A Go module mesh-catalog removed stayed in the packages state, so every
publish of mesh-catalog asked the forge for its directory, the seat git
answered not-found, and the one-directory fallback (issue 1585) ran only
on an answer without the directory: nothing of the repository was
published and its packages' digests froze unsaid.
filesVia takes the seat's not-found as a failure of the tree, so the
fallback runs; publish drops the kept key of a package whose directory
the trunk's merge no longer holds, said in the journal; a listing that
fails for another reason raises the finding packages-not-published
(scope merge, id the repository), cleared at the next publish that lists.
Builds: design 47 H H-A-232, H-A-233, H-A-234; issue 1585
The watchman's go.mod names buildrun in a replace directive with no require
line, and the walk of the module graph followed require lines only, so
buildrun's requirement of the module assigner's client was never pinned and
every test step of the watchman failed "no pin". The walk now takes the
same-repository modules a go.mod replaces by a directory as required, with
the same same-repository test, so their go.mod files are read and their
other-repository requirements pinned, transitively.
Builds: design 47 H-A-224, H-A-201
With the SDK's guard of one pull per durable (pullOnce reads the consumer's
CachedInfo and releases on the consume's Closed), the fake of a branch's
durable embedded a nil consumer and a nil consume context, so
TestABranchDurableNotThereIsMadeAndHeard panicked. The fake now answers a
ConsumerInfo with its stream, name and a created time of its own, and its
consume context closes when stopped. No other fake in mesh-delivery embeds
jetstream.Consumer; the assigner's, runner's, builder's, watchman's and
grants' unit suites pass with the new SDK unchanged.
Builds: design 47 H-A-196, H §774
The tools bundle answered the catalog seat's revision verb from its view of the
trunk in memory, which ran one or more revisions behind what the indexer kept
until it heard the event, so the runner and the builder waited on revisions the
bus held all along. The verb now reads which revision is newest from the store
at the time of the ask (a listing and one object) and takes the view's entries
only when the view holds that very revision.
Builds: design 47 H-A-222 (1), H-A-205, H-A-213
indexer_take_in_again wrote TakeInAsk{Again} and DecodeTakeIn refused it: its
hand-kept list of act fields lacked Again, so the handler never reached its
Again branch. Every field of TakeInAsk now says by its tag takein whether it is
an act, an argument or of the ask; DecodeTakeIn reads the acts from the tags,
and a table test builds every hand tool's value by its own Run path, decodes it
and acts on it through takenIn, and refuses a field left unclassed.
Builds: design 47 H §774, H-A-208, H-A-210
The seat's key form of a pipeline is <branch>.<revision>, and its four per-pipeline
subjects are said on <branch>.<revision> (the definition and generated contracts the
same as the runner branch's). The indexer pins a revision for the pipeline keyed
PipelinesByBranchRevisionKey, and this branch's runner copy follows the renamed
generated functions; the wire form is unchanged.
Builds: design 47 H H-A-213 (4), H-A-219, H-A-221
Every whole recompute of the modules state (a toolchain reported, the toolchains read, the assignments of a build module, the indexer's start, a hand act) moved fingerprints with no kept catalog revision. It now runs as a re-pin does: when at least one fingerprint moved against the trunk's newest revision, it keeps the trunk's next revision of the same source, its movers the moved modules, its call-kind recompute (the seat mesh-catalog's enumeration gains it, seatgen regenerated) and its call naming the reason; one that moves none keeps nothing.
Builds: design 47 H H-A-211, H-A-213 (2), H-A-217
The indexer kept a registry's published in its series and then recomputed every module, moving fingerprints with no kept catalog revision beside the re-pin a release's package-published starts. It keeps the version in its series and recomputes nothing: a fingerprint moves only with a kept revision, one per module release, kept by the re-pin. The seat mesh-wants' pipeline-wanted subject names its last token revision, the number of the runner's key (its form unchanged).
Builds: design 47 H H-A-211, H-A-213 (2), H-A-213 (4) (the subject's token), H-pins-spec-and-gaps leftover L3
The seat mesh-wants' pipeline-wanted gains revision, the pipeline's catalog revision number (seatgen regenerated, additive). The assigner keeps the source the pipeline-ended names in its tier progress and says it beside the revision, the number of the pipeline's key; a progress kept before the field says the key's number as both. The indexer frees the pins of the revision the event names, not of another revision of the same source, and for an event naming none takes the source's lowest-numbered revision, the take-in's.
