The eleventh cut and the move of next, read after the first rollback: release/fixtures/cut.v0-1-10.json (#197's ruling 9′, second of three)

Merged
DO
5d ago
+851-10

This records the v0.1.10 cut as release/fixtures/cut.v0-1-10.json, and tells the cut test of a rollback made after a cut. It is the second of three for ruling 9′ on "Release: cut an immutable tag, move next, and roll back in one action" (#197); the first, "The first rollback and the move forward after it, the fixture alone: release/fixtures/rollback.v0-1-10.json (#197's ruling 9′, first of three)" (#785), has merged. Part of #197; it does not close it.

Model-Lens inputs touched: none. The change is the fixture, tests/test_release.py and tools/mutants.py. It carries no record.

Size. 45,154 characters and 923 lines by git diff; 52,142 characters as the Kernel counts a diff, with its lines numbered, against a budget of 60,000.

What GitHub answered

  • Release run 43270689165 (release.yml, success) cut factory/v0.1.10 at e719181a, the merge of "An act's read again of its item asks when that item's Runs began, all of them past the cap (#724, record 0199)" (#760). It moved factory/v0-next there and recorded Deployment 7744818544. The fixture holds the run, its one job, and the job's log at the step.
  • Eval run 43267765670 (eval.yml, workflow_dispatch, success) ran on e719181a. It ended at 17:32:49Z, before the cut was asked at 17:33:27Z, and the eval workflow lists it as its only run on that commit.
  • Read at 22:45Z, after the first rollback and the move forward again (22:12Z to 22:13Z, #785's fixture). This is the one way this cut's fixture differs from every cut's before it:
    • factory/v0.1.10 and next both name e719181a: next is on it again.
    • factory/v0.1.9 still names b61f0ecb, where its cut left it.
    • next has thirteen Deployments, newest first: the move forward's, the rollback's, then v0.1.10's own down to v0.1.0's.
    • The cut's own Deployment reads inactive above its success: next moved off it at the rollback. v0.1.9's reads the same, as each earlier Release's has at the next cut.

The reads went through the Maintainer's login and were trimmed to cut.v0-1-9.json's fields and headers, read for read. tools/redact_trial.py redacted them outside the repo, and only the clean output was read. It took out the names of steps[1] and steps[6], the token's mint and its post step, as for v0.1.9. removed says what was taken.

What the cut test learns

tests/test_release.py::test_the_next_cut_moved_next_on reads each cut after the first; v0.1.10 joins its list. One helper is new, deployments_after: the Deployments recorded after a cut by steps that were no cut, read from the fixture of those steps. Today that is one entry, the eleventh cut's, from #785's fixture.

HeldHow
The read of the Deployments after the cut was answered 200asserted before its body is taken
Those two Deployments are the move's and the rollback's, by the ids each run printedagainst printed.deployments of both items
The steps were of this cut's Release and came back to its committhe rollback fixture's version and its newest Deployment's commit, against the cut's
The whole list, newest firstfor the cut and for every cut below it: the Deployments after it, then its own. So a later cut's list places the rollback's two between its own and v0.1.10's with no further change
The cut's own Deploymentinactive above success where steps followed it; success alone otherwise
The Deployment next moved off at the cutthe entry right below the cut's own, inactive above success

The next cut's fixture cannot carry the whole list. It grows by one Deployment a cut, and by two for a rollback and a move: this fixture is 52,142 of 60,000 with thirteen. The fixture of v0.1.11 will page the list, and the helper's expected list will be cut to the page. That is the next cut's pull request, not this one.

Checks it adds

  • Twelve mutants of the new fixture, each killed by the cut test: another eval workflow, run or commit; an eval that ended after the cut; another run listed on the commit; v0.1.9 moved; the cut's Deployment not v0.1.10's; its status not success; the cut's Deployment still active after the rollback; v0.1.9's Deployment still reading only success; the Deployment above the cut's not the rollback's; the newest not the move's.
  • Two mutants of #785's fixture, each killed by the cut test: the read of the Deployments answered 500, and the fixture another Release's.
  • release/AGENTS.md's cut row already names the test, and is unchanged.

Checks

CheckResult
make checksOK, at 872c9c6, main merged in once #785 had merged as a184082c
tests.test_releaseOK, 34 tests
Mutants of the cuts' fixtures and of #785's, run alone on an export104 of 104 killed, at c40a2a2
gitleaks on the fixtureNo leaks
Full make verifyNot run locally, by the Maintainer's standing change of 2026-11-09; CI's verify is the answer
ReviewerHeadUnread partsBlockingMajorMinorInfo
agy (Carbon)c0c923bpart 2 of 2 not run (quota), owed0000
musec0c923bnone0100
claude-reviewc0c923bnone0021
claude-securityc0c923bnone0000
astrac40a2a2none0000

Astra read the head after the fixes. The others' findings were gaps in the test, so none read again.

FindingWhatDisposition
muse #1, claude-review #1The Deployments after the cut were read from an answer the test did not hold to 200fixed in c40a2a2: asserted, with a mutant of #785's fixture
claude-review #2No listed mutant needed the two assertions on #785's fixturefixed in c40a2a2: the two mutants above
claude-review #3 (info)The helper could not read the next cut's fixture, and would have read the wrong Deployment as the one next moved offfixed in c40a2a2: deployments_after, and the Deployment moved off is the entry below the cut's own

Review threads, three, all resolved

ThreadDisposition
cubic: removed says the run's head commit is taken, yet head_sha is therenot changed: what is taken is the run's head_commit object, and the file holds none; head_sha is kept and the test reads it. The words are those of cut.v0-1-8.json, cut.v0-1-9.json and rollback.refused.json
cubic: the helper's docstring says None where it returns an empty listfixed in fa834e7
codeant: the rollback's Deployment is checked by its id alonenot changed: the helper returns it whole and the cut test compares commit and Release against the cut fixture's list; that it is of the Release the rules plan the rollback to is the rollback's own test's to hold, in the third pull request, with its mutant

šŸ¤– Generated with Claude Code