Commit Graph
2541 Commits
Author SHA1 Message Date
jschoubben 37dd3dac13 Merge fix/a479-catalog-answers-in-time (ea106ea8) into the catalog-answers-from-memory branch
Builds: design 47 G A-487..A-487c, A-479

# Conflicts:
#	modules/mesh-delivery-indexer/cmd/mesh-delivery-indexer-tools/catalog.go
#	modules/mesh-delivery-indexer/cmd/mesh-delivery-indexer-tools/main.go
2026-11-19 01:24:39 +01:00
jschoubben 533a9e4ca9 Revision, module and provides answer from the held trunk revision; a failed re-read is said as behind
KeptTrunk also holds the newest revision and its entries, and the manifests the entries carry by digest.
The verb revision asked of the trunk's newest and the manifests the verbs module and provides read come
from it, with no read of the bucket or the manifests state; a candidate, a branch and an older number
still read the bucket. A failed re-read says the view is BEHIND and the next heard catalog-revision
reads again. The follower's state is named catalog-half.

Builds: design 47 G §421, lead answers A-487, A-487b, A-487c (03-DESIGN/01-to-be/47-milestones/G-22-lead-answers-2.md)
2026-11-19 01:19:32 +01:00
jschoubben efd8c7b34f Catalog half reads its own objects at the start and on the heard catalog-revision; no invoke of its own seat
KeptTrunk reads the newest trunk revision from catalog-revisions once at the start (a follower of the
bundle's views.Reader, adopted on the seat's event catalog-revision) and re-reads once for each heard
event of a newer trunk revision; reads list nothing. The invoke seat:mesh-catalog.revision and the
meshcatalogclient follower are removed; the consume stays.

Builds: design 47 G §421, lead answers A-487 and A-487b (03-DESIGN/01-to-be/47-milestones/G-22-lead-answers-2.md)
2026-11-19 01:04:33 +01:00
jschoubben 80cfaa175f Catalog answers from the newest trunk revision held in memory, and module reads one module
The tools bundle's view of the trunk (KeptTrunk) is made from the half of the catalog its own follower
of the seat holds (meshcatalogclient.WireCatalog on a views.Reader: the newest revision once at the
start, again on the seat's event catalog-revision), once per revision, and says each read with its
revision, module count and time. A read no longer lists catalog-revisions or fetches the newest
revision object. The verb module reads the one module it is asked (CatalogRead.Module). The manifest
consumes mesh-catalog.catalog-revision and invokes seat:mesh-catalog.revision for the follower.

Builds: design 47 G §421, lead answer A-487 (03-DESIGN/01-to-be/47-milestones/G-22-lead-answers-2.md)
2026-11-19 01:02:43 +01:00
mesh-tools 16b795af70 Merge pull request 'Pipeline runner: a tool for every seat verb no tool reaches (A-486)' (#250) from fix/runner-verb-tools into main 2026-11-18 21:37:19 +00:00
jschoubben 59ca36e55b G1: tools runner_start, runner_run, runner_accept, runner_ask_again over the runner's seat verbs
Each tool asks its served verb with the same arguments and answers its answer; the verb records the hand act, as for runner_rerun. Declared in the manifest.
The seat has no verb release: no tool for it.
Lead answer A-486 (G-22-lead-answers-2.md): tools are how a module is spoken to.

Builds: design 47 G §83, §84, A-486
2026-11-18 02:22:54 +01:00
mesh-tools 09be0bc27f Merge pull request 'Judge: a process's own secrets belong to the account it runs as (ADR 0680)' (#249) from fix/adr0680-credential-owner into main 2026-11-18 01:18:50 +00:00
jschoubben 3a47cccbee Judge: a process's own secrets belong to the account it runs as, whatever that account (ADR 0680)
The refusals of root and the operator account as an own secret's owner and as a bus process's account end; the
secrets-owner is held to the accounts an own secret's owner is; a process naming an own secret (a word equal to its
path) it runs neither as the owner of nor as root is refused, container modules excepted (lead answer A-485).
Restic's captured manifest is re-captured from mesh-catalog cd3b8fe8 by CAPTURED's recipe.

