mesh/check checking fd7d93333e6b on 21252194d530 | novox/mesh-sdk#93 in group fix/daemon-dir-kept
Declining an ask whose probe failed would turn a truly broken toolchain into a
wait of up to the decline bound that then fails with an empty "needs a node
with" reason. With the node-engine no longer deleting a running build and
commands started in a directory that exists, the getwd cause is gone, so the
guard's refusal stays as it was. The decline refactor goes with it.
A build module runs in the directory its build was unpacked in. When the
node-engine replaced that build, every command the module started inherited a
deleted working directory, and the toolchain probe's failure became a final
refusal of the check that asked it (three checks on ace on 2026-10-26).
forgekey: ExecRunnerClean starts its command in the HOME it is given, or the
temporary directory; the other runners keep the process's directory while it
exists and use the temporary directory once it is gone.
buildstep: a guard whose probe failed answers a Refusal with Again set; the
puller declines such an ask (delivery.StepDeclined) so its asker puts it again
for another node or a later pull, instead of ending it refused. A version
mismatch stays a refusal.
Review round 2: change A must not slow builds before any slice exists, so a
module given no share sets no GOMAXPROCS. The slice's quota bounds the total;
threads only size one step, from the node's share over the module's own
capacity (shanks: 24 x 50% / 3 = 4). The share is the node-engine's own config.
Review round 1: the decoder type ships before any composer can emit it, so a
routine SDK bump never sends yields to an engine that refuses them; the
composer half moves to a later change. Every yielding process of a node goes
into one top-level slice, meshyield.slice, so the three build modules share one
quota instead of taking one each; the share is one node value, node-tools'
build-cpu-share, and each step's threads divide it by every build slot of the
node (words from the composer, a conservative fallback without them).
mesh/check being carried to 3203358c4cb5 on 2aae6f5580f2 | from mesh-delivery c8b26735c8cf | novox/mesh-sdk#92 in group fix/734-lost-ask
A deaf asker must be loud: SayHearing keeps each Hearing in the asker's state hearing, and HearingHealth answers the
module's health tool from it, unhealthy with why while DEAD_LETTERS is not heard.
An asker started before its grant was composed was refused once and stayed deaf until it started again, silently.
DeadLetterHearing tries in the background and again on an event (Again: a reconnect, an ask waited for while deaf), and
tells each change of Hearing so the module's health tool can say it. DeadLetterReads drops the flow control a pull
consumer does not use and deletes one consumer by name.
The checker's try and the builder's lanes waited for ever on an ask given up during a bus reload: no build node writes a
step for it. LostAskWaiter hands the asker a failed step on the node bus when DEAD_LETTERS keeps the ask's copy
(HearDeadLetters, granted by DeadLetterReads), and under NotTakenBound when no node took the ask within the bound.
A check ask given up during a bus reload left its try running for ever: the asker had no grant to learn that
DEAD_LETTERS holds it. DeadLetterReads grants an ordered consumer on DEAD_LETTERS filtered to the seat stream's copies,
and nothing else of the stream.
The operator's desktop slows while it builds: the Go build runtime took 1650% of
shanks' 24 processors with nothing holding it back. A build seat's holder now
runs its runtime at idle priority under its node's cpu-cap share (declaration
Yields, composed from Seat.Yields and the module's cpu-cap setting), and
buildcpu/buildstep give each step its part of that cap in threads.
A check ask's packages may need something only some build nodes have (mesh-build-node's integration tests read the
Node build module's installed toolchain). An ask now names its Needs; a holder lacking one ends the ask's step refused
as declined and drops the message for its asker to put again, skipping that worker until its next pull wait; a level's
packages whose declared needs the node lacks are left untested and named unmet, for the asker to place.
The grants part writes its list as an object of its own bucket, sealed to the bus node's key, and names it in
user-list.sealed; the node-engine of the node holding mesh-broker is granted that one key and an object by name, the
grants part its bucket (the controller's copy carries the same grants, byte for byte, golden regenerated there).
A state may name its readers (announce, NamedReaders): mesh-vault.bus-accounts is read by mesh-grants alone.
