A pipeline placed long after its take-in found its old build done but holding no
number, and was told to reserve one; the series had moved to a newer fingerprint.
The step now says it is stale and names that fingerprint and its version.
Builds: design 47 H H-A-244 (2), H-A-226
After a builder restart, the run already building a module carried its reserved number while a re-run of the same
fingerprint reserved the next, so its publish was refused "has no version reserved or released" (main.61: the builder
0.1.111 vs 0.1.112, mesh-grants 0.1.52 vs 0.1.54, and others reserved twice within seconds). A take-over by another
fingerprint dropped the older reservation but let the older build run on and publish a number that no longer existed.
Reserve now answers the standing reservation of the same fingerprint while it is unreleased and above the catalog's
highest, a build again or a re-run alike (A-490b amended so). A take-over keeps the dropped version in Withdrawn and
withdraws the older build (ended superseded, ADR 0664 §4); a late publish of it is refused as withdrawn, naming the
fingerprint that took it over.
Builds: design 47 H H-A-225; G A-490b as amended; ADR 0664 §4
One source now has several catalog revisions, so a publish naming branch and source
alone does not say which pipeline published it. The seat mesh-builds'
package-published gains an optional revision (seatgen regenerated). The version
reserved for a build keeps the catalog revision number: from the picture's half when
the picture starts the build (the same revision its source comes from), from the step
ask's run when the ask reaches the build first. Its package-published says it. A
reservation from before the field says none.
Builds: design 47 H H-A-218, H-A-220
An ask of an older kept revision was admitted, but after a builder restart its fingerprint had never been heard on
the branch, so its build neither reserved a version nor was checked against one and passed silently. Each asked
fingerprint is now recorded as the branch's (its source revision, and first as the module's registration says),
without moving what the picture heard, so the newer fingerprint's build is not superseded.
Builds: design 47 H H-A-205, H-A-155
Two pipelines of consecutive revisions ran at once (main.45, 22:5x CET): the picture held source revision 46, whose
entries had moved on, so the older pipeline's builds waited for a picture that only moves forward and failed "revision
45 of main not in the builder's picture within 2m0s". An ask of a source revision below the newest the picture holds
now reads its catalog revision by number through the catalog seat's verb revision (kept while its pipeline runs,
H-A-147) and is built when that revision is of its source and names every entry it asks; only a revision above the
newest is waited for (H-A-155).
Builds: design 47 H H-A-205, H-A-147, H-A-155
The seat generator's check failed every unit step of the contracts mover (pipeline main.43): manifestdigest.go was
written by hand under go/seats/meshcatalog (598125a1), a generated root every file of which is the generator's. The two
helpers move to the seat's hand-written client meshcatalogclient and their callers in the builder and the indexer
follow; the generator then writes the contract as the module holds it. steprun's generator test now also holds this
repository's own contracts module in step under both roots.
Builds: design 47 G §370, G §578, H §774, H-A-200 (ADR 0608 §5)
The publishes the builder heard lived only in its memory, so a build waiting for a package published before a
restart waited the whole bound and failed although the registry held the version. Before it waits, the build asks
the Go package registry seat's versions once for each unpinned path and takes the newest in the accepted series
through PackagePublished; a failed read is said and the wait goes on by the published events.
Builds: design 47 H H-A-194, H-A-192, H-A-165
A step ask waiting for a package the entry names with no pin took the release number of the module holding it
("v" + release), and the lane of an entry taken in dropped the empty pin and asked its build without one, so the Go
build module took another package's version: both built against versions never published (main.41, v0.1.25,
v0.1.97, v0.1.103 of the builder's builds). The builder now hears each package registry seat's published (the
package path and the version it carries) and keeps the newest per series; a build, the step's and the lane's alike,
waits for one in the series its module's manifest accepts (packages, else the default) up to package-wait-minutes,
and past it fails naming the package. The release-number pin (VersionReleased, NewestReleased) is gone.
Builds: design 47 H H-A-192, H-A-183, H-A-184, H-A-185, H-A-165
The process tells the builder each version-released it takes in and
gives it the newest release its record keeps, so an unpinned package's
step ask wakes on the event and pins the release; the wait's bound is
the setting package-wait-minutes (default 10), carried as a file under
the state dir like revision-wait-minutes.
