Commit Graph
855 Commits
Author SHA1 Message Date
jschoubben 7ee2b42230 issue 1025: the nested-module check is a file of its own, and the gate test keeps the app module in its repository
mesh/merge-gate head 7ee2b42230dc merged into trunk b8c810a3499c beside mesh-catalog#348@40eed3a0 — pass: builds build-agent, mesh-controller, route-proxy …
mesh/repo-check head 7ee2b42230dc merged into trunk b8c810a3499c beside mesh-catalog#348@40eed3a0 — pass: THE CHANGE ALTERS ITS OWN CHECK (merge-check.sh):…
mesh/delivery delivered
mesh/delivery-group group feat/move-a-module delivered: every member is delivered
2026-10-17 17:27:46 +02:00
jschoubben e42becb479 issue 1025: the bus test makes the controller's consumers before it publishes mesh-delivery's word
mesh/merge-gate head e42becb479a4 merged into trunk b8c810a3499c beside mesh-catalog#348@40eed3a0 — pass: builds build-agent, mesh-controller, route-proxy …
mesh/repo-check head e42becb479a4 merged into trunk b8c810a3499c beside mesh-catalog#348@40eed3a0 — fail: THE CHANGE ALTERS ITS OWN CHECK (merge-check.sh):…
mesh/delivery-group group feat/move-a-module rejected: a member's own check failed
mesh/delivery superseded: a newer head of the same pull request
2026-10-17 17:18:56 +02:00
jschoubben 43b59c5dff issue 1025: the merge check vets and tests every nested Go module (ADR 0420 §5) 2026-10-17 15:58:46 +02:00
jschoubben 81715273d3 issue 1025: the operator's Approve of a reviewed group registers a moved module from its new repository (ADR 0420 §7) 2026-10-17 15:56:49 +02:00
jschoubben 085b661936 issue 1025: the merge gate judges a group's move of a module as a move (ADR 0420 §6) 2026-10-17 15:43:28 +02:00
jschoubben 21414307c8 issue 1025: a context naming the module's own repository is its own build source (ADR 0420 §4) 2026-10-17 15:41:35 +02:00
jschoubben 0f21fe7888 issue 974: review 2: a check stopped by its key says no verdict, an error judges nothing, and an unacknowledged check goes to a second builder
mesh/merge-gate head 0f21fe7888be merged into trunk 7d80155d1636 beside mesh-catalog#345@94f65ccc — pass: builds build-agent, mesh-controller, route-proxy …
mesh/repo-check head 0f21fe7888be merged into trunk 7d80155d1636 beside mesh-catalog#345@94f65ccc — pass: its merge-check.sh passed
mesh/delivery delivered
mesh/delivery-group group feat/events-and-queues delivered: every member is delivered
2026-10-17 14:12:50 +02:00
jschoubben a83b921af0 issue 974: acceptance 5: two merges within a minute leave each open pull request one queued check 2026-10-17 14:03:26 +02:00
jschoubben 3d09d5235f issue 974: the checks are queued by pull request on two work-queue streams, taken after a delivery build, dropped when judged, stopped when their key moves; the reordering and the forge-asking end are gone 2026-10-17 14:01:33 +02:00
jschoubben f883830975 issue 974: two work-queue streams for the checks of pull requests (CHECKS_READY, CHECKS_OTHER), their workers, grants and writer 2026-10-17 13:53:22 +02:00
jschoubben 96325c295c issue 974: the serving controller follows its own plan-moved, so a plan moved from the terminal is looked at 2026-10-17 13:47:47 +02:00
jschoubben 9ca2c2bfb3 issue 974: review 1: a verb's end, a provisioner client's word and a failed key read each ask or retry a look; the silent bucket ages; the acceptance test starts the rollout through the serving wiring 2026-10-17 13:45:41 +02:00
jschoubben 20c8d9454a issue 974: the step's acceptance test: a rollout advances on events with no ticker, a silent machine is named when its key expires, a replaced controller resumes 2026-10-17 13:33:49 +02:00
jschoubben 3d192a4857 issue 974: every plan is written through the one hook that says plan-moved, proven by reading the sources and by the moves a merge and a cut make 2026-10-17 13:30:27 +02:00
