Add the replays of issues 279 and 412 (hq issues 279, 412) #285
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: novox/mesh-controller#285
Reference in New Issue
Block a user
What: two test files in internal/catalogue,
TestReplay279andTestReplay412, each written only with what the catalogue had before its fix, so mesh-lab's prover can lay them over the commit before.TestReplay279: a directory owned by${setting:puid}:${setting:pgid}is composed for a machine whose settings give 1001:2000, and must reach the node-engine owned by 1001:2000 (fix: #260).TestReplay412: a module declaring /etc/hosts without claiming node-hostname is refused naming the seat, and a node-hostname claim declaring neither /etc/hosts nor /etc/hostname is refused by CanHold naming both (fix: #267).Why: issues 279 and 412 are delivered and resolve with a replay in mesh-lab's register; the fixes' own tests read what the fixes added and do not build at the commit before.
Tested:
go test -run '^TestReplay(279|412)$' ./internal/cataloguepasses on main; mesh-labgo run ./cmd/prove R279 R412says PROVED for both: before #260 the machine does not compose (${setting:…} is not filled in this field), before #267 nothing refuses the stranger declaring /etc/hosts. Registered in novox/mesh-lab (same branch name).Tests only; no behaviour changes. Replays for issue 279 and issue 412.
Delivery
novox/mesh-controller@3bdab29a5a0e— delivered since 2026-10-13T00:45:39ZDelivery plan — builds nothing: the change touches no module of the mesh's graph
Group
fix/279-412-replays, in order: novox/mesh-controller@3bdab29a5a → novox/mesh-lab@671d56a08aTransitions
The commit's note under
refs/notes/mesh-plankeeps every transition:git log --notes=mesh-plan.Verdict: approve at
3bdab29a5aChecked:
go run ./cmd/prove --repos ... R279 R412from mesh-lab #90's branch): both PROVED, each failing before its fix and passing on it.${setting:puid}:${setting:pgid}reaches the node-engine owned by the numbers the machine's settings give (1001:2000). Before #260 it fails for the right reason: Compose refuses${setting:...}in an owner ("is not filled in this field"), which is the literal-owner limitation the issue names. It also asserts the module is not left out, so a pass cannot come from the module being dropped.Nit (not blocking): the 412 file comment lists Shelf among the pre-fix API, but the mesh-lab register comment does not. Harmless.
Merge check at
3bdab29a: head3bdab29a5amerged into trunk454a3db69a, passOut of date when it landed: out of date: judged against trunk
454a3db69a, the trunk is now 4473bc7d7315; judged again on a recheck or when it is mergedVerdict: approve at
3bdab29a5aThis is an independent review. It replaces the earlier approval from a reviewer the builder spawned. The code needs no change. The required checks are not green yet: they must be rechecked and read green before a merge.
Statuses on the head:
454a3db69a, the trunk is now 4473bc7d7315; judged again on a recheck or when it is merged").Ask mesh-delivery.recheck and merge only on the new verdict. This change has to merge before mesh-lab #90, because #90's HomeRef is this commit.
The replays use only code from before each fix. I ran the lab's prover from mesh-lab #90 (go run ./cmd/prove R279 R412) against an offline copy of mesh-controller. It laid these two files over the commit before each fix and over each fix:
5facd8b, the first parent of #260's merge9eddf36, at replay279_test.go:24 ("the machine did not compose: sonarr's resource recycle-bin holds ${setting:puid} in its owner, and ${setting:…} is not filled in this field"). It passes on9eddf36.84deb53, the first parent of #267's merge4dbd406, at replay412_test.go:18 ("a module declaring /etc/hosts without claiming node-hostname was not refused"). It passes on4dbd406.In both cases the test fails as a test, not as a build error, so every symbol it uses already existed before the fix. I also checked by name. anAdoptedAnchor, anchorRendering, byID, Layer and LeftOut existed at
5facd8b. CatalogueProblems, Shelf, SeatNamed, CanHold and Claim.Serves existed at84deb53.I ran go test -run '^TestReplay(279|412)$' ./internal/catalogue at this head and on the current trunk
4473bc7with the two files added. Both tests pass in both places.What each replay asserts matches its issue. R279 checks that a directory owned by ${setting:puid}:${setting:pgid} reaches the node-engine as 1001:2000. R412 checks two things: a module declaring /etc/hosts without the node-hostname claim is refused, and a node-hostname holder declaring neither path is refused, naming both.