Builds: design 47 H H-A-165, H-A-156, H-A-155
A run on request copied its pipeline's source onto its build step ask,
so the builder waited for the run's candidate fingerprints in its
picture of the branch, where they never come, and failed every run past
the bound. The seat mesh-builds' contract leaves a run's source absent:
the runner now does. A revision the builder's view takes in whole that
changes no entry now also wakes a waiting ask, told by the catalog
client's new Held.
Builds: design 47 H H-A-162, H-A-155
With nothing numbered of the builder's own, the tool answered that a floor
would be asked at the next claim and looked again, doing nothing.
Builds: design 47 H H-A-159, G §369
The process gives the builder its picture of each watched branch (the
source revision the branch's view holds) and the bound of a step ask's
wait for its revision, read from the setting revision-wait-minutes
(default 2) carried as a file under the state dir like
toolchain-wait-minutes.
Builds: design 47 H H-A-155, H-A-156, H-A-141
Every look numbered one more patch in the state versions (claims carry no
source revision since H-A-48): nodeengine 0.1.5 to 0.1.36, beside the
reservation's numbers.
Builds: design 47 H H-A-159, G §348, §349, §369
The seat mesh-catalog defines no manifest-accepted any more: the builder
consumes the indexer's take-in.taken in its place (module.json, names.go's
patterns) and wakes its floor on every one, looking at every module again as
before (versions.go, the wake take-in.taken). With the in-progress commit
before it the builder reads no source stamp and its untagged suite is green.
Builds: design 47 H §764 (the indexer: its events manifest-refused and manifest-accepted, with their reader); lead answers H-A-48, H-A-49, H-A-66, H-A-116 (hq 508faf0c)
Claude-Session: https://claude.ai/code/session_01JG73KHUnBze48rdua8u2sy
A module value carries no source stamp (C1d): ModuleOf takes none from the
entry's inputs, a build's claim is numbered per fingerprint naming no source
revision and no merge, and the log's started entry names none. The seat list's
parts take no callers and resolve no caller's holder (mesh-sdk's NewPart without
callers, SDK-C). internal/... is green; cmd/mesh-delivery-builder still hears
manifest-accepted (names.go:21, versions.go:192, module.json:57), which the
seat mesh-catalog no longer defines: its floor's wake is H-A-49's, asked of the
lead.
Builds: design 47 H §764 (the indexer: the field pull, the source numbering); lead answers H-A-48, H-A-66 (hq f599b761)
Claude-Session: https://claude.ai/code/session_01JG73KHUnBze48rdua8u2sy
QueueDone removed; an item is deleted only with its release, drop or takeover; releases kept
one key per module (released-kept, 16); a plan that cannot be read ends the step could not run;
the wiring of the items to the reservations and the re-ask watch is tested.
Builds: design 47 G A-456b, A-457a, A-456a
Asked when the build ends and at the start, deleted by the release, ended loudly after publish-tries
tries and readable in builder_publish_queue; settings publish-pause, publish-pause-max, publish-tries.
The publish no longer waits on running or manifest-reads-changed.
Builds: design 47 G A-457, A-457a, §588
A trunk build reserving while its series is reserved by an older build never
released drops the older reservation's bound and finding and says so in the
journal. builder_reserve_done (a hand act, reasks mode reserve-done) reserves
the series' next version for a done trunk build that reserved none, with the
done key's digests, and queues its publish as builder_publish_again does.
Builds: design 47 G A-455, §369, §588
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)
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)
The builder leaves an event it cannot take for the bus to give again, and the bus drops it after the fifth delivery its
durable allows without a word: the 0.1.6 published was lost so. On that last delivery the builder now says it gives the
message up, naming its subject, what it carries (a registry outcome's name and version, a version-released's module and
version) and the last refusal.
Builds: design 47 G A-420
oci-artifact-registry's published of mesh-delivery-pipeline/program@0.1.6 was refused as 'not numbered by the builder'
and left for the bus, up to its consumer's five deliveries, and no package-published was said: the version is the
reservation's (§369), never an entry of the state versions, design 51's numbering by name, which G keeps only until
it is removed (§372). Such an outcome is now taken: the reservations mark it and say package-published (A-415). The
log of an outcome left for the bus names its delivery count.
