Kind: short package names (builds, sshkeys, forgemsg, backupgood,
diskreadings, grantsstate, busgrants) in `packages`, which the indexer's
judge refuses by design 32's rule (a Go package is named by its go.mod's
module path). This held mesh-delivery-gate, the build modules, mounts and
restic. Each key is now the import path the module's code uses, with its
series kept; the manifest tests read the new keys, and the host/machine-name
word tests in mounts, nfs-server, mesh-build-node and mesh-delivery-gate
look past the path's `git.novox.be/novox/` prefix.
Quick fix: only the gate's manifest. Thorough fix (chosen): all 15, since
the same refusal holds each. Follow-up for the checker: the candidate
verdict lets such a name through (J7(3)), so the merge check did not stop it.
Claude-Session: https://claude.ai/code/session_01NcbgfFfQ4Z9UE5jK32PmE5
mesh-build-go keeps trunk's zero-version resolution on the group's mesh-delivery sdk types; mesh-build-node keeps per-path packages with the builder's version and seat.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
A folder module (the contracts folders) declares several package artifacts under ts/seats/<seat>/, each with its own package.json and lock and none at the module directory; the fetch asked for a lock at the module directory and failed the group's check. The fetch installs, the build packs and the test level runs at the artifact's path; an artifact naming no usable path keeps the module directory. The module's .gitignore anchors the binary's name so the command directory is not ignored.
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
catalogue's SeatReference, InRegistrySeat, Routed, Recorded, InArtifactStore and ArtifactSweepSeat leave mesh-sdk;
each package gives the two registry seats' names from their contracts (or the build step's copies of them).
Claude-Session: https://claude.ai/code/session_01NcbgfFSnn24M3GcqttoU7i
Conflicts, both sides kept: artifact-store-tools' invokes (the assigner's
and indexer's read tools, followed through views, beside the registry
seats' hold, release and remove), its main (the views reader and the
seat asker), its manifest row; gitea's invokes (the indexer's read tool,
the hand act and the builder's versions) and its consumes of the
indexer's views. gitea's account test takes the packages account's
names; repoCheckFacts, removed on main, is not imported. No catalogue
module reads the indexer's state.
Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig
The patterns mesh-build-node and mesh-build-image matched the command
directory, so publish.go and its test were never committed and the
branch did not compile. Anchored to the binaries now.
Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig
Bundles go through oci-artifact-registry and answer the seat's reference;
a trunk package is packed under the builder's version and published through
npm-package-registry, so the holder can check it with the builder. Reads are
anonymous, so the write pair, the sealed secret and the store's address go.
The one-part sender moves into buildstep for both build modules, so the move
to the registry clients is made in one place.
Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig
The unit level fails on one, a module with no tests of its own too; a build ask answers
them in its out-of-step through buildstep's compare (lead's ruling on A2k finding 1).
Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig
ADR 0607 §3, §5: the fetch's package argument also takes {name, version}; the version is placed from the registry in
each TypeScript member whatever its lock names, checked against the lock's integrity when the lock names it, else the
registry's; nothing is packed then, and a version the registry lacks fails naming it. The packed {repository, commit}
form stays for a checker that hands it. Rows in the placement and argument tests.
Claude-Session: https://claude.ai/code/session_01JG73KJNrH5EgFXMKKYSrig
mesh/delivery-group group feat/sdk1 rejected: composed together: the checker's verdict at the members' heads (5c08034be629): integration: meshbuildgo/cmd/mesh-…
mesh/check failed: integration: meshbuildgo/cmd/mesh-build-go… | 640455999e66 on ee75d4327ce5 | judged by the candidate's mesh-build-go at 785ebb388d5a
ensureModfile copied go.mod word for word into pins/<dir>/, where `go mod verify` runs, so a replace to ./x or ../x pointed
nowhere (route-proxy's failure of 2026-11-02 13:53, met a second time with the release's sibling packages). Every relative
replace target is now made absolute under the tree the go.mod came from. mesh-build-node installs a module's own npm
packages in a scratch copy of package.json, so a file:, link: or portal: dependency on a relative path would fail the same
way; it is refused naming the package and its path, as a module without a lock is.