Builds: ADR 0680 §2, §3, §4; lead answer A-485
2026-11-18 02:11:47 +01:00
jschoubben ea106ea8b9 A-479: the seat mesh-catalog's verbs answer from memory, one read of the trunk view per call
The tools bundle's verbs module, modules, provides and dependents read the modules and the module graph key by key from
the newest trunk revision, and each key listed the whole catalog-revisions bucket again (6 s per call at 11 000 objects,
35 s with 12 callers); seat-definitions read its state once per seat. The view is read once per call (SourceValues), the
seat definitions come from a watched copy, a manifest is read once by its digest.

Builds: design 47 G (the seat mesh-catalog's verbs served by the indexer's tools bundle; ADR 0604 §2, ADR 0580 clause 1)
2026-11-17 22:43:08 +01:00
jschoubben d006132c2a A-476 (b): the watchman and the view no longer name an indexer build a revision waits for
A pending revision now holds a lost or unread message or a refusal; the watchman's finding and the view's wait say so, the take-in
event no longer requires WaitsFor (old values still decode), and the hand take-in comment loses the build.

Builds: design 47 G A-476 (b)
2026-11-17 20:37:32 +01:00
jschoubben 763d6a107a A-476, principle 4: indexer_take_in_pending takes a pending revision in again, and a group
A revision whose message is unread or lost, or held for room, has a hand recovery; its description says nothing of an indexer build.

Builds: design 47 G A-476
2026-11-17 20:34:49 +01:00
jschoubben a79b811113 A-476 r1: no merge waits for the build the indexer runs; leftovers of the validation and the hold go; a commit's scope is decided inside the recompute
ErrOwnBuildUnknown, its say and the own-build-unknown wait, TakeIn.Unknown and Failed go; the build only names the writer.
indexer_take_in_pending takes only a group in. The node-build-go accept.check and output invokes go. A module of another
repository stays as held inside of(), so a base of this repository's module is not given a value that is never written
(A-476 c, A-476a); one case in the toolchains test proves it.

Builds: design 47 G A-476, A-476a
2026-11-17 20:33:52 +01:00
jschoubben 40b6ced375 A-476a: a commit moves its own repository's modules and the importers of what it moved; A-471 toolchain test data
The recompute of a merge leaves a module of another repository as held unless its imported package versions changed
(ADR 0593 §3 stands). TestIndexerCountsOnlyThePlacedBuildNodesToolchains reports language versions (A-471).

Builds: design 47 G A-476a, A-471
2026-11-17 20:22:55 +01:00
jschoubben 9768d0b67d A-476: the take-in asks no validation, holds for no indexer build, and moves only its repository's modules
(a) validation_wait.go and validate.go go with the validation ask, its end consumer, indexer_validate_again and
Pending.Validation; (b) evaluate always takes the revision in with the build the program runs (holdFor and the build
comparison go); (c) a merge's recompute writes only the modules registered for its repository.

Builds: design 47 G A-476 (a), (b), (c)
2026-11-17 20:20:45 +01:00
mesh-tools 0adcfbbbdb Merge pull request 'A-471: a build toolchain is agreed by its language version, not its patch' (#248) from fix/a471-toolchain-language-version into main 2026-11-17 14:56:46 +00:00
jschoubben 109a3e1e49 A-471: buildrun and refusal texts name the language version
buildrun tests expect the agreed toolchain as ToolchainAgreement gives it; the guard's
refusal and a decline name the language version alone.

Builds: design 47 G A-471 (G-22-lead-answers-2.md)
2026-11-17 15:53:05 +01:00
jschoubben a66583236a A-471: a toolchain is agreed, reported and named by its language version alone
A build node whose installed patch differs from the ask's takes the ask: one shared
comparison in mesh-delivery-sdk buildstep (LanguageVersion, SameToolchainVersion),
used by the agreement (indexer, builder, checker), a build node's step guard and ask
check, and the builder's digest list. A fingerprint names the language version, so
every module's fingerprint changes once.