The vault holds no value for the controller's own account, the node accounts and the enrolled devices, so it cannot
hash them; the one part that has their hashes is the node-engine of the bus node, in the list the bus loaded. The have
of the bus module now carries that list's hashes sealed to the vault's key (UserListWritten.Handed), which every reader
of the have bucket sees as a count, a list digest and a key fingerprint, and only the vault opens. HandedAccountsOf
reads them from a composed list, HandAccounts seals, OpenHanded opens and refuses a box sealed to another key.
mesh/check checking 79649c3fa9f4 on c3544b6a17a2 | novox/mesh-sdk#87 in group fix/1921-every-failure-r
A check ask stopped its level at the first test step that did not pass,
so the packages after the changed ones never ran and a check found one
fault per round. Every step now runs; the level still fails, and no
later level runs (ADR 0484 clause 13). The verdict gains Failures, every
failed test with its first line, at most MaxFailures and a count.
mesh/check checking 136fa8b2a896 on 095d41fa3e81 | novox/mesh-sdk#86 in group fix/1790-grants-create-buckets
BucketStreamConfig and ModuleStreamConfig are the stream configurations the controller's
CreateOrUpdateKeyValue (BucketConfig) and StreamConfigFor assert at its start, pure, for the
grants part's create of a bucket or stream a manifest declares and the bus lacks. A log already
has Log.Config. Never used for an update.
Issue: 1790
The copied bus composer grants mesh-grants' principal $JS.API.STREAM.CREATE.<stream> and
$JS.API.STREAM.INFO.<stream> for each bucket (KV_<module>_<state>), log (LOG_<module>_<log>) and
module stream a placed module's manifest declares, each by its own name: NATS wildcards are whole
tokens, so nothing narrower than an exact name exists and nothing wider is granted. The writers
table lets the grants part create, never update or delete. The operator allowed this one grant
on 2026-10-26 08:54, beside the memberships grant of 22:31 (ADR 0547 clause 1 and 5).
The expected lists were written again by the controller's own internal/broker with the same
change (mesh-control fix/1790-grants-create-buckets); a fixture now places mesh-grants beside the
builder's state, log and stream.
Issue: 1790
mesh/check failed: integration: github.com/novox/mesh-controller/cmd/mesh-controller.TestTheDeclarationsStreamGoesOnceEveryNodeRunsTheNewEngine: with …
The grants part reads each module's last membership before it publishes, so it needs the direct get on every
membership subject, as the controller's composer now grants (fix/1945-mesh-grants-publishes-memberships). The equality
test compares every membership the fixtures issue with the controller's MembershipFor, written by its own code.
The same line as mesh-controller f67d1553; the fixtures name a mesh-grants module and the lists were written again by
the controller's own composer on that branch, so the copy is held byte-equal to it.
mesh/check pass e0d1de047585 on 3db051be57ca | unit 1642 run, 752 kept, 2 skipped | integration 1710 run, 803 kept, 2 skipped | over bound: unit, inte…
The delivery builder exited twice on 2026-10-25 for a bucket the bus
refused at start, though it could build without it: a grant reaches the
bus with a delivery, so a refusal at start is often only a grant not
there yet. EssentialBucket stops a start naming the bucket and its grant;
OptionalBucket and NotWatched say what is not heard and go on. Here so
the seven delivery parts take one copy, not seven.
The grants part composes the bus's user list with the composer the controller uses (ADR 0547 §5), copied rather than
moved: the controller is not changed and keeps its own until it loses the lease. Left out: the bus client, the broker's
address from the environment and the client's two methods; the lease's two subjects are copied in.
One table test holds the copy to four meshes' lists and the server golden, written by the controller's own
internal/broker from the same fixtures (testdata/controller-user-lists.json).
mesh/delivery-group group fix/steps-bucket-ages held: a member is held: novox/mesh-sdk@518dae5ec9c8 (merged without a passing check: only a person decides what…
mesh/check checking 518dae5ec9c8 on 7f89cb51de15 | novox/mesh-sdk#81 in group fix/steps-bucket-ages
A Load error ended Pull, and the modules run it in a goroutine whose error nobody reads: every
build node would stop pulling when the new pauses state could not be read (or when Pull reached
restore before Serve). A node kept from building is held by capacity 0, not by a pause.
The steps state of the build modules filled to the bus's 64 MiB cap (27,370 keys of mesh-build-go,
none ever removed) and refused a passed check's end. Its keys get an age in the catalogue; the
pause key lived among them and would then have ended by itself. Puller.PausePut writes it to
delivery.PausesState instead; Put stays the fallback for a module not yet moved.
mesh-tools
jschoubben