Repository navigation
claw: the merger merges as TRI CLAW and on workflow_run of the required checks - #10474
Merged
Merged
Conversation
…ed checks (Closes #10396) Mint a merge-pr token (contents write, pull_requests write) with actions/create-github-app-token and merge with it so the merge starts the downstream runs; fall back to github.token with a step summary when the App secret is unset. Add a workflow_run trigger for the required checks' workflows judging only that run's PR, a concurrency group per PR, and treat an already-merged PR as success. --match-head-commit stays on every merge. Narrative comments trimmed so the diff does not grow (foreign-line budget). Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
gHashTag
enabled auto-merge
October 18, 2026 13:29
Contributor
This was referenced Oct 18, 2026
Merged
gHashTag
added a commit
that referenced
this pull request
Oct 18, 2026
…the new master tip (Refs #10397) Census moved: quiet 'named a path but not quiet' 210 -> 220 and shell 'run: steps' 320 -> 324 (runner-does 301 -> 303). The claw-refresh job in gate-topology.yml adds run steps; the rest of the movement came from master (the merger workflow, #10474) while this branch waited. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
gHashTag
added a commit
that referenced
this pull request
Oct 18, 2026
…#10474 #10474 rewrote auto-merge-ready-prs.yml so the merger runs as TRI CLAW on workflow_run; the rewrite added two run: steps (both handed to bash by the runner, 299 -> 301) and six steps that name a path without a quiet shape. Master's cli-tri went red on that merge; blessed with tri census pin --bless at master 6070e9b on the lab, byte-identical. Refs #10469 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
gHashTag
added a commit
that referenced
this pull request
Oct 18, 2026
… merge (Closes #10397) (#10482) * claw: refresh -- update-branch every behind, conflict-free PR after a merge (Closes #10397) specs/claw/refresh.t27 decides which open PRs get PUT .../update-branch after a push to master (held, draft, fork without maintainer edits, conflicting, unknown and current ones are skipped; at most 30, oldest first; fail closed on unreadable records). The claw-refresh job in gate-topology.yml only lists, feeds gen/c/claw/refresh.c and calls the endpoint. The foreign-code shrink rule is paid by deleting two dead wave generators (scripts/gen_w658.py, gen_w659.py: referenced only by old wave reports). Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * claw: re-bless the census for the claw-refresh job (Refs #10397) tools/census/shell.txt: jobs 91 -> 92, run: steps 320 -> 322, runner-named shell 299 -> 301 (the new claw-refresh job in gate-topology.yml, two run steps, no shell: key). tools/census/quiet.txt: named a path but not quiet 210 -> 214 (the same job's steps name paths; none is a quiet shape). Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * claw: refresh -- tests for the 27 mutation survivors that were real gaps (Refs #10397) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * t27b ledger: name specs/claw/refresh.t27, the reference passes it (Refs #10397) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * claw: refresh -- drop the blank line at the end of the spec (review suggestion, Refs #10397) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * claw: re-seal after the final-newline fix and re-bless the census at the new master tip (Refs #10397) Census moved: quiet 'named a path but not quiet' 210 -> 220 and shell 'run: steps' 320 -> 324 (runner-does 301 -> 303). The claw-refresh job in gate-topology.yml adds run steps; the rest of the movement came from master (the merger workflow, #10474) while this branch waited. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
gHashTag
added a commit
that referenced
this pull request
Oct 18, 2026