Builds: design 47 H H-A-213 (4), H-A-214, H-A-216
The seat mesh-catalog's CatalogRevision and catalog-revision gain call-kind, an enumeration the seat's schema validates (take-in, release, take-in-again), with module and version for a release and repository for a take-in again, beside call, the words for a person (regenerated by seatgen, additive). The indexer's three writers set them: a take-in's from the gate's call, a re-pin's on a module's release, a take-in again's; no reader parses the words.
Builds: design 47 H H-A-213 (3), H-A-214, H-A-215
The revision verb's source input, and the pipeline-wanted that frees pins by it, answer the lowest-numbered revision of a source (the take-in's), never a later re-pin's of the same source. TestIndexerWhole asks the exact first number: the bed seeds main.1 of source 0, so source 10's is 2 (no source is kept twice).
Builds: design 47 H-A-213 (1), (4), H-A-211
The take-in again keeps what it recomputed as a re-pin does: the trunk's next revision of the newest's source, its call 'take-in again of <repository>', its movers every module whose fingerprint moved, pinned for its pipeline, said as catalog-revision.
Builds: design 47 H-A-208, H-A-210 as H-A-211 and H-A-212 amend it, H-A-213 (1), (3)
A branch's revision number is the indexer's own count (the newest + 1); the gate's source is the revision's field source, which the take-in's dedupe and source-to-take read. A catalog revision carries its call (the seat mesh-catalog's CatalogRevision and catalog-revision gain call, regenerated by seatgen). A re-pin on a module's release runs on every watched branch holding a module that names its package (the first on the modules state, another on its newest revision's picture), and keeps, where a fingerprint moved, the branch's next revision of the same source with the call 'release of <module> <version>' and the moved modules as its movers, pinned for the release's pipeline, and says catalog-revision; no take-in.taken.
Builds: design 47 H S11, S12, S13, S14 (H-pins-spec-and-gaps item 1), H-A-211, H-A-212, H-A-213 (1)-(4: the pin of every revision a pipeline runs on), H-A-147
An entry names an unpublished package with an empty version, never left out, so a run on request's candidate checks its packages as an entry's (CheckEntry), not as an ask's. Built against the SDK head aee7e4c2 (fix/47-H1-sdk-empty-pin), which goes in the same merge call.
Builds: design 47 H S8 (H-pins-spec-and-gaps item 4, L10), H-A-204
The registry's published is offered again when the recompute of its users fails, as a failed series write already was.
Builds: design 47 H S15 (H-pins-spec-and-gaps item 3), H-A-167
A repository with no merge since the walk of the module graph keeps its
modules' direct requirements only, and no tool took it in again. The tool
asks the indexer, through the state takeins, to read the repository's trees
at the merge its taken key keeps and apply them as a take-in does, whose
recompute of the modules state is the one a publish re-pin runs; kept
requirements and pins gain the transitive packages. An unknown repository,
or one never taken in, is refused in the answer and as the tool's error.
It keeps no catalog revision yet: H-A-211's revision follows with the
re-pin's.
Builds: design 47 H H-A-208, H-A-210, H-A-201 (hq a7b876c1, d7d18ef0)
A reader twice could not read one entry's pins because the whole revision
(1.3 MB) exceeds what the bus carries; the tool answers one module's
packages, paths per artifact, fingerprint and manifest digest, at the
newest trunk revision (read as entryOf reads it) or at a given one from
its kept list and that one entry object.
Builds: design 47 H H-A-204 (hq f977c0e2); tools-are-how-a-module-is-spoken-to
A release re-pinned only the entries whose pins already named one of its
packages, so an entry whose module did not change and whose pins left the
package out was never re-pinned. Every kept entry whose pins or kept
requirements name a package of the released module is now re-pinned, of
any repository and changed or not; its fingerprint moves with the pin.
Builds: design 47 H H-A-204, H-A-165, H-A-167, H-A-201
Build asks were refused for a requirement with no pin: the import form's
versions name only what the last build imported, and the version line
leaves out a package none publishes, so the pins kept only those. The pins
now name every other-repository package of the module's go.mod graph (its
kept packages), at the newest published version of its accepted series, or
with no version while none is published, so the builder waits for it.
Builds: design 47 H H-A-204, H-A-201, H-A-165, H-A-167
The Go build module borrowed a sibling's version for a mesh package the
entry did not pin (a transitive requirement), and given no pin it now
refuses. The indexer now pins, in the entry's packages, every mesh package
of another repository that the go.mod of the module's directory, of each
Go package artifact's directory and of every same-repository module they
require transitively names. A module of the module's own repository is
never pinned, whatever a build reported: it is built from the tree.
