9e5c00d716
The MUTATION_GATE_STRYKER_REGRESSION.md note shipped in #4134 was WRONG: there is no Stryker 9.6.1 OptionsValidator regression. The false "concurrency must match pattern" failure was operator error — `stryker -c <file>` sets --concurrency, not the config file (which is the positional arg), so concurrency became the string "stryker.conf.json". `npx stryker run` (auto-discovery) validates and runs fine: it instruments 15488 mutants across the 8 modules. Renamed -> MUTATION_GATE_SPIKE_VERDICT.md and rewritten with: - the corrected record + the `-c` CLI gotcha; - the real Task 12 per-test spike verdict (GO, Plan A): a scoped run on open-sse/utils/error.ts killed 86/422 mutants, ALL 86 with a populated `killedBy` resolving to a test file -> per-test attribution works (per-file granularity, as expected for the tap-runner). Onda 2 can use killedBy directly. - the standing Onda-2 blocker is budget (god-files), not tooling.
2.9 KiB
title
| title |
|---|
| Mutation gate — Task 12 per-test spike verdict (GO) + Stryker CLI gotcha |
Mutation gate works; Task 12 per-test attribution is GO
Supersedes an earlier draft of this file that wrongly reported a "Stryker 9.6.1 OptionsValidator regression". There is no regression — the mutation gate runs fine. The earlier failure was operator error (see the gotcha below). This note corrects the record and captures the real spike verdict.
Task 12 spike verdict — GO (Plan A, killedBy)
Ran a scoped Stryker mutation on open-sse/utils/error.ts (2 covering test files) on the
v3.8.29 toolchain (Stryker 9.6.1 + ajv 8.18.0, coverageAnalysis: perTest, tap-runner):
- 422 mutants — 86 Killed, 64 Survived, 272 NoCoverage.
- All 86 killed mutants carry a populated
killedBy(86/86), and the report'stestFilessection resolves eachkilledByid to a covering test file. - So per-test (Plan A) attribution works: Onda 2 radiography can read
killedBydirectly. Granularity is per test FILE (the tap-runner spawns one node process per test file) — exactly the unit the redundancy heuristic needs ("a test file that kills no mutant another file does not already kill").
The full config also instruments cleanly: npx stryker run --dryRunOnly reports
"Found 8 of 5632 file(s) to be mutated. Instrumented 8 source file(s) with 15488 mutant(s)"
— the 8-module nightly is healthy. This matches (and refreshes, on the current deps) the GO
recorded back in v3.8.27.
⚠️ Gotcha that caused the false alarm: stryker -c is --concurrency, NOT --config-file
npx stryker run -c stryker.conf.json sets concurrency = "stryker.conf.json" (a string),
which fails schema validation with a mis-attributed error:
Config option "concurrency" must match pattern "^(100|[1-9]?[0-9])%$". Stryker still loads
the real config via auto-discovery, so the run looks like the config is broken when it is
not. jsPropertySyntax (an ajv-6 option, a no-op under ajv 8) makes the validator's error
path unhelpful, but the data is the giveaway: options.concurrency === "stryker.conf.json".
Correct invocations:
- Default config (
stryker.conf.jsonin cwd):npx stryker run(no flag). - Explicit config file: pass it as the positional arg —
npx stryker run my.conf.json. -c/--concurrencytakes a number or percentage (-c 4,-c 50%).
Standing Onda-2 blocker (unchanged): budget, not tooling
The real blocker for a complete nightly remains runtime budget — the 8-module run is
15488 mutants and the god-files (chatCore, combo) dominate ~⅔; the conc=4 nightly timed
out at 180min (run 27705123780). Levers: exclude god-files until the Onda-3 splits, seed
the incremental file, batch per night, or accept partial. The per-test spike does not change
that — it only confirms the attribution mechanism Onda 2 will rely on.