Commit Graph
56 Commits
Author SHA1 Message Date
jschoubben 04bb2d1398 Show placement candidates and each node's answer in indexer_placement, add indexer_placement_reread, and drop a group's candidates when its head changes no placement file, so a missed message or a revert is recoverable by tool (issue 1126, ADR 0559 §3, §7, §8) 2026-10-27 02:18:21 +01:00
jschoubben ad60ba3a46 Write a catalogue pull request's changed placement as placement-candidates <group>.<node> and delete them when it merges or closes, so each placed node composes it in shadow and answers (issue 1126) 2026-10-27 02:13:55 +01:00
jschoubben 0b980203fe Taken keys are filled only from merges the indexer applied or a drained forge stream's heads (issue 1280) 2026-10-27 00:18:23 +01:00
jschoubben 73bf290f58 The indexer fills every repository's taken key from the source's head at start and when the state appears (issue 1280) 2026-10-27 00:14:41 +01:00
jschoubben c12b870104 The indexer binds its taken state when it appears, as issue 1790 asks of every state a delivery adds: an indexer started before the bus holds it still takes merges in and says the key it did not write (issue 1280, ADR 0556 clause 1)
mesh/delivery merged: source.main moved to novox/mesh-delivery@4af36c69a761
mesh/delivery-group group fix/delivery-records-true merged: every member is merged
mesh/check checking c12b870104e8 on bd107d20154c | novox/mesh-delivery#80 in group fix/delivery-records-true
2026-10-26 22:25:39 +01:00
jschoubben 971cdb44e2 Merge remote-tracking branch 'origin/main' into fix/delivery-records-true
# Conflicts:
#	modules/mesh-delivery-indexer/cmd/mesh-delivery-indexer/main.go
#	modules/mesh-delivery-indexer/go.mod
#	modules/mesh-delivery-indexer/go.sum
#	modules/mesh-delivery-indexer/manifest_test.go
#	modules/mesh-delivery-indexer/module.json
#	modules/mesh-delivery-view/cmd/mesh-delivery-view-tools/main.go
#	modules/mesh-delivery-view/go.mod
#	modules/mesh-delivery-view/go.sum
#	modules/mesh-delivery-view/internal/view/read.go
#	modules/mesh-delivery-view/manifest_test.go
#	modules/mesh-delivery-view/module.json
2026-10-26 22:20:45 +01:00
jschoubben 6201f58c20 The indexer says per repository the last merge it took in, so the view knows its modules speak for that merge (issue 1280, ADR 0556 clause 1) 2026-10-26 20:49:00 +01:00
jschoubben 352066dde6 Review of L2: the assigner binds the indexer's placement state when it appears, with no restart (D1); the three new unit tests folded into existing ones, so no count rises (D2); pin mesh-sdk 3968849 (issue 1126)
mesh/delivery closed: a newer head of the same pull request
mesh/check checking 352066dde611 on 3dd94f714d6e | novox/mesh-delivery#76 in group feat/1126-nodes-compose
2026-10-26 20:12:00 +01:00
jschoubben 4f75925410 Nodes compose: placement from the catalogue, node records, self wants, own and pair credentials, core boxes and the authority from the vault (issue 1126)
mesh/delivery closed: a newer head of the same pull request
mesh/check failed: unit: git.novox.be/novox/mesh-catalog/checks/nobinaries.TestTheCatalogueHoldsNoCompiledProgram: modules/mesh-config/cmd/mesh-config…
One commit for the branch feat/1126-nodes-compose of mesh-delivery: its 16 commits squashed so its history does not carry the 17.8 MB compiled mesh-delivery-assigner a commit added by mistake and a later one removed (issue 1577). The tree equals the former head ff6c038d. The operator allowed this rewrite of the feature branch on 2026-10-26.
2026-10-26 19:30:18 +01:00
jschoubben 907e092abc mesh-delivery-indexer reads unread registered modules again in one background run, so a forge timeout holds no merged message or source revision past its ack wait (issue 1945, review round 1)
mesh/delivery merged: source.main moved to novox/mesh-delivery@b3a4b09f5b6f
mesh/check being carried to 907e092abc45 on b3a4b09f5b6f
The take-in tool's bound now says the take-in is still running and where its answer will be.