…tforms' safe zone, a marketing clock (Closes #10469) (#10471) * t27f token_floor: drone flights instead of hard cuts, text in the platforms' safe zone, a marketing clock (Closes #10469) The owner on the #10451 reels: the text runs over itself, the transitions are too jerky (fly like a drone camera, as in The Matrix), and the reel should work as marketing. - reel_marketing.t27 (new): seven measured overlay sources and the safe rect taking the widest margin of each edge, the hook clock, the reset window, the CTA hold and size, the 15-30 s length; sources listed. - token_floor_scenes: each shot after the first opens with a flight -- a quintic Hermite in ln z and one in the screen px the centre sweeps, matching the drifts at both ends, a pull-back for long flights and a hover. Worst shove 20.1 px/frame^2 tall, 19.9 wide (bound 24); a hard cut would throw it more than 300. Tall layout inside reel_safe. - token_floor: a label leaving the scene's safe rect fades whole. - token_floor_mux: hook, CTA and safe-zone tests against reel_marketing. - clamp01 written once (pbr_sphere), so duplicate-bodies is green; the t27b ledger names pbr_sphere (pass) and token_floor (blocked, type [N]T), cap 74 -> 75 for the one new non-pass spec. Lab: test-report 0 FAIL, 0 vacuous on all five; seals rewritten; both reels 15.000 s at -14.0 LUFS; four tall shards equal the full render. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * census: re-bless quiet 210 -> 216 and shell 320 -> 322 run steps after #10474 #10474 rewrote auto-merge-ready-prs.yml so the merger runs as TRI CLAW on workflow_run; the rewrite added two run: steps (both handed to bash by the runner, 299 -> 301) and six steps that name a path without a quiet shape. Master's cli-tri went red on that merge; blessed with tri census pin --bless at master 6070e9b on the lab, byte-identical. Refs #10469 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * t27b ledger: max_not_pass 130 -> 132 for token_floor_scenes and token_floor_mux Both were pass at 1ef60ae and are now blocked on ExprCall(@sin): the drone flight's hover in token_floor_scenes calls @sin, and token_floor_mux uses scenes. The CI ratchet on f0d69e4 counted 132 non-pass entries against the cap of 130 (OVER CAP 1, every other column 0); the merge from master had dropped this branch's cap edit. Refs #10469 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #10396
Lane 3 of the TRI CLAW queue epic (#10393). Only
.github/workflows/auto-merge-ready-prs.ymlchanges.What changes
actions/create-github-app-token@v2step mints a token with onlypermission-contents: writeandpermission-pull-requests: write(themerge-prrow ofspecs/claw/github_surface.t27, named in a comment). The merge step uses it, so the merge starts the push-triggered workflows on master. IfCLAW_APP_PRIVATE_KEY/CLAW_APP_IDis unset (or the mint fails,continue-on-error), the merge falls back togithub.tokenand says which one it used in$GITHUB_STEP_SUMMARY.workflow_run: completedof the four workflows that carry the base branch's required checks (Schema Validation, Issue Gate, Spec Parse Ratchet, Own Language), pull-request runs only, not cancelled ones. The PR comes fromworkflow_run.pull_requests[0], else from the open PR whose head is the run's head sha (forks); with no single PR the run judges nothing and never falls back to a sweep. The cron stays as the backstop.--match-head-commitis untouched on the merge; the gate script (find-ready) is untouched, so the merger self-test passes unchanged.workflow_runand the bee's label event, and onesweepgroup for cron/manual runs. A single global group would not do: GitHub keeps one pending run per group and cancels older pending ones, and four required workflows finish per push. Because a sweep and a per-PR run can still meet on the same PR, the merge step now treats "already merged by another run" as success instead of a failure.Budget note
The own-language budget is shrink-only (added lines must not exceed deleted lines, 40 per file), so this diff is +32/-32: historical narrative comments (the 2026-09-19 laptop story, the 2026-10-03 measurement paragraph, one
|| echonote) were trimmed to a line or two to pay for the new lines. No logic was removed.Verification (Railway lab, not the Mac)
python3 tools/bees/merger_gate_selftest.py: exit 0, "0 failure(s) of 41".schedule, pull_request_target, workflow_run, workflow_dispatch;--match-head-commitis still on both merge paths.actionlintis not installed on the lab; run on the Mac as a read-only lint, it reports only shellcheck info/style notes (SC2086, SC2129) inside the unchangedfind-readyscript.Not verified yet
github.token.🤖 Generated with Claude Code