jschoubben 8af77c2c75 issue 974: a rollout advances on events (a look asked by each plan saved, report and expired silent key), the 30 s ticker goes, and each tier send is said as sent 2026-10-17 13:28:34 +02:00
jschoubben fd522f4322 issue 974: a rollout's wait for a machine is a key on the bus that its report deletes and its expiry names silent 2026-10-17 13:18:27 +02:00
jschoubben fbf5a3f909 The held-merges integration test waits for the batch's window to close, as a joined batch's window opens again (issue 921) 2026-10-17 12:42:30 +02:00
jschoubben a79182e53b Give the forge's merges a durable consumer of their own: no max-deliver, no reset, a growing delay while held (issue 973) 2026-10-17 12:28:27 +02:00
jschoubben e3f22c737f Round a merge window key's time to live up to a whole second, as the server rounds down (issue 973) 2026-10-17 12:20:24 +02:00
jschoubben 2b99d455fc Integration test of the merge path on a test bus: a held merge is recorded and carried, a window closes by its key's expiry (issues 921, 973) 2026-10-17 12:14:43 +02:00
jschoubben 9e99f31c23 Remove the five-minute merge catch-up, its signal S5 and its watchdog: the event and its durable consumer keep every merge (issue 973) 2026-10-17 12:12:19 +02:00
jschoubben df39a499ef Close a batch's merge window by the expiry of a key on the bus, heard by a watcher, and drop the five-second cutter (issue 973) 2026-10-17 12:10:30 +02:00
jschoubben 379a52c2e8 Hold a merge the controller cannot record now, and hand it over again, instead of taking it (issue 921) 2026-10-17 12:03:00 +02:00
jschoubben 300d3c1bf4 Cut a closed batch beside the open walks it shares nothing with, queue a related one with the reason, fold only related waiting walks (issue 973)
mesh/merge-gate head 300d3c1bf466 merged into trunk 6a62a7e552ea — pass: builds build-agent, mesh-controller → ace, g14, novox, shanks; no bus step; every …
mesh/repo-check head 300d3c1bf466 merged into trunk 6a62a7e552ea — fail: its merge-check.sh failed: --- FAIL: TestSSHClientOwnsTheOperatorsSSHConfig (0.01s)
mesh/delivery superseded: a newer head of the same pull request
2026-10-17 00:49:53 +02:00
jschoubben baf1501d81 A walk, a retry and a let-go wait only for a started walk they are related to (issue 973) 2026-10-17 00:46:27 +02:00
jschoubben a5ba47e23c Name the open walk a change is related to, own path related to all (issue 973) 2026-10-17 00:45:00 +02:00
fix-969 262b39ae09 Test the gate's handing end to end: snapshot, store, and the composition either side (issue 969) 2026-10-16 23:16:38 +02:00
fix-969 2f2a8bb57e Compose each machine in the merge gate with its engine's contract, so moved modules are handed there as on the mesh (issue 969) 2026-10-16 23:12:16 +02:00
jschoubben 87410d79ed Issue a verified sender again only over the membership that was read, so a newer push is not overwritten (issue 930)
mesh/merge-gate head 87410d79edd6 merged into trunk e231185572ba — pass: builds build-agent, mesh-controller → ace, g14, novox, shanks; no bus step; every …
mesh/repo-check head 87410d79edd6 merged into trunk e231185572ba — pass: its merge-check.sh passed
mesh/delivery delivered
2026-10-16 17:46:18 +02:00
jschoubben 1c3c3edc7d Prove on a real bus a lost verified sender is issued again once the judgement passes (issue 930) 2026-10-16 17:43:25 +02:00
jschoubben b2e2bf0638 Give a channel's verified sender back when the root-free judgement passes again, and say when it is dropped (issue 930)
A push composed during a failed judgement issued the router a membership without the verified sender and nothing composed it again, so every approval was refused. The self-check now issues it again within one run; a composition says which kind lost it and why.
