Reconcile master divergence (merge github mirror) + fix ingest observability path #12

Merged
nsaspy merged 50 commits from reconcile/master-2026-09-14 into master 2026-09-14 15:56:03 +00:00
Owner

Merges the 47 commits that landed on the github mirror master back into canonical master (Forgejo). No history rewritten, both parents preserved; the merge was textually clean and both parents' gates were verified.

Evidence (.prolog/runs/run-1599100*.pl, now tracked in-repo):

  • schema-lock PASS, doc-coverage 851/851, gen-api-docs zero drift
  • Integration suite at merge head: 33/36 HTTP-API (RED)
  • Attribution run at parent 8533dc4: identical 33/36 -> merge introduced NO regressions; canonical master was already broken (PR #11 called observability-active-p unqualified in STAR.ACTORS)
  • Fix: star:observability-active-p (448a0ff)
  • Fixed head: 36/36 HTTP-API, 25/25 valkey-lease, 9/9 couchdb-view (GREEN)

After merge, github mirror master will be fast-forwarded to the PR merge commit.

Merges the 47 commits that landed on the github mirror master back into canonical master (Forgejo). No history rewritten, both parents preserved; the merge was textually clean and both parents' gates were verified. Evidence (.prolog/runs/run-1599100*.pl, now tracked in-repo): - schema-lock PASS, doc-coverage 851/851, gen-api-docs zero drift - Integration suite at merge head: 33/36 HTTP-API (RED) - Attribution run at parent 8533dc4: identical 33/36 -> merge introduced NO regressions; canonical master was already broken (PR #11 called observability-active-p unqualified in STAR.ACTORS) - Fix: star:observability-active-p (448a0ff) - Fixed head: 36/36 HTTP-API, 25/25 valkey-lease, 9/9 couchdb-view (GREEN) After merge, github mirror master will be fast-forwarded to the PR merge commit.
Concurrent acquires could surface :outcome-unknown or :timeout under
load even when the backend answered and no authority was violated:

- wait-for-input also returns NIL on EINTR, so a GC/scheduler interrupt
  of select was mistaken for a lost reply; retry the reply wait until
  the operation deadline is actually exhausted.
- the pool condition-wait raised valkey-pool-timeout without re-checking
  the pool predicate, so a notify racing with timeout expiry stranded an
  idle connection and cascaded to other waiters; hoist the pool deadline
  and re-check closed/idle/slot at every wake.

Provision the 100-way concurrency test store with operation-timeout-ms
2500 (bounded by its 3000ms caller deadline) and add a deterministic
regression test that parks an idle connection without a notify.
The checked-in generated docs were stale relative to source at
4721efa (the drift that PR #151-era #153 already fixed in 16b06d2);
regenerate here so the first-merged branch is doc-coverage green.
fix: make valkey lease concurrency outcomes deterministic
- add negative validation: an operation missing a required field must
  fail strict v0.9 validation with document-schema-validation-error
- assert schema lock release_version 0.9.1 and operation in
  required_dtypes in the python contract test
- make the schema-lock checker fail closed when release_version is
  missing or a required dtype lacks a required-fields mapping
- refresh the converged star-cl pin in the runtime lock docs
feat: consume StarIntel operation 0.9.1 contract
fix: enforce tenant isolation on target listing
fix: retry transient Rabbit consumer startup failures
test: exercise non-default target tenant over HTTP stack
chore: add issue-first OpenCode YAGNI execution loop
Reunifies canonical master (Forgejo, 11 unique commits: valkey lease
runtime + observability) with the 47 commits that landed on the github
mirror since 4721efaf. No history rewritten; both parents preserved.
PR #11 wired observability hooks into star.actors:publish but called
observability-active-p unqualified; STAR.ACTORS neither uses nor imports
it, so every ingest publish (POST /new/document/*, /documents/bulk)
failed with STAR.ACTORS::OBSERVABILITY-ACTIVE-P undefined. HTTP-API
integration suite went 33/36 on canonical master and on the
reconciliation merge; qualifying as star:observability-active-p (the
defun in observability-gserver.lisp, exported from :star) restores the
ingest paths.
chore: track .prolog verification and KB state
Some checks failed
Docstring coverage / doc-coverage (pull_request) Failing after 1m15s
cca6b60099
Operator directive: verification + KB state must ride with checkouts,
branches, and worktrees (worktrees at detached heads previously had no
.prolog at all). Un-ignore .prolog/ entirely; keep only the transient
.facts.lock ignored. Digest computation in prolog-verify already
excludes runs/, facts.kb, verify.pl, result.json, so tracking them adds
no verification friction.

Evidence in runs/: merge-head gates (schema-lock, doc-coverage 851/851,
doc drift clean), integration RED 33/36, attribution run proving parent
8533dc4 fails identically (merge added no regressions).
nsaspy merged commit e4ba2f63d3 into master 2026-09-14 15:56:03 +00:00
Sign in to join this conversation.
No reviewers
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
starintel-labs/starintel-server!12
No description provided.