Builds: design 47 G A-471 (G-22-lead-answers-2.md)
2026-11-17 15:53:05 +01:00
mesh-tools 80a7061efa Merge pull request 'G-24 tests on main: the assigner hears presence once; grants composes each module's consumer from its own subjects' (#247) from fix/g24-tests-on-main into main 2026-11-17 14:35:31 +00:00
jschoubben 10d87dcc77 mesh-grants: a module's consumer is composed from its own subjects, not from its account's wide grant
ConsumerFor drew its filters from PermissionsFor, which since every account may subscribe to every subject (e1476cfd) is ">": no filter
named an event, so no module got an account consumer on EVENTS. The scoped permissions give them again. The integration subtests that
read each account's own subjects run with the wide grant off, as the unit tests do.

Builds: design 47 G A-434 (G-22-lead-answers-2.md: every account may do everything, the mechanisms stay whole), the grants part's account consumers (ADR 0561 §6)
2026-11-17 15:31:00 +01:00
mesh-tools 1cd984a4c7 Merge pull request 'A-473: the builder asks the tagged publish of a reserved version when its build ends' (#246) from fix/a473-builder-publishes-at-build-end into main 2026-11-17 14:09:34 +00:00
jschoubben b31924a2f5 A-473: the builder asks the tagged publish of a reserved version when its build ends
Reserved build ends (built-at set, every digest known) now queue that version at each
artifact's registry seat holder and drain it, as a holder's running does.

Builds: design 47 G A-473 (design 51 Publishing, A-417)
2026-11-17 15:06:38 +01:00
mesh-tools 7654838182 Merge pull request 'A-472: a pipeline's end waits on no other part' (#245) from fix/a472-runner-end-waits-on-nothing into main 2026-11-17 14:00:25 +00:00
jschoubben 3f51d83147 Runner: widens-read-bound and revisions-kept are the module's settings (A-472a)
Declared in module.json's config and written to state files as test-step-bound is; runner_cancel keeps its note.

Builds: design 47 G A-472a
2026-11-17 14:59:25 +01:00
jschoubben 0be2b3d01a assigner: the watchman's machine-away and machine-back have one hearer, the world's, so a durable on their streams no longer splits them
The composition step's own hub handler for machine-away and machine-back shared the durable of hearDelivery on the kept streams, so
each event went to one of the two and a machine away was often not known away to the delivery judge. The step now asks the world.

Builds: design 47 G §389, §64
2026-11-17 14:57:12 +01:00
jschoubben 871f75acd3 Runner: a pipeline's end waits on no other part (A-472)
The runner keeps each catalog revision it read, bounds every read widens makes to 5 s and says each read's duration, failure or cut in the journal and in could-not-run; runner_cancel reads the pipeline after an ask that did not answer.

Builds: design 47 G A-472
2026-11-17 14:55:27 +01:00
mesh-tools 4b75791d3d Merge pull request 'G-24: one composition run per candidate and machine, one test run per level, every reading step timed' (#244) from fix/47-G-shared-runs into main 2026-11-17 12:15:35 +00:00
jschoubben 4205f63786 Merge branch 'fix/47-G-shared-runs-watchman' into fix/47-G-shared-runs 2026-11-17 13:14:44 +01:00
jschoubben 2438f4a108 Merge branch 'fix/47-G-shared-runs-runner' into fix/47-G-shared-runs 2026-11-17 13:14:44 +01:00
jschoubben 9081966682 assigner: a case holding that a take-up says an end again under the open run's ask, a first end under the ask's id (A-470q)
Builds: design 47 G §719; A-470q.
2026-11-17 13:12:57 +01:00
jschoubben 16cb874ccc watchman: the integration test proves the end of a run's ask begins a test step's clock again (W1)
Builds: design 47 G-24 §732, §733 (watchman case)
2026-11-17 13:10:25 +01:00
jschoubben 20d6a68bfb assigner: a reader that leaves a run not yet taken, or whose place a rerun takes, has its wait ended superseded at once (A-470u)
Builds: design 47 G §727, §723; A-469y, A-470u.
2026-11-17 13:09:18 +01:00
jschoubben 21bef6c044 assigner: a drop ends each reader whose end is owed, once, and moves no reader (A-470r)
A carried reader that ended passed keeps its verdict; a reader marked rerun is ended by the drop alone, not again by the run's
move.