2026-10-16 17:42:23 +02:00
mesh-tools f45b4cb213 Merge pull request 'Issue the provisioner client its own account when its module declares no broker (issue 917)' (#359) from fix/917-provisioner-account into main 2026-10-16 14:11:26 +00:00
mesh-tools b5da0b2161 Issue the provisioner client its own account when its module declares no broker
mesh/merge-gate error: builds mesh-controller → novox; no bus step; the check could not run: killed by hand
mesh/repo-check error: the check could not run: killed by hand
mesh/delivery superseded: a newer head of the same pull request
The catalogue's mesh-provisioner-client declares only provisioner-broker:
the client speaks on the bus as itself. module issue refused it for
lacking a broker, assign minted nothing, and push refused it for lacking
an account, so no machine could run the client (issue 917).
2026-10-16 13:01:03 +00:00
jschoubben 65f24cd45d Judge catalogue tests as the check does: the predefined list and templates read from the tree (issue 914) 2026-10-16 14:50:40 +02:00
mesh-tools 5eb5ea1774 Seal core credentials by the controller while the vault does not hold them, so the bus is never left out of a declaration (issue 908, ADR 0396)
mesh/merge-gate head 5eb5ea177482 merged into trunk 18c58b4903ad — pass: builds mesh-controller → novox; no bus step; every machine composes with the chang…
mesh/repo-check head 5eb5ea177482 merged into trunk 18c58b4903ad — fail: its merge-check.sh failed: 1 store test(s) listed in testdata/store-tests did not …
mesh/delivery superseded: a newer head of the same pull request
2026-10-16 10:27:53 +00:00
mesh-tools d984816844 Feature 57 (config and provisioning): release mesh-controller, the reviewed release branch merged with main, on top of main 2026-10-16 08:28:38 +00:00
mesh-toolsandClaude 151faa10a1 Judge a delivery group's composed check by the group's own controller, release branch included (issue 867, ADR 0239 §14)
mesh/merge-gate head 151faa10a13b merged into trunk a07dd78c3e90 — pass: builds build-agent, mesh-controller → ace, g14, novox, shanks; no bus step; every …
mesh/repo-check head 151faa10a13b merged into trunk a07dd78c3e90 — pass: its merge-check.sh passed
mesh/delivery delivered
The composed check of mesh-controller#346 and mesh-host#114 into feat/57-config-and-provisioning was judged by the release branch's controller, which cannot build the node-engine member's validator, so the group could never pass. ADR 0239 §14 has the group's own controller judge when a member moves it; ADR 0358 §2 changes the judge only for a pull request into a release branch. A group member in mesh-controller now judges, with the mesh-host member's validator, built in the isolated judge environment and kept by content; the statuses say the group's controller judged.

Co-Authored-By: Claude <noreply@anthropic.com>
2026-10-16 00:59:27 +00:00
mesh-toolsandClaude 3fdbea4204 Ask mesh-delivery for a recheck's group on a bus connection of its own (issue 867, ADR 0382 §1)
mesh/merge-gate head 3fdbea420440 merged into trunk af60d2719c0c — pass: builds mesh-controller → novox; no bus step; every machine composes with the chang…
mesh/repo-check head 3fdbea420440 merged into trunk af60d2719c0c — pass: its merge-check.sh passed
mesh/delivery delivered
`delivery check` runs as a command of its own, where no serving connection exists, so a recheck said its siblings could not be learned ("this controller is not on the bus") and mesh-host#114 was judged without its sibling. groupsAsked now asks through onTheBus, which lends the serving connection or dials one, as the hand-act log does. TestTheRecheckVerbAsksMeshDeliveryOnABusOfItsOwn drives the verb against a real responder.