Builds: design 47 G §369, §372, §52, A-415
distribution's tag of mesh-delivery-pipeline 0.1.6 was refused: reserved answered 'no entry carries the name
mesh-delivery-pipeline/program', since the lookup matched only an artifact's package name and a bundle's entry
names none. A bundle's or an archive's name is <module>/<artifact>, read from its reference through NameOf; a
name no entry carries is said in the builder's log with the entries looked at.
Builds: design 47 G A-418, §478, §572
builder_publish_again for mesh-delivery-pipeline 0.1.6 queued nothing after a restart: the holder was learnt only
from node-engine running events heard since the builder started. The Republisher now reads each registry seat's
holder from the catalog's SeatHolders, the holder function the builder's FollowSeat of the assignments takes;
publish again and a node's returned queue at it. A running still queues the versions waiting on its module.
A seat with no holder is logged by name.
Builds: design 47 G §588, A-416, A-416a
A trunk build's registry publishes during the build, before the builder records the digests on the reservation;
such a published was acknowledged and lost, so package-published was never said. It is now kept on the reservation
(name, version, digest) and marks its artifact when the digests are recorded; one that never matches is logged.
Builds: design 47 G §588, §52, §334, A-73, A-415
The seat event's own stream is heard from its first message through the SDK's HearSeatEvents with FromFirst; nothing of EVENTS is read and no new grant is needed.
Builds: design 47 G A-411 (replaces A-408's read of EVENTS)
The builder kept the catalog's versions-filled in memory only and its durable
consumer took the event once, so a restart left every build wait-held. It now
reads the last message of the subject from the events stream at its start (a
read that fails is retried with backoff in the background).
Builds: design 47 G A-408, §369, §372
The versions, bounds, reserved, publish-queue, acted, toolchains-heard and instances buckets were constants naming
mesh-delivery-builder; ownBucketsOf takes them from the module the credential names.
Builds: design 47 G A-296, §428
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
Its states' buckets, its finding observer, its build log entries and its step asks' asker take the credential's module (tools: the module the runtime serves them as) instead of constants of its name; builds.StateBucket/BuildsBucket replace the Bucket constants. The readers of the builds bucket (checker, view, grants) still name the builder: the next D1 subtask finds it by the seat mesh-builds.
Builds: design 47 G A-296, §428
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
buildrun's AssignerHolder asked the seat's verb for the holder every 5 seconds. The builder and the checker now follow
mesh-module-assigner's assignments through views.ProcessReader.FollowSeat with the catalog half's holder function (the
checker gains the catalog half's reader of §614 for it); the indexer reads the claimant from its own catalog
(Indexer.ClaimantNow) and follows its seat-holders view again at each refresh of its build modules. AssignerHolder goes; the
manifests consume the seat-holders view (the checker also the catalog half's events and verbs), and the assigner's
seat-holders first-view tool answers the checker and the indexer. The beds give each part a claimant and a seat-holders view
in place of the seat verb's answer; the indexer's beds hold the claimant from the start, so the merge test reads the newest
revision.
Builds: design 47 G §717, §614, A-298, A-302
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The builder, its tools bundle, the assigner (process and bundle) and the view's tools bundle answered
their seat list's again tool to the modules node-tools and mesh-healers. Their lists are now the
operator, seat:node-tool-runner and seat:mesh-healers, resolved by a holder function:
- meshcatalogclient's Branches.Claimant (moved from the builder's versions.HolderOf, which now asks it)
names the claimant of mesh-module-assigner from the trunk's half, and Branches.SeatHolders follows
that claimant's seat-holders view through the part's wiring. It drops the view and follows the new
claimant's on a catalog change, asks nothing on a timer, and answers seatlist.HoldersIn.
- the assigner answers HoldersIn from its own seat-holders picture (its world in the process, the
watched copy in the bundle).
The builder's and the view's manifests consume seat:mesh-module-assigner.views.seat-holders. The
assigner's seat-holders first-view tool answers both, by module as A-131 has it.
Builds: design 47 G A-298, A-295, A-296, A-210, A-131, §717, §616, §614
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The three constants named modules the builder no longer calls or reads: the gate and the checker
by nothing, the indexer only by a test's stand-in of a tool the builder stopped calling when the
catalog half named the requirements (A-119). A deletion: nothing reads them.