Builds: design 47 H H-A-201, H-A-183, H-A-157, H-A-202
A Go build resolves the same-repository modules its go.mod files require at
the zero version, transitively, from the fetched tree; a change in one of
them must move the fingerprint, whatever the module's form. The indexer
walks the go.mod graph of the module's directory and its Go package
artifacts at the take-in and appends a line per derived directory and its
git object to the tree input; the entry's paths keep the manifest's form,
and the Go build module derives the same directories itself.
Builds: design 47 H H-A-202, H-A-172, H-A-157, H-A-198
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)
mesh-build-go and other modules import packages of this repository that no manifest declared, so no release
published them and the registry holds no version: their importers stay unpinned and fail past the bound. Each
module now declares its imported packages in the form they had before issue 1966's bootstrap took them out:
asks (assigner), builds and buildrun (builder), forgemsg (forge bridge), takein (indexer), tellerevents
(teller), viewstate (view), grantsstate and busgrants (grants), each a Go package of series 0.1. Each
module's manifest test asserts its packages again.
Builds: design 47 G §370, H-A-186
The pin fell back to the path's newest publish while no series value
held contents, which live none does, so a publish in a series a module
does not accept pinned it. The series a module accepts is the one its
manifest names, or the default; the pin is the newest version published
for the package path within it.
Builds: design 47 H H-A-185, H-A-184
A pin by the package path across every series moved a module to a
series it does not accept while its contents came from the accepted
one. The pin is the accepted series' newest published version where
that series gives the contents; only where none does yet is it the
newest published of the path. A publish in another series pins nothing
and moves nothing.
Builds: design 47 H H-A-184, H-A-183
At main.39 every importer of a nested package was pinned to the release
number of the module holding it (v0.1.97 of the builder for its package
builds), and the Go seat found no tag <package dir>/v0.1.97. A Go
package's version is its tag: the pin is now the newest version the
registry published for that package path (the seats' published, kept in
the package's series), never a module's release; a package the
registry never published stays unpinned.
Builds: design 47 H H-A-183, H-A-167, G §370
TestIndexerBuildsARunsCandidateAndDropsItWhenEveryMembersPipelineEnded failed on the build node
under load (pipelines main.29 and main.31). It restarted the indexer as soon as the candidate was
written, before the run-started was acknowledged (the write comes first), and it polled both
waits with sleeps: the same kind as the toolchains bed's restart fixed in 20a7bd0f. A stop
between the write and the acknowledgement leaves the run-started for redelivery after its
acknowledgement wait, which writes the candidate again once its run has ended.
The run is now one more step of TestIndexerWhole around its restart: run-started of run 9 writes
candidates.run-9 and the run's key (watched, not polled), the restart waits for the run-started's
acknowledgement beside the source-moved's, and the member's pipeline-ended after the restart drops
both. The separate test is removed: the module's integration tests go from 541 to 540; no test
added.
Builds: design 47 H §743
A tier-late pipeline's later-tier movers released a version, so the newer revision holds their
release record, yet their wants were not written: freeing their older release on the tier-late
event would remove the revision their want names. HeardPipelineWanted now reads the event's outcome:
tier-late frees nothing, and the same pipeline's later pipeline-wanted with outcome running frees as
before. The pins test's freeing sequence says a tier-late event first (nothing freed), then the
running one (the older release freed); the events test's row hands the outcome on. No test added.
Builds: design 47 H-A-174 (hq design/47-H1-answers-59 84e2514d), H-A-173
A passed pipeline's movers include ones the assigner never takes (superseded, no version released,
a tier-late pipeline's later tiers); freeing their older release on the newer pipeline-wanted would
remove the revision their want still names, and a mover releasing nothing would pin a revision per
pipeline for good. The release record is now written only for a mover that released a version in
the pipeline, and pipeline-wanted frees a module's older record only where the newer pipeline's own
revision holds that module's release record.
Tests: a pins table row (a passed mover with no version pins nothing), the freeing sequence (a
passed pipeline releasing no version frees no older release, the next one that releases does), and
TestIndexerWhole now says a pipeline-wanted on the bus after main.12's passed end and waits for
main.11's release to be freed and main.11 removed, so the indexer's hearing of pipeline-wanted is
tested end to end (review round 1). No test added.