Co-Authored-By: Claude <noreply@anthropic.com>
2026-10-16 00:27:57 +00:00
mesh-toolsandClaude de5292323f Find a rechecked pull request's group by its repository and number (issue 867, ADR 0382 §1)
mesh/merge-gate head de5292323f33 merged into trunk a47976be4e69 — pass: builds mesh-controller → novox; no bus step; every machine composes with the chang…
mesh/repo-check head de5292323f33 merged into trunk a47976be4e69 — pass: its merge-check.sh passed
mesh/delivery delivered
A recheck rebuilds the pull request without its head branch name, so the sibling lookup found no group and said nothing: mesh-host#114's recheck was judged without its sibling. With no branch name, the group is the one whose member_details hold the pull request; an older mesh-delivery says the siblings could not be learned. TestReplay867Recheck replays it.

Co-Authored-By: Claude <noreply@anthropic.com>
2026-10-16 00:13:06 +00:00
mesh-toolsandClaude 37fd73f378 Check a delivery group's member beside its siblings' merges, judged by the group's controller (issue 867, ADR 0382)
mesh/merge-gate head 37fd73f3781e merged into trunk 4446027108fd — pass: builds build-agent, mesh-controller → ace, g14, novox, shanks; no bus step; every …
mesh/repo-check head 37fd73f3781e merged into trunk 4446027108fd — pass: its merge-check.sh passed
mesh/delivery delivered
mesh/delivery-group group fix/867-group-siblings delivered: every member is delivered
A mesh-host change and a mesh-controller change that need each other could not both pass their own checks: each read the other at its base. The check now asks mesh-delivery for the group's open siblings into a base of the same name, reads each at its head merged onto its base, builds the judge from a controller sibling's merge in the isolated judge environment keyed on its inputs, and names the siblings in the verdict. With no answer from mesh-delivery the verdict says it could not learn them; a pull request with no sibling is checked as before. TestReplay867 replays the failure.

Co-Authored-By: Claude <noreply@anthropic.com>
2026-10-15 23:27:12 +00:00
jochen a18c17d3cc Name the machine in the secret-given message (issue 835, review round 2)
mesh/delivery-group group feat/mail-seats delivering: 1 of 3 delivered
mesh/merge-gate head a18c17d3cc5f merged into trunk 0ab61d1be8c5 — pass: builds build-agent, mesh-controller, route-proxy → ace, g14, novox, shanks; no bus…
mesh/repo-check head a18c17d3cc5f merged into trunk 0ab61d1be8c5 — fail: its merge-check.sh failed: --- FAIL: TestTheInstallersFirstUserListIsWhatTheContro…
mesh/delivery superseded: a newer head of the same pull request
ADR 0380 §1 says the message names the module and the machine it is
for; the words named no machine at the terminal and only the desk's
machine at the desk. The headline now ends "on <node>" and the tell
carries "subject": <node> in both cases.

Also covers two gaps review round 2 flagged: a tell failure other
than ErrNothingServes/ErrGrantRefused surfaces and raises no
condition, and the terminal's hand act now names who asked (not just
where), extracted into terminalGiveHandAct so it is testable on its
own.
2026-10-15 18:45:03 +02:00
jochen 435468c151 A secret given tells the operator once, opens no condition (issue 835)
ADR 0380: giving a module's own secret now asks the operator channel's
tell over the bus, naming the secret, where, when and who asked, and
raises no secret-given condition. Until nothing serves tell yet, or the
controller's grant is refused (link.ErrNothingServes / the new
link.ErrGrantRefused), the condition is raised exactly as before, so the
change composes whichever side of issue 797 merges first. A successful
tell also retires every open secret-given condition. The terminal's own
secret accept now records a hand act too, as the desk path already did.

The controller's bus grant for operator-channel.tell is added to the
composed config (golden fixture updated).
2026-10-15 17:45:45 +02:00
jochen 9aeb5f4493 Keep a restricted message past its stream's age as a record without its body, as ADR 0376 §3 says (issue 728)
Reverts 45b6c347, which kept nothing of such a message and raised an aged-out
condition: that departed from an accepted decision on a lead's word alone. The
concern behind it (the record keeps the subject, with no age) is its own issue.