Builds: design 47 G §7, §428, A-210
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
A-163: the asker in `accept.<queue>.<asker>` is the asking seat's name (mesh-pipeline-runner,
mesh-catalog, mesh-wants, mesh-builds), never a nickname: the builder asks as mesh-builds (was
"builder"), the indexer's validation as mesh-catalog (was "indexer"), the assigner's runs as
mesh-wants (was "assigner"), the runner's steps on `accept.check.mesh-pipeline-runner` and
`accept.delivery.mesh-pipeline-runner` (were `accept.check.step` and `accept.delivery.step`, where
the holder read `step` as the asker). An asker naming none (the checker) asks on main's subject
(A-184). A build seat's queues are delivery, check and recheck alone (A-170, A-184): buildstep's
AskKinds holds no step, a holder takes a step from check or delivery, and the grants part makes a
worker per queue, never per kind. A holder keys a named ask's step `<asker>.<node>.<id>` and main's
`<node>.<id>` (§481, §483); readers watch their own name's keys. The composition step's step-progress
names the one caller mesh-wants' definition lists for step, read from the seat list (A-294).
Builds: design 47 G §481, §483, §543, §600, A-163, A-170, A-184, A-248, A-294
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
Cause: the bed reports go1.27.1 for both nodes at its start, and the test then reports go1.27.2. The builder's new
durable on EVENTS hears the old reports first, so for a moment it agrees on go1.27.1 (said as toolchain-agreed
go1.27.1 by n1, seen in a run's EVENTS); a netcheck entry heard in that moment is built at its old fingerprint
("the old fingerprint, naming go1.27.1, was built with go1.27.2", integration_test.go:767, twice on 2026-11-12).
The builder acts as ADR 0465 says on what it heard; the bed told it a story the test does not mean. The test now
purges the bed's go reports before its own, and asserts that no toolchain-agreed of go1.27.1 is ever said.
Builds: design 47 G §492, §604, A-282
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The floor wake (half.floor), the builder (half.b), ready and branches.Changed were set after reader.Start, so a
catalog view heard on the reader's goroutine read them while main wrote them: the race detector named
catalog.go:117 against main.go:501 in TestBuilderMakesADoneKeyWithTheDigests. They are now set before the start;
the builder cmd integration package then ran 21 times under -race with no race.
Builds: design 47 G §614, A-119
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
c5a6871f created the key once and never wrote it again, so a step that kept moving was bounded from its first
movement (issue 1930). Arm now creates the key when absent; one ending no earlier is armed already; one ending
earlier is written again by an update against the revision read, carrying the bound's time to live; a waiter
whose update loses reads the key again and takes the newer bound as armed. The test moves a step twice past its
first bound, with two waiters each time, and reads one write per movement, its time to live, and no delete.
Builds: design 47 G A-282, A-115, §544
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
stepBounds.Arm deleted the key and created it again, so two builder instances waiting on one ask raced: one's
create met the other's and failed "wrong last sequence", and the ask's build failed (the builder bed's two
instances together). Arm now creates the key only if absent; one already there is armed, and nothing deletes
it to make it again.
Builds: design 47 G A-282, A-115, §544
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The SDK's DeclareAskWorkers now declares each queue's two workers, main's and the named asks' (accept.<queue>.<asker>),
and the puller takes from both; the bed declared main's worker again filtering both forms, which overlaps the named
worker and the work queue refused ("filtered consumer not unique", 15 tests). The kind is a bed declaring by hand
what the SDK declares: the builder's bed was the only one whose own consumer overlaps the SDK's, and now declares
none. Behind it, the builder's seat server admits no caller of mesh-builds.reserved until the seat list holds
mesh-builds' definition (A-278); the bed's seat-definitions answers it from the contracts module's own file.
Builds: design 47 G §544, §600, A-115, A-278
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
AnswerSeat checks a call's caller against the seat's definition its caller
reads from the seat list it follows (A-186), refusing every caller while the
list is unread. The assigner (mesh-wants input), the builder (mesh-builds
reserved) and the indexer (mesh-catalog input) hand it the definition their
seat list holds once heard; the pipeline runner, which followed none, follows
one over its calls connection and hub, read again once its events are live,
and its manifest invokes seat-definitions and consumes the list's two events.
The runner's and the indexer's integration beds answer the seat list, and the
indexer's input test calls with a caller the bus proved (A-179).
Builds: design 47 G lead answers A-179, A-186 (join 11 area 6)
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h