Builds: design 47 H-A-173 (hq design/47-H1-answers-59 21297155), H-A-170, H-A-168
The pin judge removed main.23 and its entries when main.24 became current, while the assigner had
just written main.23's builder as the want: the node-engine asked its entry and was told none is
kept, holding the want for good. A passed pipeline-ended now pins the pipeline's kept revision
under a record release.<branch>.<module>, one per mover, beside the pipeline's own pin; the seat
mesh-wants' pipeline-wanted of a newer pipeline on that branch (its wants written to the end,
tier-late included) frees that record from every revision below the newer one, for each module the
newer revision's movers name. The indexer consumes pipeline-wanted (its manifest says so). A
revision a release pin holds keeps its entries, as removeUnlisted already does for any kept revision.
Tests: rows in the pins table (a passed pipeline's revision kept once no longer current; a failed
one's goes), the freeing in the pins test (a release past a failed pipeline, freed by the next
passed one's pipeline-wanted, twice changing nothing), and a pipeline-wanted row in the events
test. TestIndexerWhole now reads main.11's release beside its validate base, and waits for the held
source-moved's acknowledgement before it reads the consumer (the write comes first). No test added.
Builds: design 47 H-A-168, H-A-170 (hq design/47-H1-answers-59 87b43ec8), H-A-147, H §749
The import form's fingerprint left the pins out, so an importer whose
trunk build reported its imports never moved on a re-pin and was never
built again against the release. Every form now takes its sorted pins;
a build that stands keeps its fingerprint while its pins are the ones
it was built with, and a fingerprint with no pin stays as it was.
Builds: design 47 H H-A-172, H-A-167
A re-pin that left the fingerprint as it was left the build made
against the older pin as the one the mesh runs. The version line's
fingerprint now takes the pins, so a re-pin moves the importer, which
becomes a mover built against the release. An accepted series' older
version no longer stands over a newer release. A re-pin that fails
returns its error before the heard artifacts are forgotten, and a
package-published heard again for a released version re-pins, so the
redelivery recovers it. The release path and the seed's pins carry
tests that fail without them (review round 2).
Builds: design 47 H H-A-165, H-A-167, H-A-161
Pins moved to a new released version only at the next merge, toolchain
report or hand recompute. The release now starts the pass of H-A-161
for the modules whose packages hold a package of the released module
alone; every other module stands as held.
Builds: design 47 H H-A-165 (the indexer's part), H-A-161
With the go.mod's mesh requires carried at every take-in, requirementsOf
returned early for every module requiring a mesh package: its go.sum was
never read and keepRequirements never cleared an old version it named
(indexer_requirements for mesh-sdk still names v0.1.13). The go.sum is
read whenever its hash moved or this process has not kept it yet.
Builds: design 47 H H-A-157 (review round 1)
The indexer filled a module's packages with the commit of a module value
it holds, which Go turned into a pseudo-version of whatever that tree
was (mesh-delivery-sdk at an 11-19 commit, nodeengine pre-H1). A pin is
now the newest released version of the module holding the package (the
versions state version-released says), an accepted series' version
stands, and a package with none released is kept with no version: its
kind-1 edge stands and the builder asks no pin for it. A commit of
another repository moves a module only through a package of it the
module imports (A-476a), now that pins name no commit.
Builds: design 47 H H-A-161, G §370, H §746, H-A-157 (kind-1 edges)
An import form whose imports name no package of another repository, or a
build that stands with versions held empty from the go.sum era, wrote an
empty package set while the go.mod requires a mesh package: no kind-1
edge. The version line's versions of the go.mod's set stand then.
Builds: design 47 H H-A-157, G §47 (kind 1)
At main's revision 21, 64 of 82 Go modules carried no package edge: the
version line read the mesh packages go.sum names, and go.sum names none
required at the zero version (ADR 0573 §2). The take-in now reads them
from the go.mod's require lines, at any version, with no build needed;
the go.sum is read only for the requirements state. The sibling rule
(replaces by a directory) stands.
Builds: design 47 H H-A-157, H-A-136 (sibling rule), G §47 (kind 1)
A take-in that moves nothing keeps no revision for its source number, so validate's case before,
read as the step's revision minus one, named a revision never kept and could not run. The take-in
now writes, on the catalog revision it keeps, the field base: the newest kept revision below it,
the one it pins for its pipeline, in the same write; absent on the branch's first take-in. The
assigner's validate reads the base from the revision it already asks for; where the revision names
none the case before is empty and its end says so, never a could-not-run.
- mesh-delivery-contracts: the seat mesh-catalog's CatalogRevision gains base (schema, fixture,
generated Go and TypeScript by seatgen).
- mesh-delivery-indexer: keepBranchEntries names the base it pins; the pin test's rows check it.
- mesh-delivery-assigner: before reads the base; the validate test asks a revision with none, and
both stand-ins name the base as the take-in does.
