct (0.39s)
./cmd/mesh-controller --- PASS: TestNodeShowAndStatusReadTheLastSendFromTheStore (0.43s)
./cmd/mesh-controller --- PASS: TestTheConditionsVerbReadsTheLooksTheBucketNoLongerHolds (0.29s)
./cmd/mesh-controller --- PASS: TestAWalkLetGoAndStoppedByVerbsIsSaidStartedAndEnded (0.45s)
./cmd/mesh-controller --- PASS: TestAWalkStoppedOrClosedByAPersonIsSaidEndedStopped (0.45s)
./cmd/mesh-controller --- PASS: TestAWalkRetriedByAPersonIsSaidStartedAgain (0.44s)
./cmd/mesh-controller --- PASS: TestABacklogReleasedByAPersonIsSaidStarted (0.39s)
./cmd/mesh-controller --- PASS: TestAWalkEndedByCancelOrClearIsSaidEnded (0.44s)
./cmd/mesh-controller --- PASS: TestAWalkRestartedByARebuildIsSaidStarted (0.39s)
./cmd/mesh-controller --- PASS: TestAWalkStoppedByHandAndRetriedEndsFailedWhenItFailsAgain (0.43s)
./cmd/mesh-controller --- PASS: TestReplay237 (0.43s)
./cmd/mesh-controller --- PASS: TestAnUnassignThatLeavesItsNodeUnresolvedSaysSoOnceInItsAnswer (0.40s)
./cmd/mesh-controller --- PASS: TestAnUnassignThatLeavesItsNodeResolvedSaysNothingOfIt (0.40s)
./cmd/mesh-controller --- PASS: TestAnAssignThatLeavesItsNodeUnresolvedSaysOnlyThePlansRefusal (0.43s)
./cmd/mesh-controller --- PASS: TestADirectoryOwnersKeyIsRefusedThroughAVerb (0.44s)
./internal/inventory --- PASS: TestTheLastSendSaysWhatTheMachineAnswered (0.23s)
./internal/inventory --- PASS: TestAReportIsTheLastSendsAnswerOnlyWhenItIsAboutThatSend (0.24s)
./internal/inventory --- PASS: TestAnIssuedValueIsSealedToEveryHolderOnceAndNotAgain (0.24s)
./internal/inventory --- PASS: TestAnIssuedValueIsSealedAgainAfterAKeyChangeOrAnAcceptance (0.24s)
./internal/inventory --- PASS: TestAConditionsLooksAreKeptBesideTheBucket (0.22s)
./internal/inventory --- PASS: TestAWalksStartIsKeptAndARefusedSaveLeavesItUnsaid (0.23s)
```
Re-review of review 80. Both blocking findings are fixed, each with a test. The new rule refusing the root or a top-level directory is also in place.
The fixes
1. The step-back covers rendered files only. The resource loop is gone from stepsBack. An uplink declaring /etc/resolv.conf as a resource beside a node-resolver holder is refused at Resolve again ("both declare the path"), as on the base. The test for it is TestAnUplinkDeclaringTheResolverFileAsAResourceIsRefusedBesideTheResolver. I put a3bf5da's node_resolver.go and seat_paths.go back under the new tests, and this test fails there, as it should.
2. Declared paths are cleaned.declaredPaths cleans both resource paths and rendered-file paths, and missingPaths, unclaimedSeatPaths, seatPathsOnMachine and stepsBack all go through it. TestAnUncleanSpellingOfASeatsPathIsTheSeatsPath covers all four spellings, both for refusing a stranger and for accepting a holder. It fails on a3bf5da.
3. A seat names no root or top-level directory.seatPathProblem refuses a path whose directory is /, and TestASeatNamesNoRootOrTopLevelDirectory tests it. It fails on a3bf5da. None of the starting set's paths is affected.
Mutations
I re-ran the 13 earlier mutations and added 4 new ones. Each was caught by at least one test, except one:
Turning off the step-back of rendered files now also fails the new resource test.
Not cleaning resource paths fails 2 tests.
Turning off the top-level rule fails 1 test.
Not cleaning the dropped rendered-file path fails the resource test (the rendered file at /etc//resolv.conf).
Not caught: not cleaning a rendered-file path in declaredPaths. No test has a stranger declare a rendered file at /etc//hosts. The code is right; the test is missing. I am not blocking on it. A one-line case in TestAnUncleanSpellingOfASeatsPathIsTheSeatsPath would close it.
Today's mesh
I re-ran this with mesh-catalog main (cd2b6aa) and mesh-media-catalog main (285b237).
module check gives byte-identical output with the base and this head, with exit 0 and no ADR 0287 problem.