Quick fix: rewrite the one replace route-proxy had, in ensureModfile (minutes). Thorough fix: the kind, a go.mod or
package.json copied away from its tree, for every replace form (single, block, with a version, ./ and ../) in the one place
all copies go through (ensureModfile, which versions.go and the pins share), and the same kind in the other build module
that copies a manifest. Chosen: thorough, since it was met twice and hides a failure until the delivery (quick-or-thorough,
fix-a-bug-met-twice). Test: TestARelativeReplaceStillResolvesInTheTemporaryModfile runs real go mod verify in the pins directory.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
The runner, the toolchain guard, the fetch, the parsers, buildcpu, the conformance kit and mesh-pack move from mesh-sdk's
buildstep into modules/mesh-build-go/buildstep (own go.mod). mesh-build-go, mesh-build-node and mesh-build-image require it
by a same-repository replace; what the delivery parts read (the failure contract, the fetch request, the stale sweep, the
module tree shape, NewID, CapacitySetting) stays imported from mesh-sdk as sdkstep.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
The levels a build module judges hand the tests MESH_BUILD_NODE=1; the tests that run a step in bubblewrap fail on it when bwrap is missing. Issue 1309
A replaced mesh-build-go left a check ask's step running and its start
sweep spared it (issue 1890, note 21). mesh-build-go, mesh-build-node and
mesh-build-image pin mesh-sdk 981a35b5, whose steps name their owner and
whose sweep ends another owner's at once; each runs the sweep through one
method that main calls and that its test runs through the conformance
kit's left-running case. The failurekit check now refuses a build module
whose tests do not run that case, test step or not.
47 packages carried the marker and were never kept: about 2000 tests per level ran on every check, whatever it
changed. The marker now lists, as paths or globs from the repository's root (? before a path whose absence is read),
what the package's tests read outside it; the digest hashes those files with the package's own non-Go files. A marker
that lists nothing, names an unclean path, or a line that matches nothing keeps today's never-kept digest.
The 46 markers list their reads; checks/deps and checks/trunk lose theirs (their tests read only temporary
directories). Digest form 6.
Quick fix of issue 1936, chosen by the operator 2026-10-26 21:56; the thorough fix (the Go test log's files kept with
each verdict) stays on the issue.
mesh/delivery-group group fix/capacity-follows-cpu-share rejected: composed together: the checker's verdict at the members' heads (63694fe9222d): unit: ..test
mesh/check checking 3f8b0b42480b on 7c5c29421f02 | novox/mesh-catalog#511 in group fix/capacity-follows-cpu-share
mesh-build-go, -node and -image declare build-hours (node=HH:MM-HH:MM in the node's local time,
empty = any hour) and hand it to the SDK puller (the operator, 2026-10-26: g14 builds only from
18:00 to 06:00); their SDK pin moves to mesh-sdk d00dac1, which has it.
Each of mesh-build-go, -node and -image declares cores-per-step (6, 2, 2) and hands it to the
SDK puller, which bounds the configured capacity by floor(slice quota cores / cores-per-step),
read live; thread limits and the status tools follow the effective capacity (issue 1806's
follow-up: three Go checks on shanks' 12-core slice starved past the 30-minute step bound).
mesh-build-go says its lint and test phases with the ask id (issue 1890).
mesh-build-node kept the first fifteen lines of node --test's output, so a passing line of the first file stood for the
failure. Both build modules now parse through the SDK (go test -json, node --test's TAP report written to a file, the
type-check and go vet as console failures), name themselves in the result, and run the SDK's captured cases; the
catalogue check failurekit refuses a node-build-* module with a test step whose tests do not run the kit.
mesh-catalog #495 gave read on every repository to the tools' account alone. The build modules
fetch over ssh as nox-builder (the forge-account provider's builders account, ADR 0466 §3), so
ace's and g14's fetch of jschoubben/car-hunter and jschoubben/snake still failed; the machines'
agents (nox-agent, ADR 0410 §1) could not read them either.
- gitea: readall.ts grants every account of one list, meshAccounts in tools-account.ts (mesh-tools,
nox-builder, nox-agent; a Go test holds the two Go names to internal/forgeaccount's constants),
on the same start pass, forge events and tools. gitea_unreadable_repos answers per account and
names the accounts the forge does not hold yet; gitea_grant_read takes accounts as well as
repositories. An absent account is asked nothing; the start pass runs again ten minutes apart
until every account was there once.
- mesh-build-go, mesh-build-node, mesh-build-image: pin mesh-sdk ff81bdf8 (fix/fetch-reads-all),
whose forgekey.Said keeps git's words past the bare "error:" line: the mirror's line said only
"git -C: error:". Their own git reasons (ls-files, ls-remote, rev-parse, repack) go through it too.
buildah writes STEP lines on stdout and its error on stderr, so the last line of both hid every image build's cause.
mesh-build-image's reason is now stderr's last lines (600 bytes) with the exit status, and the journal gets both tails;
mesh-build-node keeps stderr's end and stdout's start apart; mesh-build-go's cause skips go build's '# package' line.
Change A rebuilds the three build modules, so their runtimes restart once; it
changes nothing else about their builds until change B gives them the share.
Review round 1: a cap per module let three build modules take three shares
of the node. The share is now one node value held by the node-engine's
yielding slice; each step divides it by every build slot of the node, from the
composer's words, or assumes every build seat at its own capacity without them.
jschoubben
jochen
mesh-tools