fix(t27b): ready uses master's last verdict past a running tip, or waits (Closes #6334) - #6342
Merged
Merged
Conversation
…its (Closes #6334) #6333 merged at 02:02Z with coverage=FAILURE (master PENDING): pr_ready acted on a non-required BLOCK but not on a non-required WAIT, so the Q29 wait fell through to READY. steward.t27 now waits on it, and a new master_state(running, last) decides what a check is judged against when master's newest run is still going: red below stays red, green below or no completed run in the window is PENDING (WAIT, never MERGE). fold_master keeps the verdict, its commit and the running flag; ready reports e.g. coverage=FAILURE (master FAILURE@f329e27c1, newer run going). 111 spec tests; gen-c regenerated, lab gen-check SAME. Refs #6063 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
6 tasks
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 #6334
Refs #6063
What happened
At 02:02Z
tri t27b watchmerged #6333 withcoverage=FAILURE (master PENDING). Master's tip run of coverage was only queued, so the comparison proved nothing (it happened to be safe: coverage failed on the 3 master commits below).Root cause:
pr_readyinspecs/tri/t27b/steward.t27acted on a non-required BLOCK (other == 2) but not on a non-required WAIT, so the Q29 wait (check_effect(2, 1, false) == 1) fell through to READY. Confirmed before the fix:pr_ready(True, "MERGEABLE", True, [None], ["WAIT"]) == "READY".Where the decisions live (steward.t27)
pr_ready:if (other == 1) { return 1; }after the required WAIT.master_state(running, last) -> u8: what master state a PR check is judged against when master's newest run of it has no conclusion. Red below a running tip stays red (master was already broken: READY, as fix(t27b): reference cache writes whole rows, keys on t27c content #6333 should have been, but now with the evidence); green below is PENDING (Q29, WAIT not BLOCKED); no completed run in the window is PENDING (WAIT, never MERGE).gen/c/tri/t27b/steward.cregenerated witht27c gen-c(/tmp/t27shared/release/t27c, which reproduces the previously committed C byte for byte from the old spec);tri t27b gen-checkon the Railway t27c lab:SAME fbaea4652dba(the lab notes its binary mtime precedes its source checkout d566906, a tooling-only merge).Plumbing (t27b.py)
fold_master(outputs, shas)over the last 12 master commits keeps{last, sha, running}per check: the scan goes past a running tip to the newest completed verdict and its commit. Cancelled/skipped/neutral are still passed over; a check with none of either stays ABSENT.readyasksr.master_state(...)thenr.check_effect(...); it reports e.g.coverage=FAILURE (master FAILURE@f329e27c1, newer run going)orcoverage=FAILURE (master PENDING, no completed run in window).{name: state}and the new{name: {last, sha, running}}.Tests (fixture-driven, no network)
scripts/ci/test_a_t27b_tick_reads_before_it_acts.py: fold past a queued tip findsFAILURE@f329e27c1; no verdict in window; ready for the fix(t27b): reference cache writes whole rows, keys on t27c content #6333 case (WAIT), red below running tip (READY + commit), green below (WAIT), plain PENDING (WAIT), settled red (READY), required check in new shape; watch with no completed master run -> WAIT, not MERGE. PASS: 0 failure(s).scripts/ci/test_the_t27b_steward_decides_in_t27.py: 111 tests in the generated C,master_statetable,pr_readyWAIT; withTRI_T27Cset, gen-c byte-identical. PASS.pr_readyline from the gen C turns 4 checks red (3 ready, 1 watch -> MERGE), reproducing fix(t27b): reference cache writes whole rows, keys on t27c content #6333.Live read-only:
tri t27b readynow prints e.g.#6252 BLOCKED Check L1 TRACEABILITY=FAILURE (master SUCCESS@6cf985a44).AGENTS.md debt line:
t27b.py1481 -> 1568.🤖 Generated with Claude Code