I composed checkResources with the manifests of each node's assignment (ace, g14, novox and shanks, as mesh-controller.node gave them at 16:40). It gives 0 problems on every node, and the output is identical to the base, step-back included.
These refusals depend only on what a module declares, not on where it is assigned, so I conclude no live node is refused.
Checks
merge-check.sh with a throwaway postgres:17 (MESH_TEST_POSTGRES) exits 0. That covers gofmt, go vet, go test -race ./... and all 26 listed store tests, which ran against the store and passed. The timing failure I saw in review 80 did not recur.
Statuses on the head: mesh/merge-gate, mesh/repo-check and mesh/delivery are all success, and the forge says it is mergeable.
Verdict: approve at 24347ea5a02b53a9b2a7f44bcf4ca503ff23a1ca
Re-review of review 80. Both blocking findings are fixed, each with a test. The new rule refusing the root or a top-level directory is also in place.
## The fixes
**1. The step-back covers rendered files only.** The resource loop is gone from `stepsBack`. An uplink declaring `/etc/resolv.conf` as a resource beside a node-resolver holder is refused at `Resolve` again ("both declare the path"), as on the base. The test for it is `TestAnUplinkDeclaringTheResolverFileAsAResourceIsRefusedBesideTheResolver`. I put a3bf5da's node_resolver.go and seat_paths.go back under the new tests, and this test fails there, as it should.
**2. Declared paths are cleaned.** `declaredPaths` cleans both resource paths and rendered-file paths, and `missingPaths`, `unclaimedSeatPaths`, `seatPathsOnMachine` and `stepsBack` all go through it. `TestAnUncleanSpellingOfASeatsPathIsTheSeatsPath` covers all four spellings, both for refusing a stranger and for accepting a holder. It fails on a3bf5da.
**3. A seat names no root or top-level directory.** `seatPathProblem` refuses a path whose directory is `/`, and `TestASeatNamesNoRootOrTopLevelDirectory` tests it. It fails on a3bf5da. None of the starting set's paths is affected.
## Mutations
I re-ran the 13 earlier mutations and added 4 new ones. Each was caught by at least one test, except one:
- Turning off the step-back of rendered files now also fails the new resource test.
- Not cleaning resource paths fails 2 tests.
- Turning off the top-level rule fails 1 test.
- Not cleaning the dropped rendered-file path fails the resource test (the rendered file at `/etc//resolv.conf`).
- **Not caught:** not cleaning a *rendered-file* path in `declaredPaths`. No test has a stranger declare a rendered file at `/etc//hosts`. The code is right; the test is missing. I am not blocking on it. A one-line case in `TestAnUncleanSpellingOfASeatsPathIsTheSeatsPath` would close it.
## Today's mesh
I re-ran this with mesh-catalog main (cd2b6aa) and mesh-media-catalog main (285b237).
- `module check` gives byte-identical output with the base and this head, with exit 0 and no ADR 0287 problem.
- I composed `checkResources` with the manifests of each node's assignment (ace, g14, novox and shanks, as `mesh-controller.node` gave them at 16:40). It gives 0 problems on every node, and the output is identical to the base, step-back included.
- These refusals depend only on what a module declares, not on where it is assigned, so I conclude no live node is refused.
## Checks
- `merge-check.sh` with a throwaway postgres:17 (`MESH_TEST_POSTGRES`) exits 0. That covers gofmt, `go vet`, `go test -race ./...` and all 26 listed store tests, which ran against the store and passed. The timing failure I saw in review 80 did not recur.
- Statuses on the head: mesh/merge-gate, mesh/repo-check and mesh/delivery are all success, and the forge says it is mergeable.
Merge check at 24347ea5
Judged: head 24347ea5a02b merged into trunk 84deb530ee5a; read beside it: mesh-catalog 0863ec2a059d, mesh-controller-main 84deb530ee5a, mesh-host 303d4037e84e, mesh-tools 954e1a53eef9
mesh/merge-gate: PASS — every machine composes with the change as it did without (4 of 4 compose) (modules: build-agent, mesh-controller)
mesh/repo-check: PASS — its merge-check.sh passed
Delivery plan — builds build-agent, mesh-controller → ace, g14, novox, shanks; no bus step
tier 0: mesh-controller
tier 1: build-agent
ace: receives build-agent
g14: receives build-agent
novox: receives mesh-controller, build-agent
shanks: receives build-agent
build-agent keeps data (ADR 0233): what it holds is backed up before it moves
grants step: if this changes what the bus's user list says (a new tool, subject or user), novox is sent that list alone, every build there kept, before build-agent is judged on its first machine
read by no module's build: cmd/mesh-controller/seats_test.go, internal/catalogue/seat_paths_test.go, internal/catalogue/uplink_verbs_test.go
Run by the build seat on ace as build-1791832036763757323 (builds --log build-1791832036763757323).
