ComposeSent no longer runs the node-engine's own culprit search (fixed at eight modules), and the composer's
own loop, which reran that search at every try, is gone: ComposeSent holds the case's changed modules on the
machine one at a time, one composition each, at most blame_tries, and blames the first whose hold ends the
problem; past the bound the untried are named. Without blame_tries there is no search. A module the node
holds composing without the change is held as there, and two owners of one path are blamed outside the search.
Builds: design 47 G §721, A-468i, A-468l, A-469j, A-469k, A-470c
A module awaited on a machine whose revision gives it no fingerprint (the case's none) is held as its
node-engine holds it, the machine judged from that composition, and the module blamed with the revision's
none words, so only its own pipelines fail. An awaited build with a fingerprint keeps the machine not judged.
Builds: design 47 G §721, §722, A-470j, A-470k
The composer program has no settings, so the assigner bounds the search through the request: only the candidate's
changed modules placed on the machine are held, one at a time, the search ending at the first whose hold ends the
problem; past the bound the untried are named. A request with no blame_tries gets no search.
Builds: design 47 G §721, §733; A-468h, A-468i, A-469i, A-469j, A-469k
The assigner fails only the composition steps of the modules a machine blames, so the composer names them: the modules
the change refuses that the node held neither in its read nor composing without the change, answered structured by
ComposeSent, and both owners of a path two modules own, from the resolution's pairs. Blamed is always written, so a
verdict blaming none is told from an old composer's. The two request-file tests are one, for the test budget.
Builds: design 47 G §721, §733; A-468h, A-468l, A-469h, A-469j
The composer asks the trunk build of another node's module through the seat mesh-catalog's verb module (declared in the manifest); the golden grant lacked its call subject, so the ask hung to the node's bound in every pass and held the integration tests.
Builds: design 47 G-22 A-461b
A container's image counts as fetched for the parts check only when the composition
names the seat it came through; an upstream image its manifest pins by digest is no
part and is never 'fetched and not in the want's parts'.
Builds: design 47 G A-454 (§421's parts check, A-432)
A module not composed for want of its entry (entry-asked, entry-missing) and a want
whose parts differ write their have with a Wait of step compose and that reason, as
the kept data's holds do; the hold for a failed backup carries none.
Builds: design 47 G A-453 (ADR 0539 §16, §541)
The mesh-wide refusal ("nothing in this mesh provides") holds its consumers as the
node-wide one does; a module that wants the name only optionally is left absent
(ADR 0330 §3), never held. A held why still names the nodes that offer nothing
here. The two tests of the first commit are folded as rows into one table test
that replaces two existing ones.
Builds: design 47 G A-452
A requirement nothing answers no longer refuses the whole composition: each
module that wants it is held with the "nothing provides" words, as a refused
claim's claimant is, and the node composes the rest. Held modules stay in the
set, so what they provide still counts. The node is refused when every module
is held or no module of the set wants the name. Tests that used an unanswered
requirement to make a node's own set fail now use an ambiguous one.
Builds: design 47 G A-452 (extends A-405, A-407; ADR 0636 §3)
The running build's entry stays held (kept in the composition state) while the node runs it, though the catalog half dropped it; the agent no longer tells the composer it waits for a composition it holds, and the composer hands again when the entry answers.
Builds: design 47 G §296, §297, A-113, §614
A want of a released build names the module's version and no parts. Each artifact of the build's entry is
resolved at that version through the verb versions of its kind's registry seat: oci-artifact-registry for a
bundle or an archive, container-image-registry for an image, under <module>/<artifact> or a package's own
name. An upstream copy, which is not numbered, takes the digest its manifest pins. partsMissing, buildOf and
KeptBuildOf (and the shadow's added modules) read those digests; a want with parts is read as before.
A lookup that fails is said in the journal, naming the module, the version and the seat, and holds the module
whole: nothing of it is composed. It is asked again after a minute. The manifest invokes both verbs.
Builds: design 47 G A-430, A-430a; lead answer A-430b
Live on novox 2026-11-14 root-free answered 'judged no account yet' minutes after the start and nothing was logged:
the seat's serving and the first looks were silent.
Live on novox 2026-11-14 the G node-engine never ended a composing pass, so it never named its own module and never
served its seat node-engine: grants got no responders for root-free, judged novox not free, and the operator channel
held every Approve ask (the seed, hq#609) as no channel could prove who answered.
mesh/check failed (2): mesh-delivery-checker: no published of the tagged version of go-meshconditions of me…; and 1 more | 5caa760f5eaa on fd023f7473bc
The apply gives each process the declaration's For as MESH_NODE beside its own env, so a process that reads MESH_NODE (mesh-sdk, the zfs follower's writer and its node-service-manager calls) names its node.
Builds: design 47 G A-342
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
The whole test's web manifest keeps G's entry form (no build artifact, the entry read by fingerprint) and takes main's resources under if-config and config-lines.
Builds: design 47 G §421; design 42 zfs-owns-zfs H1-H5
Claude-Session: https://claude.ai/code/session_01NcbgfGKaxw984XyAPfLm4h
mesh/check being carried to 96a869063977 on f8acfaeed79d | from 0193949efe9f, mesh-catalog 117f933dacf0, mesh-delivery 7221a1e5cd7f, mesh-sdk e42b9491…
A size of a million MiB or more reached the file in exponent form and was refused.
Builds: design 42 zfs-owns-zfs §3, A-24
A key a layer holds that the manifest does not declare (scrub-pool during a rename) is not a config value of the module.
Builds: design 42 zfs-owns-zfs §9, A-15, A-24, A-25
The node record's reads come from mesh-sdk's ManifestReads, so building with S1
and S2 states word:resources.if-config, form:config-lines and form:config-json;
the record test now asserts them.
Builds: design 42 zfs-owns-zfs §4, answer A-13
`${config-lines:<key>:<prefix>[:mib-bytes]}` in a file's content writes one line
per entry (nothing for an unset value or empty list; a newline in an entry
refuses the composition); `${config-json}` writes the module's set config values
as one JSON object. Both join the sweep's table of fillers. Their rows join the
config table test (A-22).
Builds: design 42 zfs-owns-zfs §3, §9; answers A-15, A-24, A-22
When one record of a unit goes while another still declares it, what the going
record found of the unit is kept for the staying one when that has no finding of
its own, so the module's leave gives back the state found before the mesh. The
preview test and the new case are one test (A-22).
Builds: design 42 zfs-owns-zfs §2, answers A-21, A-22
The composer leaves out each resource whose `if-config` holds none of its
conditions against the module's config values (with defaults), and drops the key
from the rest, as the node-engine parses strictly. A resource that leaves is
removed by the existing orphan removal. One table test replaces the two tests of
setting_into_test.go.
Builds: design 42 zfs-owns-zfs §2, A-21 (declaration side), A-22
A failed resource of a module gates the rest of that module, which is reported
skipped with the failed resource named; the module's result names each one left.
A resource of no module, and other modules, go on as before. One table test
replaces two in apply_test.go; two fixtures adapted to the rule.
Builds: design 42 zfs-owns-zfs §1 (ADR 0679 §16 (a)), A-22