[dinwwwh]

Requested by dinwwwh · project thread

What

Test-only. Stryker on main (94c60c7) leaves 77 of 2,511 mutants alive, where 0.5.0 left 70. Of the 10 new survivors, 8 are untested branches added by #313, #314 and #318 (the other 2, the converting: [] initial value and isOneWay's final return false, are equivalent):

Mutant that survived	New test
cutHolders.clear() removed (#313)	A $ref object cut at recursion in one call still told no oneOf branch apart in the next call.
follow's check for a $ref into a target being inlined removed (#313's surroundings)	A $ref to a component whose own $ref into that target is cut only after the component's other keywords convert. Without the check, the oneOf stays although its branch { allOf: [{}] } now overlaps the other one, so 3.0 rejects { tag: "u" }, which 3.1 accepts.
isOneWay ignoring writeOnly (2 mutants), or not following a $ref (#318)	A oneOf tagged only by writeOnly tags, and one tagged by $refs to readOnly consts, both become anyOf.
repeats()'s additionalProperties clause (3 mutants, #314)	A repeated additionalProperties beside the same (no) properties merges, and a repeated description beside different properties merges too.
mayRepeatUnseen counting null as an object	A oneOf of literals where one enum holds null merges.

I checked each mutant by hand: each now fails a test, and the unmutated code passes.

Checks

pnpm run check and pnpm test (303 tests) pass. No source change, so coverage stays at 100%.

---

Generated by Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[codspeed]

Merging this PR will not alter performance

1 warning

⚠️ 1 benchmark is dominated by system calls

System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.

Learn more about system calls

See affected benchmarks

✅ 29 untouched benchmarks
⏩ 2 skipped benchmarks¹

---

_{Comparing claude/project-thread-rixcs1 (c848f6e) with main (f5bf5c9)}

Footnotes

1. 2 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[pullfrog]

✅ No new issues found.

Reviewed changes — test-only PR adding coverage in packages/downgrader/tests/v3.1-to-v3.0.test.ts for ten mutants Stryker leaves alive on main (eight from #313/#314/#318 plus the two equivalent ones). No src/ change.

• cutHolders.clear() reset — a $ref object cut at recursion in one downgradeSchemaV31ToV30 call must not suppress branch-telling in the next call.
• follow's inlining cut — a $ref to a component whose own $ref back into the target is cut only after its other keywords convert still forces the oneOf to anyOf.
• isOneWay writeOnly / $ref follow — a oneOf tagged only by writeOnly, and one tagged via $refs to readOnly consts, each become anyOf; a $ref loop holds neither keyword.
• repeats() additionalProperties clause — a repeated additionalProperties beside the same (no) properties merges; a repeated description beside differing properties merges too.
• mayRepeatUnseen null handling — a oneOf of literals whose enum holds null merges.

I verified the coverage claim independently: applying each of nine targeted mutations to src/v3.1-to-v3.0.ts and re-running the corresponding test kills all nine (source restored byte-identical). Full suite (362 tests) and uncheck (oxlint/oxfmt/tsc) pass on the tree.

  ^{｜ View workflow run ｜ Using deepseek-v4.1-flash (free via Pullfrog for OSS) ｜ 𝕏}