--- the gate: build-agent, mesh-controller
mesh-controller: ok, 58 tool(s); WARNING: merge-window ask(s) for a setting and do(es) not say whether it is trusted, so it counts as trusted: set at the terminal alone; say "trusted" false where nothing trusts it (novox/hq issue 339); WARNING: controller stay(s) up and say(s) not how it is ready — judged by liveness alone, refused from 2026-11-18 (ADR 0240 rule 8)
long-running resources without health: 1
files asking for a setting without saying whether it is trusted: 1
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 1 module(s) changed against 4 machine(s), as the snapshot of 2026-10-12T19:03:57Z 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 2026-10-12T19:03:57Z
modules the change moves:
- mesh-controller (changed)
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 2 module(s) in 2 tier(s): build-agent, mesh-controller
notes:
- cmd/mesh-controller/seats_test.go, internal/catalogue/seat_paths_test.go and internal/catalogue/uplink_verbs_test.go read by no module's build, and rebuild nothing (issue 280)
--- its merge-check.sh (go toolchain)
ok github.com/novox/mesh-controller/cmd/mesh-builder 1.039s
ok github.com/novox/mesh-controller/cmd/mesh-controller 268.905s
ok github.com/novox/mesh-controller/examples/objectstore-provisioner 1.015s
ok github.com/novox/mesh-controller/examples/postgres-provisioner 1.034s
ok github.com/novox/mesh-controller/examples/redis-provisioner 1.015s
ok github.com/novox/mesh-controller/examples/route-proxy 1.138s
ok github.com/novox/mesh-controller/internal/artifacts 1.057s
? github.com/novox/mesh-controller/internal/beside [no test files]
ok github.com/novox/mesh-controller/internal/broker 3.945s
ok github.com/novox/mesh-controller/internal/builder 38.913s
ok github.com/novox/mesh-controller/internal/catalogue 7.547s
ok github.com/novox/mesh-controller/internal/conditions 2.569s
ok github.com/novox/mesh-controller/internal/envfile 1.104s
ok github.com/novox/mesh-controller/internal/facts 1.097s
ok github.com/novox/mesh-controller/internal/identity 4.295s
ok github.com/novox/mesh-controller/internal/inventory 93.984s
ok github.com/novox/mesh-controller/internal/lease 17.757s
ok github.com/novox/mesh-controller/internal/licences 3.921s
ok github.com/novox/mesh-controller/internal/licences/adapters 1.096s
ok github.com/novox/mesh-controller/internal/link 60.574s
ok github.com/novox/mesh-controller/internal/lint 2.295s
ok github.com/novox/mesh-controller/internal/outward 1.121s
ok github.com/novox/mesh-controller/internal/overlay 1.049s
ok github.com/novox/mesh-controller/internal/secrets 1.419s
ok github.com/novox/mesh-controller/internal/store 3.349s
ok github.com/novox/mesh-controller/internal/testbus 1.219s
ok github.com/novox/mesh-controller/internal/token 1.017s
store tests, each run against the store:
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeWalksTheOthersAgain (0.41s)
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeEndsTheMergesThatContainIt (0.41s)
./cmd/mesh-controller --- PASS: TestAStopWalksTheOthersAgainInTheOpenBatch (0.43s)
./cmd/mesh-controller --- PASS: TestAStopNamesWhatStaysOnTheMachinesAndWhatStillCarriesIt (0.40s)
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeLeavesItOutOfABatch (0.38s)
./cmd/mesh-controller --- PASS: TestSavingPlansAndMovingMergesIsOneAct (0.36s)
./cmd/mesh-controller --- PASS: TestAPlanFollowedFromTheStoreAnswersWhenItMoves (4.39s)
./cmd/mesh-controller --- PASS: TestNodeShowAndStatusReadTheLastSendFromTheStore (0.44s)