Builds: design 47 H-A-151, H-A-154 (hq design/47-H1-answers-59 01842bbe), H-A-147, H-A-108 (3), design 47 H §743
TestIndexerCountsOnlyThePlacedBuildNodesToolchains failed on the build node under load
(pipeline main.18) in two ways, both of the test, not of the indexer's count of the
placed build nodes' toolchains:
- it restarted the indexer as soon as the module was written, before its take-in
ended: the stop cancelled the take-in, its source-moved stayed unacknowledged, and
the branch's durable (one message at a time) held the next one behind it for its
acknowledgement wait, past the test's 20 s bound. Each restart now waits for the
source-moved's acknowledgement; TestIndexerWhole's restart the same way.
- a source-moved said right after a restart could reach the stopped program's pull,
still counted until the bus closed its connection. restart now returns only once
the bus has closed every connection of the stopped program, heard on the system
account's disconnect events, with no sleep.
Under load (8 runs side by side, GOMAXPROCS=1) the test failed 5 of 80 runs before,
2 of 240 with the first change alone, and 0 of 600 with both; TestIndexerWhole 0 of 600.
No test added.
Builds: design 47 H §743
The pin judge removed the branch's previous revision in the very write that took the next one in
(pipelines main.13 and main.15 on 2026-11-20), so the validate step's case before could not be read
and the composition was never asked. The take-in now pins the newest revision it kept before its
write under the same pipeline record as its own, and the pipeline's end frees the record on every
key it pins; a take-in with no revision before it pins only its own and says so.
Builds: H-A-147 (hq design/47-H1-answers-59 310f8955ba16a1ca8ae9661d8196cab16e3d9597), design 47 H §743, §749, H-A-108 (3)
As the grants part gives a module's consumer on a seat event's stream kept by age:
without a start from now a made durable would replay every past source-moved of the
branch past the start's guard; the current revision is read at the start already.
Builds: design 47 H §743, H-A-134 (hq 8d30cc42), H-A-129, H-A-130
Claude-Session: https://claude.ai/code/session_01JG73KHUnBze48rdua8u2sy
catalog-revisions kept trunk.<n> and branch.<branch>.<n> objects beside the
branch-keyed revisions, the newest trunk one marked current, so a reader heard
revision 528 while its picture held 1. A branch's take-in now drops every such
object no record pins, says the count in the journal, and the revisions tool marks
each branch's newest revision current.
Builds: design 47 H §743, H-A-135
Claude-Session: https://claude.ai/code/session_01JG73KHUnBze48rdua8u2sy
The grants part composes no per-branch durable on the source-moved stream, so the
indexer never heard a source revision after its start. Its bind now makes the
durable itself, as a module's seat-event consumer is made (SeatEventConsumerConfig,
named <node>_<indexer>_<branch>, filtered to the branch, one message at a time),
and binds an existing one as it stands.
Builds: design 47 H §743, H-A-134
Claude-Session: https://claude.ai/code/session_01JG73KHUnBze48rdua8u2sy
At a clean state the grants part makes the per-branch consumers of source-moved only from the modules view, which the
indexer answers only after its first take-in, so nothing ever took the trunk in. At its start, and again at each
failed bind, the indexer now reads the branch's current source revision through mesh-delivery.source and takes it in
as a heard source-moved of it would, when its view keeps no revision of the branch at or above it: start true exactly
when it keeps none, every repository the answer names a head. No timer; binding on registration-changed is unchanged.
Builds: design 47 H §743, §768, H-A-129, H-A-130
Claude-Session: https://claude.ai/code/session_01JG73KHUnBze48rdua8u2sy
Review round 1, H-A-121 (2): beside the one-item form, all true drops
every pin of the state pins, each recorded through mesh-hand-acts and
asked of the indexer as the one-item form is; it answers the count
dropped and the pipelines the pins named.
Builds: design 47 H §765, H-A-121, H-A-124
Claude-Session: https://claude.ai/code/session_01JG73KHUnBze48rdua8u2sy
packageEntries, manifestBytes, writtenModules, WriteHeard, heldOldSections, keepOldSections,
the overlay's written and changed, moduleOfDir with moduleAt, EncodeSeriesPackage,
InputDigest, HoldSeats and the barBlock shape's functions are referenced by no file of the
repository; with HoldSeats gone the seat list's lock is a plain mutex WithSeats takes.
barShows and barOption stay: a judge test reads them.
Builds: design 47 H §764; lead answer H-A-48
Claude-Session: https://claude.ai/code/session_01JG73KHUnBze48rdua8u2sy