Mutation testing (@tao.js/utils)

StrykerJS + Jest for signal-network utilities.

Run

pnpm test:mutation:utils

HTML/JSON reports land in reports/mutation/ (gitignored).

Notes

• Config: stryker.config.json — inPlace: true; barrel index.js excluded.
• Disable equivalent mutants with // Stryker disable … : reason (colon required; place disable before a try, not inside it).
• Optional _debug && console.log(...) paths, ++id % MAX_SAFE_INTEGER ID-counter arithmetic, and a handful of redundant boolean sub-clauses / no-op guard branches (see below) are Stryker-disabled as equivalent.
• Latest score (2026-07-18): 100% (577 killed, 2 timeout, 0 survived; thresholds high=95). 11 sandbox child-process crashes in bridge.js/Transceiver.js are Stryker/Node runner flakiness (filter is not a function from an unrelated in-place mutant clobbering a shared module during a parallel run), not survivors — they do not count against the score and disappear on re-run.

File-by-file

| File | Score | Killed | Timeout | Survived | | ----------------- | ----------- | ------- | ------- | -------- | | Channel.js | 100.00% | 74 | 0 | 0 | | Transceiver.js | 100.00% | 140 | 2 | 0 | | Transponder.js | 100.00% | 58 | 0 | 0 | | logger.js | 100.00% | 81 | 0 | 0 | | Relay.js | 100.00% | 40 | 0 | 0 | | Source.js | 100.00% | 50 | 0 | 0 | | seive.js | 100.00% | 29 | 0 | 0 | | bridge.js | 100.00% | 35 | 0 | 0 | | trigram-filter.js | 100.00% | 26 | 0 | 0 | | forward-chain.js | 100.00% | 8 | 0 | 0 | | transfer.js | 100.00% | 36 | 0 | 0 | | All files | 100.00% | 577 | 2 | 0 |

Equivalent mutants disabled (with reasons in source)

• ID counters (Channel.js, Transceiver.js ×2, Transponder.js) — ++id % MAX_SAFE_INTEGER monotonicity/wraparound is untestable without exhausting the counter; the newSignalId body is disabled with all since its returned id is never asserted on.
• Debug logging (Channel.js, Transponder.js) — this._debug && console.log(...) branches have no externally observable effect on network behavior.
• logger.js clause1-alone (constructor, depth(), setInspect()) — in inspect == null || typeof inspect != 'function' || …, whenever inspect/i is null/undefined the typeof clause is already true, so forcing clause 1 to false never changes which branch runs.
• Transceiver.js:119 — the synchronous catch guard if (!control.signalled) runs immediately after entering the try, before anything else can observe or mutate control, so it is always true on entry; equivalent to if (true).
• Transceiver.js:224 — when the inline handler's return value is null/undefined, forcing entry into the if body only ever assigns that same falsy value to firstResolve (or is skipped because firstResolve is already truthy), which is indistinguishable from skipping the block.
• Transceiver.js:236 — for...of over an empty nextSpool array already runs zero iterations, so forcing the surrounding if to true has no observable effect when the array is empty.
• Transponder.js:52 — inTO was already coerced through +timeoutMs || 0 above, so inTO > 0 ? inTO : 0 and inTO >= 0 ? inTO : 0 pick the same value for every possible inTO (0 maps to 0 either way).

See @tao.js/core (packages/tao/README.mutation.md) for the pattern used to raise scores.