./cmd/mesh-controller --- PASS: TestTheConditionsVerbReadsTheLooksTheBucketNoLongerHolds (0.31s)
./cmd/mesh-controller --- PASS: TestAWalkLetGoAndStoppedByVerbsIsSaidStartedAndEnded (0.46s)
./cmd/mesh-controller --- PASS: TestAWalkStoppedOrClosedByAPersonIsSaidEndedStopped (0.46s)
./cmd/mesh-controller --- PASS: TestAWalkRetriedByAPersonIsSaidStartedAgain (0.42s)
./cmd/mesh-controller --- PASS: TestABacklogReleasedByAPersonIsSaidStarted (0.44s)
./cmd/mesh-controller --- PASS: TestAWalkEndedByCancelOrClearIsSaidEnded (0.50s)
./cmd/mesh-controller --- PASS: TestAWalkRestartedByARebuildIsSaidStarted (0.41s)
./cmd/mesh-controller --- PASS: TestAWalkStoppedByHandAndRetriedEndsFailedWhenItFailsAgain (0.46s)
./cmd/mesh-controller --- PASS: TestReplay237 (0.44s)
./cmd/mesh-controller --- PASS: TestAnUnassignThatLeavesItsNodeUnresolvedSaysSoOnceInItsAnswer (0.41s)
./cmd/mesh-controller --- PASS: TestAnUnassignThatLeavesItsNodeResolvedSaysNothingOfIt (0.42s)
./cmd/mesh-controller --- PASS: TestAnAssignThatLeavesItsNodeUnresolvedSaysOnlyThePlansRefusal (0.42s)
./cmd/mesh-controller --- PASS: TestADirectoryOwnersKeyIsRefusedThroughAVerb (0.45s)
./cmd/mesh-controller --- PASS: TestTheSendMakesNoValueForAnOutsideSecretNobodyGave (0.52s)
./cmd/mesh-controller --- PASS: TestAQuestionComposesWithoutAnOutsideSecretNobodyGave (0.51s)
./cmd/mesh-controller --- PASS: TestReplay201 (0.35s)
./cmd/mesh-controller --- PASS: TestARollbackWhoseSendWasRefusedIsCompletedByTheNextSend (0.35s)
./cmd/mesh-controller --- PASS: TestAGateJudgesTheOtherMovesBesideAKeptCore (0.35s)
./cmd/mesh-controller --- PASS: TestReplay401 (0.51s)
./internal/inventory --- PASS: TestTheLastSendSaysWhatTheMachineAnswered (0.27s)
./internal/inventory --- PASS: TestAReportIsTheLastSendsAnswerOnlyWhenItIsAboutThatSend (0.26s)
./internal/inventory --- PASS: TestAnIssuedValueIsSealedToEveryHolderOnceAndNotAgain (0.28s)
./internal/inventory --- PASS: TestAnIssuedValueIsSealedAgainAfterAKeyChangeOrAnAcceptance (0.27s)
./internal/inventory --- PASS: TestAConditionsLooksAreKeptBesideTheBucket (0.25s)
./internal/inventory --- PASS: TestAPlanIdHoldingAQuoteIsReadAsAPlainId (0.25s)
./internal/inventory --- PASS: TestAWalksStartIsKeptAndARefusedSaveLeavesItUnsaid (0.26s)
**Merge check** at `24347ea5`
Judged: head `24347ea5a02b` merged into trunk `84deb530ee5a`; read beside it: mesh-catalog `0863ec2a059d`, mesh-controller-main `84deb530ee5a`, mesh-host `303d4037e84e`, mesh-tools `954e1a53eef9`
- `mesh/merge-gate`: **PASS** — every machine composes with the change as it did without (4 of 4 compose) (modules: build-agent, mesh-controller)
- `mesh/repo-check`: **PASS** — its merge-check.sh passed
**Delivery plan** — builds build-agent, mesh-controller → ace, g14, novox, shanks; no bus step
- tier 0: mesh-controller
- tier 1: build-agent
- ace: receives build-agent
- g14: receives build-agent
- novox: receives mesh-controller, build-agent
- shanks: receives build-agent
- build-agent keeps data (ADR 0233): what it holds is backed up before it moves
- grants step: if this changes what the bus's user list says (a new tool, subject or user), novox is sent that list alone, every build there kept, before build-agent is judged on its first machine
- read by no module's build: cmd/mesh-controller/seats_test.go, internal/catalogue/seat_paths_test.go, internal/catalogue/uplink_verbs_test.go
Run by the build seat on ace as `build-1791832036763757323` (`builds --log build-1791832036763757323`).