Builds: design 47 G §728; A-470r, A-470m.
2026-11-17 13:07:21 +01:00
jschoubben ae105b4631 runner: test 109 reads a step's end said again from EVENTS through a durable consumer
The runner keeps no message id on its step events (A-470t), so the bus's deduplication drops no end
said again (A-470q); test 109 now reads the tools' unit step-ended from the stream through a durable
consumer, its first end failed and the end said again by the move passed, both kept.

Builds: design 47 G §727; A-470d, A-470q, A-470t
2026-11-17 13:06:46 +01:00
jschoubben b58a8b2e7f assigner: an end said again is published under a message id naming the run that makes it say again, so the bus never drops it (A-470q)
A move names the moved run's ask, a take-up of a reader that ended the open run's, a drop the dropped run's; a first end keeps
the ask's id. Test 108 reads the ends through a durable consumer, as the runner hears them.

Builds: design 47 G §727; A-470d, A-470q.
2026-11-17 13:06:36 +01:00
jschoubben 533a076eb3 runner: test 109 asserts a queued ask has no bound in the state bounds
The case of a test run waiting with its holder present was asserted just after placing, so a
five-minute bound put back on the ask failed no test; test 109 now reads the state bounds after the
asks are put and holds only drops' keys there.

Builds: design 47 G §647 (as A-466y amends it), §535
2026-11-17 13:05:11 +01:00
jschoubben 4e33533bd0 assigner: a case holding that an end names its run's ask, so a replaced run's late end ends nothing (A-470n)
Builds: design 47 G §719; A-470n.
2026-11-17 13:04:19 +01:00
jschoubben 1f344bd47c assigner: a step asked again that joins a run not yet taken says its wait held, as join does (A-469y)
Builds: design 47 G §727; A-469y.
2026-11-17 13:04:02 +01:00
jschoubben 5b8fa23a8d watchman: the integration test seeds steps through the runner's stands verb; its findings name the runner's verbs (W1)
Builds: design 47 G-24 §731, §732, §733 (watchman case); A-469p
2026-11-17 13:03:57 +01:00
jschoubben ed78003b23 assigner: a composition reader names when its step said its end, so a take-up says no end again for a carried reader that passed (A-470l, A-470m)
The step's end, once said, is written on its reader in each run of its candidate naming it, by compare-and-set; a rerun's run
carries it with the reader; the take-up at the start skips a reader that ended and is not marked rerun.

Builds: design 47 G §719, §727; A-470l, A-470m.
2026-11-17 12:50:38 +01:00
jschoubben 20e966b9cd contracts: a composition reader names when its step ended, so a take-up tells a reader that passed from one that joined later (A-470m)
Builds: design 47 G §719; A-470l, A-470m. Generated by seatgen from the mesh-wants definition.
2026-11-17 12:47:39 +01:00
jschoubben 7697ef5988 assigner: a run's bound holds its ask, armed by compare-and-set, and its end comes before the input's delete (A-470n, A-470o, A-470p)
A bound was armed by a delete and a create with no revision, holding nothing, so a late ask-taken or
step-progress of a replaced run armed a clock that later ended the run asked next, and an expiry heard
after the bound was armed again ended a run that kept saying progress. A bound key now holds its run's
ask, published at the key's latest revision with its own time to live, and is deleted at its own
revision when the run read again no longer names that ask; a late signal moves nothing and is
journaled; an expiry ends the run only while its bound is still absent when read again and the run is
taken (bound.) or untaken (untaken.), the untaken expiry's readers said superseded as the key is gone.
A step-ended's run end is written before the kept input is deleted; a failed delete asks the event
again, whose redelivery finds the run ended and still deletes the input.

Builds: design 47 G §720, §719; A-470n, A-470o, A-470p
2026-11-17 12:39:12 +01:00
jschoubben 768213b816 Merge branch 'fix/47-G-shared-runs-watchman' into fix/47-G-shared-runs 2026-11-17 12:30:47 +01:00
jschoubben f9dd20e86d Merge branch 'fix/47-G-shared-runs-runner' into fix/47-G-shared-runs 2026-11-17 12:30:47 +01:00
jschoubben a96f873e21 runner: test 109, the test runs whole against the real bus
The runner whole against an embedded bus: one unit and one integration run per level read by both
artifacts, a failing run counted once, a rerun moving the other reader and not the accepted one, a
restart's take-up, progress past the bound, a replaced reader, a silent ask ending could not run with
its readers and its ask dropped, a second revision's own runs, a seat with no holder failing its
pipeline at once with its other steps not asked, test-runs and test-runs-drop over the bus, and the cap.
The finding writer's stream is hosted, as on the mesh.