2026-10-15 14:56:24 +02:00
jochen ce77a2caaa Merge branch 'feat/mail-hardening' into mh-785 2026-10-15 14:27:37 +02:00
jochen 0d1bb035cb Let a recorded module's move that restarts nothing roll out with its walk (issue 785)
Under a module's own record (its manifest's, or its irreplaceable data's), every
move waited for a person's push, even one that only changed a tools bundle: the
database server stayed behind on ace on 2026-10-15 although its walk said it
recreated none of its containers. ADR 0377 lets such a move roll out.

A move is judged per machine against the build that machine was last sent: it
goes when no container, mounted file, data path, data declaration, provider or
(for a module with data) long-running process changes. A person's record, the
bus and any move that cannot be judged still wait. The walk keeps per machine
which way each recorded module went and why, says it in its note and log, and
status names a machine behind with `push <node>` only for a move that waits.
2026-10-15 14:27:37 +02:00
jochen 45b6c347b2 Keep nothing of a restricted message past its stream's age, only count it (issue 728)
A record kept without its body still carried its subject, which names whose mail it was, with no age. The lead decided such a message is not kept at all: the notice is taken, the message counted, and the log and an aged-out condition say how many messages of the stream aged out undelivered. Unrestricted streams are unchanged.
2026-10-15 14:22:07 +02:00
jochen d03e24f423 Age a restricted stream's dead letters out with it and never show their body (issue 728)
DEAD_LETTERS kept a message given up on MAIL forever and showed it whole to any caller of dead-letters, outside the grant and past MAIL's week. Per ADR 0376 §3-§6: the kept copy carries a per-message time to live of what is left of its stream's age (a copy already past it is kept without its body), DEAD_LETTERS and EVENTS allow per-message TTL, the list and show give no body or own headers for it, delivering it again needs a minute left and the copy keeps the time left, and its max-deliveries condition clears saying how many aged out undelivered and from which stream.
2026-10-15 14:11:37 +02:00
jochen f75cd0239d Hear a restricted stream only from a module's own account, never the tool runner's (issue 727)
The tool runner's bus account carried the heard grants, consumer and bind word of every module it runs, and every agent session on that machine reaches the bus as that account, so any session could pull a granted module's consumer on MAIL. Per ADR 0376 §1 the controller now composes a restricted stream's consumer, grants and bind word only for a module that declares runs-as; registration refuses a consumes of a restricted stream without runs-as (§2), and a grant that composes nothing is said in doctor (D-heard) and in the controller's log on every raise and send.
2026-10-15 14:11:37 +02:00
jochen f7928a2bf7 Let an own secret's manifest declare a desk prompt (issue 789)
A family like mesh-mail's mail-credential-* names only the module and the secret's key at the desk, so the operator cannot tell which account is asked. An own secret (or a family) may now declare a one-line prompt in its manifest, with {member} filled by a family member's suffix; the module check refuses one that is not one line, over 120 characters, or holds another placeholder. desk_secret.go reads it from the controller's own catalogue and sends it to the desk as what, never from the call or the module's runtime; omitted when none is declared, so an old desk or module behaves as today.
2026-10-15 13:47:20 +02:00
jschoubben 9e11790bd7 Say a read-only mesh-cli call in no line of its own, and count them hourly (hq issue 762)
mesh/merge-gate head 9e11790bd74a merged into trunk 3ed08c3464e3 — pass: builds mesh-controller → novox; no bus step; every machine composes with the chang…
mesh/repo-check head 9e11790bd74a merged into trunk 3ed08c3464e3 — pass: its merge-check.sh passed
mesh/delivery delivered
A page polling plans every few seconds wrote about 14,000 journal lines a day. A line is judged a
read by the generic command verb's reading forms (ADR 0266), on its every word, so plans go,
conditions silence and settings set are still said; a refused or failed read is said too.
2026-10-15 00:28:17 +02:00