```
--- the gate: build-agent, mesh-controller
mesh-controller: ok, 58 tool(s); WARNING: merge-window ask(s) for a setting and do(es) not say whether it is trusted, so it counts as trusted: set at the terminal alone; say "trusted" false where nothing trusts it (novox/hq issue 339); WARNING: controller stay(s) up and say(s) not how it is ready — judged by liveness alone, refused from 2026-11-18 (ADR 0240 rule 8)
long-running resources without health: 1
files asking for a setting without saying whether it is trusted: 1
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 1 module(s) changed against 4 machine(s), as the snapshot of 2026-10-12T19:03:57Z 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 2026-10-12T19:03:57Z
modules the change moves:
- mesh-controller (changed)
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 2 module(s) in 2 tier(s): build-agent, mesh-controller
notes:
- cmd/mesh-controller/seats_test.go, internal/catalogue/seat_paths_test.go and internal/catalogue/uplink_verbs_test.go read by no module's build, and rebuild nothing (issue 280)
--- its merge-check.sh (go toolchain)
ok github.com/novox/mesh-controller/cmd/mesh-builder 1.039s
ok github.com/novox/mesh-controller/cmd/mesh-controller 268.905s
ok github.com/novox/mesh-controller/examples/objectstore-provisioner 1.015s
ok github.com/novox/mesh-controller/examples/postgres-provisioner 1.034s
ok github.com/novox/mesh-controller/examples/redis-provisioner 1.015s
ok github.com/novox/mesh-controller/examples/route-proxy 1.138s
ok github.com/novox/mesh-controller/internal/artifacts 1.057s
? github.com/novox/mesh-controller/internal/beside [no test files]
ok github.com/novox/mesh-controller/internal/broker 3.945s
ok github.com/novox/mesh-controller/internal/builder 38.913s
ok github.com/novox/mesh-controller/internal/catalogue 7.547s
ok github.com/novox/mesh-controller/internal/conditions 2.569s
ok github.com/novox/mesh-controller/internal/envfile 1.104s
ok github.com/novox/mesh-controller/internal/facts 1.097s
ok github.com/novox/mesh-controller/internal/identity 4.295s
ok github.com/novox/mesh-controller/internal/inventory 93.984s
ok github.com/novox/mesh-controller/internal/lease 17.757s
ok github.com/novox/mesh-controller/internal/licences 3.921s
ok github.com/novox/mesh-controller/internal/licences/adapters 1.096s
ok github.com/novox/mesh-controller/internal/link 60.574s
ok github.com/novox/mesh-controller/internal/lint 2.295s
ok github.com/novox/mesh-controller/internal/outward 1.121s
ok github.com/novox/mesh-controller/internal/overlay 1.049s
ok github.com/novox/mesh-controller/internal/secrets 1.419s
ok github.com/novox/mesh-controller/internal/store 3.349s
ok github.com/novox/mesh-controller/internal/testbus 1.219s
ok github.com/novox/mesh-controller/internal/token 1.017s
store tests, each run against the store:
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeWalksTheOthersAgain (0.41s)
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeEndsTheMergesThatContainIt (0.41s)
./cmd/mesh-controller --- PASS: TestAStopWalksTheOthersAgainInTheOpenBatch (0.43s)
./cmd/mesh-controller --- PASS: TestAStopNamesWhatStaysOnTheMachinesAndWhatStillCarriesIt (0.40s)
./cmd/mesh-controller --- PASS: TestAStopWithoutAMergeLeavesItOutOfABatch (0.38s)
./cmd/mesh-controller --- PASS: TestSavingPlansAndMovingMergesIsOneAct (0.36s)
./cmd/mesh-controller --- PASS: TestAPlanFollowedFromTheStoreAnswersWhenItMoves (4.39s)
./cmd/mesh-controller --- PASS: TestNodeShowAndStatusReadTheLastSendFromTheStore (0.44s)
./cmd/mesh-controller --- PASS: TestTheConditionsVerbReadsTheLooksTheBucketNoLongerHolds (0.31s)
./cmd/mesh-controller --- PASS: TestAWalkLetGoAndStoppedByVerbsIsSaidStartedAndEnded (0.46s)
./cmd/mesh-controller --- PASS: TestAWalkStoppedOrClosedByAPersonIsSaidEndedStopped (0.46s)
./cmd/mesh-controller --- PASS: TestAWalkRetriedByAPersonIsSaidStartedAgain (0.42s)
./cmd/mesh-controller --- PASS: TestABacklogReleasedByAPersonIsSaidStarted (0.44s)
./cmd/mesh-controller --- PASS: TestAWalkEndedByCancelOrClearIsSaidEnded (0.50s)
./cmd/mesh-controller --- PASS: TestAWalkRestartedByARebuildIsSaidStarted (0.41s)
./cmd/mesh-controller --- PASS: TestAWalkStoppedByHandAndRetriedEndsFailedWhenItFailsAgain (0.46s)
./cmd/mesh-controller --- PASS: TestReplay237 (0.44s)
./cmd/mesh-controller --- PASS: TestAnUnassignThatLeavesItsNodeUnresolvedSaysSoOnceInItsAnswer (0.41s)
./cmd/mesh-controller --- PASS: TestAnUnassignThatLeavesItsNodeResolvedSaysNothingOfIt (0.42s)