2026-10-26 14:11:12 +01:00
jschoubben 07b5f13474 mesh-delivery-indexer reads a registered module again once its repository becomes readable, and takes one in on ask (issue 1945)
A registered module whose repository the forge answered 404 (car-hunter, snake, before the forge's account could read
them) was read again only when the indexer started again or the controller wrote a new build record of it, so the grants
part's composed user list stayed short of their subjects and the take-over was refused. A failed read now keeps the
module as unread, read again on the next source revision and the next merged message (no timer of its own), and the new
tool indexer_take_in (the tools bundle mesh-delivery-indexer-tools, through the indexer's takeins state, as the
builder's builder_ask_again) takes one in now and answers what it stored or why not.
2026-10-26 14:01:34 +01:00
jschoubben 699cbfb7c1 Indexer: a registered module takes its registered ref, not a newer branch's record; no read at an older commit after a failure
mesh/delivery merged: source.main moved to novox/mesh-delivery@ead9ee2d138d
mesh/check checking 699cbfb7c131 on ead9ee2d138d
2026-10-26 12:52:29 +01:00
jschoubben 319d56be7e Indexer: a registered module is read from every repository it was built from, whatever the branch, and at the registered commit when the head is gone 2026-10-26 12:45:35 +01:00
jschoubben 646f83642a Indexer: read a registered module at the source's head when it names one, else at the registered commit
mesh/delivery merged: source.main moved to novox/mesh-delivery@11d3678984af
mesh/check pass 646f83642a01 on 11d3678984af | unit 171 run, 502 kept | integration 181 run, 568 kept | composition passed in 39ms on novox
2026-10-26 11:41:46 +01:00
jschoubben d9bfba1e62 Indexer: read one directory of a registered module's repository, leave a repository the source names to its merges, accept master as a trunk 2026-10-26 11:36:22 +01:00
jschoubben c44b9f91cd Indexer: gofmt 2026-10-26 11:23:27 +01:00
jschoubben 82da252cf0 Indexer: index the manifests of modules registered from their own repositories, so the grants part has no gap for them 2026-10-26 11:23:21 +01:00
jschoubben 8d38c6bac9 Write the source revision a candidate was previewed onto, so the builder merges its preview onto that trunk head
mesh/delivery closed: a newer head of the same pull request
mesh/check failed: integration: mesh-delivery-indexer/cmd/mesh-delivery-indexer.TestIndexerCountsOnlyTheAssignedBuildNodesToolchains: within 5 s of n2…
Candidate.Base was never written non-zero (review round 2): the builder's base lookup would have found no trunk for any
live candidate and built no preview at all. The indexer now writes, whenever it writes a candidate's fingerprints, the gate's
source revision whose trunk heads it read the previews onto; that head of the pull request's repository is the preview's
trunk. No mesh-sdk change: the field and its meaning were there. The builder's test now takes the candidate the indexer's
own test writes (testdata/candidate-written.json, kept equal to it), not a hand-made one.
2026-10-25 21:16:20 +01:00
jschoubben eaf27f36af Write a module's value only when the module changed, so a merge no longer rewrites every module of its repository
The operator, 2026-10-25: only if a module has a new version in commit x is a new value written naming commit x. A merge
wrote every module of the repository at the new commit, fingerprint unchanged (about 186 keys per mesh-catalog merge), and
the controller's take-in took each in again: twice tonight it flooded. Now a merge that gives a module the value it holds
but for the commit writes nothing, and the value keeps the commit its fingerprint first appeared at. The kind: a write
that differs only in where it was computed. The other instance, a source revision re-stamping Src on every module at its
repository's head after any merge anywhere, now stamps Src once. The trunk head stays the gate's source.main.
2026-10-25 21:13:59 +01:00
jschoubben 3e7f852472 Count a build node's toolchain report only once written since its assignment, and keep a computed fingerprint while builders disagree (issue 1940)
mesh/delivery held for a person: merged without a passing check: only a person decides what follows
mesh/check failed: integration: mesh-delivery-indexer/cmd/mesh-delivery-indexer.TestIndexerCountsOnlyTheAssignedBuildNodesToolchains: within 5 s of n2…
On 2026-10-25 20:56 assigning mesh-build-node to ace again brought ace's report of 2026-10-23 (v26.12.0) back into the
agreement before the module ran there; the indexer blanked every module's fingerprint to "builders disagree" and the
controller's take-in refused modules already computed and built.

- The toolchain gate (indexer, builder, checker; still one file in three) leaves a report out, said once, until its node
  writes it since its current assignment: the assignment's own time, or the revision that opened it.
- The indexer keeps a module's held fingerprint while a language has no agreed toolchain, when its kept inputs with the
  toolchains it was computed with give it again; only a new fingerprint (a merge, a base, an agreed toolchain moved)
  needs the agreement.
