Skip to content

spec(queen): review_first.t27 -- no backlog dispatch while finished work waits in review (Closes #10555) - #10562

Merged
gHashTag merged 2 commits into
masterfrom
spec/review-first-10555
Oct 18, 2026
Merged

gHashTag merged 2 commits into
masterfrom
spec/review-first-10555

Conversation

@gHashTag

Copy link
Copy Markdown
Owner

Closes #10555

What

specs/queen/review_first.t27: the owner's rule of 2026-10-18 -- the Queen takes nothing from the backlog while finished work waits in review.

  • in_review_queue() -- the review sweep's own filter, the rows the status page counts as reviewable (started, finished, review_state NULL or 'wait', not reaped, issue open).
  • admission() / admits() -- a backlog candidate is refused while that queue is longer than REVIEW_QUEUE_MAX (0).
  • kind_of() -- owner-assigned beats send-back beats backlog. Rework and assigned work are admitted at any queue length, so freed workers go to what the review asked for.
  • round_refused() + REFUSAL_TAIL -- the sentence a round reports when the rule stopped everything.

Why

Measured 2026-10-18 17:09Z on api.t27.ai/queen/status: 69 of 70 workers busy; in the last hour 102 dispatched, 56 finished, 11 reviewed; reviewable 29, unreviewed 362. The swarm started about nine bees per verdict, so the review queue only grew.

Evidence (Railway lab, t27c at 6ea5f51, zig 0.16.0)

  • t27c gen plus the zig test runner: 8/8 tests pass.
  • t27c seal --save then --verify: all hashes MATCH.
  • Negative controls: 17 hand mutants, 17 killed by a named test or a comptime invariant (bound 0 -> 1, <= -> <, rework refused, assigned refused, kind order swapped, each sweep-filter clause dropped or flipped, refusal clauses flipped, a code moved). The two that only the invariants catch were rerun with the invariant neutralized, and named tests kill them too.

Not here

The wiring into gHashTag/trios (queen-tick.ts candidate filter behind a flag, vendored wasm, t27c gen-ts constants) is a follow-up PR against queen, as is the Railway redeploy.

🤖 Generated with Claude Code

…ork waits in review (Closes #10555)

The owner's rule, 2026-10-18: the Queen takes nothing from the backlog while
there is a queue in review. Measured 17:09Z: 102 dispatched, 11 reviewed in
the last hour; reviewable 29.

- in_review_queue(): the review sweep's own filter (the status page's
  reviewable count).
- admission(): a backlog candidate is refused while that queue is longer
  than REVIEW_QUEUE_MAX (0); send-back rework and owner-assigned issues are
  admitted at any length, so freed workers go to review work.
- round_refused() + REFUSAL_TAIL: the sentence the round reports.

Railway lab: 8 tests pass, seal verify all MATCH, 17 of 17 hand mutants
killed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The ratchet reported it UNLISTED: the reference passes it and t27b passes it.
The four analog UNLISTED entries are inherited from master and left alone.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@github-actions

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-10-18 19:00:10 UTC

Summary

Status Count
Total Open PRs 50
PRs with Failing Checks 46
PRs with All Checks Green 4
READY 0
FAILING 46
PENDING 0
NO CHECKS YET 0

These columns do not partition: 0 + 46 + 0 + 0 = 46, and there are 50 open PRs. A PR is being counted twice or not at all.

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=fad12492937a != manifest seal=87e5cbd3ad94.
    The committed NMSE numbers were certified against an older compiler.rs.
    Run scripts/reseal-check.sh locally for the two-step reseal command (advisory; not a merge gate).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

queen: review first -- no backlog dispatch while work waits in review

1 participant