./cmd/mesh-controller --- PASS: TestAnAssignThatLeavesItsNodeUnresolvedSaysOnlyThePlansRefusal (0.42s)
./cmd/mesh-controller --- PASS: TestADirectoryOwnersKeyIsRefusedThroughAVerb (0.45s)
./cmd/mesh-controller --- PASS: TestTheSendMakesNoValueForAnOutsideSecretNobodyGave (0.52s)
./cmd/mesh-controller --- PASS: TestAQuestionComposesWithoutAnOutsideSecretNobodyGave (0.51s)
./cmd/mesh-controller --- PASS: TestReplay201 (0.35s)
./cmd/mesh-controller --- PASS: TestARollbackWhoseSendWasRefusedIsCompletedByTheNextSend (0.35s)
./cmd/mesh-controller --- PASS: TestAGateJudgesTheOtherMovesBesideAKeptCore (0.35s)
./cmd/mesh-controller --- PASS: TestReplay401 (0.51s)
./internal/inventory --- PASS: TestTheLastSendSaysWhatTheMachineAnswered (0.27s)
./internal/inventory --- PASS: TestAReportIsTheLastSendsAnswerOnlyWhenItIsAboutThatSend (0.26s)
./internal/inventory --- PASS: TestAnIssuedValueIsSealedToEveryHolderOnceAndNotAgain (0.28s)
./internal/inventory --- PASS: TestAnIssuedValueIsSealedAgainAfterAKeyChangeOrAnAcceptance (0.27s)
./internal/inventory --- PASS: TestAConditionsLooksAreKeptBesideTheBucket (0.25s)
./internal/inventory --- PASS: TestAPlanIdHoldingAQuoteIsReadAsAPlainId (0.25s)
./internal/inventory --- PASS: TestAWalksStartIsKeptAndARefusedSaveLeavesItUnsaid (0.26s)
```
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.
What
Builds hq ADR 0287 in
internal/catalogue:Seat.Paths(compiled, never stored, likeReceives):node-hostnamenames/etc/hostsand/etc/hostname;node-uplinknames/etc/resolv.confstepping back fornode-resolver;node-resolvernames/etc/resolv.conf; every other seat names none.CanHoldrefuses a holder that does not declare every path its seat names (as a resource or as a file the mesh renders), naming the seat and the missing paths. That covers registration (CatalogueProblems) and handover (seat --to/--add)./etc/resolv.conf. Access (reading) stays allowed.paths(SeatDeclaration.Paths), under the same rules: node seats only, literal, clean and absolute. A path another seat already names is refused, and the second declaration loses. A holder of such a seat must declare its paths.checkResourcesapplies the same rule to what the controller generates for a machine, so a computed module cannot write a seat's path.stepsBackinnode_resolver.go), replacing the uplink/resolver special case. It now drops a resource on the path as well as a rendered file.seatsshows a PATHS column, and below the table who owns each seat's path on each machine, with the step-back applied (path-ownersin--json).Why
Which module owns
/etc/hostsor/etc/resolv.confcould be found only by reading every manifest. A module not holdingnode-hostnamecould declare/etc/hostson a machine where nobody held it, and nothing refused it.How it was tested
The new
internal/catalogue/seat_paths_test.gocovers the rows of ADR 0287 "How it is checked":CanHoldwith the path declared as a resource, declared as a rendered file, under the operator's home, and not declared at all;/etc/hostsfrom a module that does not claimnode-hostname, and accepting the uplink and the resolver together;${, or when it names/etc/hosts;/etc/hosts;PathOwners;CatalogueProblemsover the catalogue, with no ADR 0287 problem: run against the captured copy and against mesh-catalog main viaMESH_CHECK_BESIDE.The existing uplink/resolver composition tests pass with the special case removed. The
seatslisting is tested incmd/mesh-controller/seats_test.go. WithmissingPathsandunclaimedSeatPathsturned off, the four refusal tests fail.sh merge-check.shpasses locally; the store tests were skipped because no postgres was available.Fixes issue 412
Delivery
novox/mesh-controller@24347ea5a02b— delivered since 2026-10-12T19:31:22ZDelivery plan — builds build-agent, mesh-controller, route-proxy → ace, g14, novox, shanks; no bus step
Machines
Transitions
The commit's note under
refs/notes/mesh-plankeeps every transition:git log --notes=mesh-plan.Merge check at
a3bf5da9mesh/merge-gate: PASS — every machine composes with the change as it did without (4 of 4 compose) (modules: build-agent, mesh-controller)mesh/repo-check: PASS — its merge-check.sh passedDelivery plan — builds build-agent, mesh-controller → ace, g14, novox, shanks; no bus step
Run by the build seat on novox as
build-1791814810148047906(builds --log build-1791814810148047906).Verdict: request changes at
a3bf5da9b1Independent review for issue 412 / hq ADR 0287. Two small defects need fixing before merge. Neither one touches today's mesh, but both open gaps in the very rules this PR adds.