2026-10-25 21:03:16 +01:00
jschoubben 32542ebc14 Keep an unreadable manifest's inputs and judge its verdict again at start, so an upgraded SDK clears it (issue 1923)
mesh/delivery delivered
mesh/check pass 32542ebc14c3 on 7a88eb0a5609 | unit 122 run, 366 kept | integration 132 run, 414 kept | composition passed on novox
2026-10-23 22:17:09 +02:00
jschoubben fae4d48e09 Merge pull request 'R4g+R5 (mesh-delivery): builder and checker put asks, the assigner and its composition check, the view's wants and backlog (design 47)' (#27) from feat/r4g into main
Reviewed-on: #27
2026-10-23 17:49:01 +00:00
jschoubben d32ec6b7f8 Say what the builder, the indexer and the watchman do since R4g in their comments: no build-agent is asked, no build record moves a fingerprint, no checks-running bucket is mirrored (issue 1887, review round 2 item 6) 2026-10-23 16:47:36 +02:00
jschoubben 527f87f94c Merge branch 'feat/r4g-asks' into feat/r4g
# Conflicts:
#	modules/mesh-delivery-checker/cmd/mesh-delivery-checker/judge.go
#	modules/mesh-delivery-checker/manifest_test.go
#	modules/mesh-delivery-checker/module.json
2026-10-23 16:20:17 +02:00
jschoubben a4629bc649 Compute every module's fingerprint with no builder input and read no build-agent record for it, so a change of the builder rebuilds nothing (issue 1887 I1, ADR 0530 §1) 2026-10-23 16:14:19 +02:00
jschoubben d5868c178f Take the words of a module the change removes from mesh-sdk in the indexer and the assigner, so the assigner's reading of a removal cannot drift from what the indexer writes (issue 1297, B3b) 2026-10-23 15:44:00 +02:00
jschoubben 1dc9d80cfb Keep a key written ahead of the source head and a key with no path when reconciling (issue 1900) 2026-10-23 14:44:14 +02:00
jschoubben 26bfdbb637 Drop the indexer's modules keys the trunk no longer holds, at each merge and at start (issue 1900) 2026-10-23 14:31:55 +02:00
jschoubben 612658e788 Ask the forge for the objects of a merge's named paths in the indexer, the one reader of a manifest's paths, and narrow the module's tree from the answer (issue 1758, I6 reworked; ADR 0499 §2 §3, ADR 0495 §1, L8)
The manifest invokes gitea.gitea_commit_trees. A forge that cannot say gives the module none with its words.
2026-10-22 18:24:38 +02:00
jschoubben ece328d0cc Give a candidate's module its named paths' objects from the preview, so the builder's fetch for it names them (issue 1758, ADR 0499 §4; mesh-sdk CandidateModule.Paths) 2026-10-22 18:20:37 +02:00
jschoubben b39090aab9 Read manifests with mesh-sdk's ParseManifestForInputs in the indexer and the builder, so claude-code, dunst and xterm are built too (issue 1751)
The indexer's own copy goes. Pins mesh-sdk feat/r4e-sdk e76a5a3 in every module of the group.
2026-10-22 18:20:37 +02:00
jschoubben 23daa2c03c Take a module's tree input from the paths its artifacts name, by mesh-sdk's NamedTree, and keep their objects and each artifact's paths, so a module outside them moves for no change there (issue 1758, I7 and the indexer's I8, ADR 0499 §2 §3)
A merged message or a preview without the objects gives the module none, naming the paths. Every other module keeps its
directory's tree. Module.ArtifactPaths is filled for the checker (the lead's interface).
2026-10-22 18:14:30 +02:00
jschoubben e81ef8bb63 Read a manifest whose config value falls back to a predefined key, so claude-code, dunst and xterm get a new-form fingerprint (issue 1751, ADR 0348 §7)
The indexer holds no published list of predefined keys; each key a value names is taken as the value declares it.
No fingerprint input comes from the list: the gate and the controller still judge the value.
2026-10-22 18:11:21 +02:00
jschoubben af4eaefb14 Read the controller's build records only for the builder's own input and to fill a key's missing directory once, so a key written before paths gets one with no merge (issue 1851 I3, issue 1614)
Both reads go in R4g with build-agent. A path a key names is never overwritten, and no fingerprint changes.