Builds: design 47 G §733, §724, §725, §726, §727, §728, §535, §647; A-468f, A-470f, A-470g, A-470h, A-470i, A-470n
2026-11-17 12:30:19 +01:00
jschoubben ab9d9604b4 runner: a run's bound holds its ask, armed by compare-and-set, and its end is written before the step's finding (A-470n, A-470o)
A bound was armed by a delete and a create with no revision, so a late ask-taken or step-progress of a
replaced run could arm a clock for the run asked next, an expiry heard after the bound was armed again
ended the run, and a step-ended's run end waited behind its finding's event (3 s in test 109, past its
bound). A bound key now holds the run's ask and is published at the key's latest revision with its own
time to live, read against the run after it is written and deleted at its own revision when the run no
longer names that ask; a late signal moves nothing and is journaled; an expiry ends the run only while
its bound is still absent when read again; the run's end is written as its step-ended is heard, before
the finding.

Builds: design 47 G §720, §725, §726; A-470n, A-470o
2026-11-17 12:28:26 +01:00
jschoubben 2a90323af3 runner: no write of a test run acts on a read made before its latest change
The age and test-runs-drop deleted a run's key whatever was written after their read, an expiry or the
last reader's leave ended whatever run the key held by then, a rerun answered a write between its read
and its write as an error, and a step-progress heard as its run ended wrote a clock that later ended
the run asked next. A delete of test-runs is now a compare-and-set at the revision read, retried from
the latest read; a run's end ends only the run under the ask its caller judged; the last reader ends
the run in its own leave; a rerun writes its new run from the latest read; a clock is read against the
run after it is written, and a clock of a run not taken ends nothing.

Builds: design 47 G §725, §726, §727, §728; A-469l, A-469z, A-470d
2026-11-17 12:23:40 +01:00
jschoubben 00e1588ef9 assigner: no delete of composition-runs lands on a value newer than the one it judged
Every value write of the state was already a compare-and-set from the latest revision, retried on
conflict; the deletes were not. The sweep and the drop deleted a run's key whatever it held, so a
rerun's run replacing the key between the read and the delete was lost; the end of a run, the
ask-taken and a rerun deleted bound keys a newer run of the key may have armed. Each delete is now
at the revision just read, while the key holds the run it was judged from, judged again on a
conflict (the kind the runner's test 109 met in test-runs).

Builds: design 47 G §719 (readers added and removed by compare-and-set; the key deleted after composition-runs-kept), §720 (the bound keys)
2026-11-17 12:23:38 +01:00
jschoubben a439b8da47 runner: a carried reader with rerun false says no step-ended again at a take-up (A-470l)
The runner's take-up of a rerun's test run already ends only the readers marked rerun and the open
ones (endReader); this table case holds it there, so a reader that passed is never said again.

Builds: design 47 G §725 (take-up), §727; A-470d, A-470l
2026-11-17 12:16:01 +01:00
jschoubben a2f4e8b499 assigner: the running builds whose manifest is not read are said once per change of their set
The assigner said the same list of about 100 modules every two or three seconds, once per
pass (live on the stable node since 2026-11-17 11:02 CET, the second meeting); it now says
the list once each time it changes. Why their manifests cannot be read is the builder's
stream of build ends, which holds no end of a build made before 2026-11-11, said apart.

Builds: design 47 G G-23 (the assigner's long line, met a second time)
2026-11-17 12:11:38 +01:00
jschoubben 8e95c2486f assigner: a composition run's request names the revision's modules with no fingerprint (A-470j)
A module whose build can never be made left its machine not judged, so every step of the
candidate could not run; the case now carries none, the revision's words for each module it
gives no fingerprint, for the composer to blame that module alone.

Builds: design 47 G §721, §722; A-470j, A-470k
2026-11-17 12:08:44 +01:00