Blocking
1. The step-back now drops a resource in the check but not in the declaration.
stepsBack(node_resolver.go) now also leaves out a resource on the seat's path. OnlycheckResourcessees that result throughsteppedBack. The declaration itself (declaration.go:1124) callsstepsBackonly forFactsFrom; the module's own resources are copied from the unsteppedm.Resources.I checked this with a scratch test: systemd-networkd with
/etc/resolv.confdeclared as a resource, plus a node-resolver holder rendering it. On this head,Resolveaccepts it andDeclarationcomposes the file twice (systemd-networkd.resolversandlocal-resolver.fact-resolvers). On the base (9eddf36) the sameResolveis refused.So the check now passes a machine that composes two writers of the resolver file, where it used to refuse it. This is not live today: every uplink holder (networkmanager, systemd-networkd) renders the file as a fact.
Fix: apply the same step-back to the resource loop in
Declaration, or keep the step-back to rendered files only. Add a declaration-level test with the uplink declaring the path as a resource;TestTheUplinkStepsBackByTheSeatsDefinitioncovers onlystepsBackandcheckResources.2. A path that is not clean gets past "only a claimant declares a seat's path".
declaredPathscompares a resource's path byte for byte. A module that claims nothing and declares any of these as a file resource passes bothCatalogueProblemsandcheckResources(I checked each one throughParseManifest):/etc//hosts/etc/./hosts/etc/../etc/hosts/etc/hosts/The machine then writes
/etc/hosts. A seat's own path is required to be clean, but a declared path is not.Fix: compare
path.Clean(p)indeclaredPaths, whichmissingPathsandunclaimedSeatPathsboth use. Add one of these paths toTestOnlyAClaimantDeclaresASeatsPath.No manifest in mesh-catalog or mesh-media-catalog main uses such a path today.
Checked and fine
Check table (item 1). Every row of ADR 0287's table has a test. I ran 13 mutations, one per rule; each was caught by at least one test:
seatPathProblem,SeatPathProblems,missingPaths,unclaimedSeatPathsandseatPathsOnMachinemade to return nothing;declaredSeatPathProblemsturned off;UseSeatsnot copyingPaths, and thePathOwnersstep-back turned off.One small gap: the handover row is proven through
CanHolddirectly, not throughseat --to. The command calls the sameCanHold, so I accept it.Today's catalogues (item 2).
module checkover mesh-catalog main and over mesh-media-catalog main gives byte-identical output with the base, main and PR binaries, and no ADR 0287 problem.I also composed
checkResourceswith the manifests of each node's current assignment (ace, g14, novox, shanks, frommesh-controller.node). Base, main and the PR each give 0 problems on every node.Nine assigned modules have no manifest in either catalogue (mesh-host, node-tools, the mesh-wireguard overlay generator, and six app repositories). node-tools' manifest names none of the three paths, and the overlay generator writes only
/etc/wireguard/mesh0.conf.A composition refusal from this PR depends only on a module's own declared paths against the seats' paths, not on where it is assigned. So no live node can be refused by it.
Same step-back result (item 3). g14 and shanks run networkmanager and systemd-resolved. On base and PR, the uplink loses only its
resolversfact and keeps its three resources, and systemd-resolved keeps the file. ace and novox are unchanged.Dropping the resolver file (item 4). The path is dropped only when the node-resolver holder on that machine declares it, so the file always keeps one writer in the composition. No uplink declares it as a resource today. Finding 1 is the remaining risk: the opposite of losing the file, it is composed twice.
Module-declared seats (item 5). Paths must be literal, clean and absolute, on a node seat only. A path already named loses to the first declaration, and a module's seat never steps back. Only literal comparison is applied; see finding 2 for the declared side.