2026-10-22 18:09:56 +02:00
jschoubben 9f7e082688 List a repository's modules from the merged message's manifests and the indexer's own bucket, not the controller's build records, so a new module gets a value and a key with no build asked (issue 1851, I1 I2 I4, ADR 0501 §6, ADR 0453 §6)
I1: a merge writes each module its manifests name, at its directory. I2: a pull request's changed directory, a
preview's touched module and a source revision's modules come from the modules bucket (a module a pull request adds is
found in the preview by its manifest). I4: the value of today's form (fp, none) is no longer written, since it was
computed from the build records; mesh-sdk keeps it readable until R4g. One commit: the three share applyMerged and its
tests. Pins mesh-sdk feat/r4e-sdk 9cb4d59.
2026-10-22 18:08:48 +02:00
jschoubben a2a5167d72 Write a preview of the conflicts for a head that does not merge, and give the candidate's modules the reason; pin mesh-sdk to the R4c-5a head (issue 1445) 2026-10-21 21:25:45 +02:00
jschoubben 4242fb5b0b The indexer gives a candidate's touched module the fingerprint its merge preview gives it, and wakes on the checker's preview.written without a restart; pin mesh-sdk to the R4c-5a head (issues 1445, 1681) 2026-10-21 21:17:09 +02:00
jschoubben 61fea8314f Compute from a build of build-agent only once a toolchain is read, so a restart drops no fingerprint; fixtures as live (issue 1727)
mesh/merge-gate head 61fea8314fe8 merged into trunk 318064ce206d — pass: builds mesh-delivery-indexer → novox; no bus step; every machine composes with the…
mesh/repo-check head 61fea8314fe8 merged into trunk 318064ce206d — pass: its merge-check.sh passed
mesh/delivery delivered
2026-10-21 16:45:51 +02:00
jschoubben 5f9ed2b886 Take the builder's Builder input from build-agent's newest trunk build record, not its absent fingerprint, so modules get a new-form fingerprint (issue 1727) 2026-10-21 15:49:33 +02:00
jschoubben 3bd2a065f9 The indexer writes '.' for a root module's path and the toolchains each fingerprint was computed with (issue 1615) 2026-10-21 11:33:49 +02:00
jschoubben 07a28b03c4 The indexer writes each module's directory into modules.<m> and keeps it (issue 1606, ADR 0478 clause 8) 2026-10-21 11:16:16 +02:00
mesh-tools a5da6079de Write each module once per merge and watch a build module's toolchains bucket on its toolchain.reported event, with no timer (issue 1444, review round 1)
mesh/repo-check head a5da6079de01 merged into trunk 340bf643883d beside mesh-catalog#389@bcee61ed — pass: its merge-check.sh passed
mesh/merge-gate head a5da6079de01 merged into trunk 340bf643883d — pass: group feat/r4c3-build-modules composed: every machine composes with the change as …
mesh/delivery covered after its failure: a later walk that contains it was delivered: plan-1792582266656254343 (novox/mesh-delivery at d68d7ef3 into main…
mesh/delivery-group group feat/r4c3-build-modules merged: every member is merged
2026-10-20 21:02:44 +00:00
mesh-tools 127ac64cce Leave an artifact without its lock file out of Locks (issue 1459) and read the build modules' toolchains (issue 1444), part 2 of 2 2026-10-20 20:48:30 +00:00
mesh-tools 74fef7dd5a Recompute the new-form fingerprints when a build module reports its toolchain (issue 1444), part 1 of 2 2026-10-20 20:47:35 +00:00
mesh-tools f69a561a16 Test how a manifest and lock hashes become a module's new-form inputs (issue 1450, design 47 R4c-2) 2026-10-20 18:55:54 +00:00
mesh-tools 27bfd83fa0 Test the kept manifests, the new-form reasons and the candidate's manifests against real files (issues 1449, 1444, 1451, design 47 R4c-2) 2026-10-20 18:55:29 +00:00
mesh-tools a75d9cc851 Keep manifests and say why no new-form fingerprint exists, so the group check can start from facts (issues 1449, 1450, 1444, 1451, design 47 R4c-2) 2026-10-20 18:54:28 +00:00
mesh-tools d650a41ba6 Copy the real manifests the indexer's new-form tests read (issues 1450, 1444)
Fixtures are copied from the writer, never written by hand: four catalogue manifests at mesh-catalog a67f0196 and build-agent's at mesh-controller f2d7bc94.
2026-10-20 18:48:08 +00:00
mesh-tools 05c180cf87 Add the indexer's source revision logic and its tests that the last commit named (issue 1344) 2026-10-20 00:31:19 +00:00
mesh-tools 409c41d4ae Stamp the modules with the gate's source revision, read from source.main (issue 1344) 2026-10-20 00:30:52 +00:00