Minor: a module seat may name
/or/etc. That is harmless with literal matching, but the "role's own path" clause is not machine-checked.seats --json(item 6). The change only adds: per-rowpaths(omitempty) and a top-levelpath-owners(omitempty). Every existing key is unchanged. Theseatsseat verb passes the JSON through.Local checks (item 7). gofmt and
go vetare clean.go test -race ./...passes, apart fromTestATierPutsABrokenModuleBackAtOnce/gate_kept_on_node-toolsin cmd/mesh-controller. That is delivery-plan code this PR does not touch, and it failed once on timing while another race run was loading the machine. It passed 3 of 3 runs alone on this head.All 26 store tests in testdata/store-tests pass against a throwaway postgres:17 (
MESH_TEST_POSTGRES).The PR merged locally onto current main (
3cfc121, 7 commits ahead of its base) builds, vets, and passes the catalogue and cmd suites.Statuses on the head (item 8). mesh/merge-gate success ("every machine composes with the change as it did without"), mesh/repo-check success, mesh/delivery success ("ready"). The forge says it is mergeable.
Once 1 and 2 are fixed, each with a test, I expect to approve.
Merge check at
24347ea5mesh/merge-gate: PASS — every machine composes with the change as it did without (4 of 4 compose) (modules: build-agent, mesh-controller, route-proxy)mesh/repo-check: PASS — its merge-check.sh passedDelivery plan — builds build-agent, mesh-controller, route-proxy → ace, g14, novox, shanks; no bus step
Run by the build seat on g14 as
build-1791817762661447750(builds --log build-1791817762661447750).Verdict: approve at
24347ea5a0Re-review of review 80. Both blocking findings are fixed, each with a test. The new rule refusing the root or a top-level directory is also in place.
The fixes
1. The step-back covers rendered files only. The resource loop is gone from
stepsBack. An uplink declaring/etc/resolv.confas a resource beside a node-resolver holder is refused atResolveagain ("both declare the path"), as on the base. The test for it isTestAnUplinkDeclaringTheResolverFileAsAResourceIsRefusedBesideTheResolver. I put a3bf5da's node_resolver.go and seat_paths.go back under the new tests, and this test fails there, as it should.2. Declared paths are cleaned.
declaredPathscleans both resource paths and rendered-file paths, andmissingPaths,unclaimedSeatPaths,seatPathsOnMachineandstepsBackall go through it.TestAnUncleanSpellingOfASeatsPathIsTheSeatsPathcovers all four spellings, both for refusing a stranger and for accepting a holder. It fails ona3bf5da.3. A seat names no root or top-level directory.
seatPathProblemrefuses a path whose directory is/, andTestASeatNamesNoRootOrTopLevelDirectorytests it. It fails ona3bf5da. None of the starting set's paths is affected.Mutations
I re-ran the 13 earlier mutations and added 4 new ones. Each was caught by at least one test, except one:
/etc//resolv.conf).declaredPaths. No test has a stranger declare a rendered file at/etc//hosts. The code is right; the test is missing. I am not blocking on it. A one-line case inTestAnUncleanSpellingOfASeatsPathIsTheSeatsPathwould close it.Today's mesh
I re-ran this with mesh-catalog main (cd2b6aa) and mesh-media-catalog main (285b237).
module checkgives byte-identical output with the base and this head, with exit 0 and no ADR 0287 problem.checkResourceswith the manifests of each node's assignment (ace, g14, novox and shanks, asmesh-controller.nodegave them at 16:40). It gives 0 problems on every node, and the output is identical to the base, step-back included.Checks
merge-check.shwith a throwaway postgres:17 (MESH_TEST_POSTGRES) exits 0. That covers gofmt,go vet,go test -race ./...and all 26 listed store tests, which ran against the store and passed. The timing failure I saw in review 80 did not recur.Merge check at
24347ea5Judged: head
24347ea5a02bmerged into trunk84deb530ee5a; read beside it: mesh-catalog0863ec2a059d, mesh-controller-main84deb530ee5a, mesh-host303d4037e84e, mesh-tools954e1a53eef9mesh/merge-gate: PASS — every machine composes with the change as it did without (4 of 4 compose) (modules: build-agent, mesh-controller)mesh/repo-check: PASS — its merge-check.sh passedDelivery plan — builds build-agent, mesh-controller → ace, g14, novox, shanks; no bus step
Run by the build seat on ace as
build-1791832036763757323(builds --log build-1791832036763757323).