test(apple,vega): give two runner/Vega probes assertions that can lose

[thymikee]

Summary

Two tests passed for every production bug of the kind they claim to catch. Both now lose for a
specific one, and I ran the mutant against both the old and the new body.

Test	Why it could not fail	Mutant planted	Old body	New body
runner-io.test.ts — "a tail whose file disappears stops on its own instead of throwing"	onOutput: () => {} threw every chunk away; the body had zero assertions, so the "stops on its own" half of its own name was never observed	drain()'s catch { stop(); } → swallow and keep polling (runner-io.ts:211), i.e. precisely the give-up-then-resume bug the in-code comment forbids	8 passed	1 failed | 7 passed
inventory.test.ts — "Vega inventory … accepts only its virtual device"	expected value was parseVegaDeviceList(...), a production export — the parser was its own oracle, so both sides moved together	kind: 'emulator' → 'physical' in parseVegaDeviceList (inventory.ts:26)	2 passed	1 failed | 1 passed

parseVegaDeviceList is the more interesting half: that oracle call was the function's only
reference outside inventory.ts, so its kind, booted, target, id and display-name format
were pinned nowhere in the repo. The literal expectation is now that parser's first real test, and it
still carries the discovery claim — the mocked device list output contains a physical : tv row
which must not reach the result.

Dropped after a second read, rather than shipped as noise: runner-sequence.test.ts's unasserted
validateRunnerSequenceSteps(...) accept-paths (the cap case and the low-durationMs case). An
unasserted call still fails when it throws, and cap + 1 rejection is pinned two tests above the cap
case, so the boundary is genuinely closed. Wrapping them in assert.doesNotThrow would have been
cosmetic.

Validation

Tested at 3d11a8ff.

• --project apple-runner packages/platform-apple/src/runner/__tests__/runner-io.test.ts → 8 passed
• --project unit-core packages/platform-vega/src/inventory.test.ts → 2 passed
• pnpm check:affected --run — all runnable checks passed
• oxlint . --deny-warnings, repo-wide oxfmt, tsc -b packages/platform-vega packages/platform-apple
• Every mutant reverted; git status clean of production changes before committing

Touches no production file, so no device lane is selected; check:production-exports reports no new
finding for the now test-unreferenced parseVegaDeviceList, which stays exported.

[github-actions]

Size Report

Metric	Base	Current	Diff
Installed (including dependencies)	4.81 MB	4.81 MB	0 B
Package (unpacked)	4.81 MB	4.81 MB	0 B
Package (download)	1.44 MB	1.44 MB	+9 B

Startup median (7 runs, lower is better):

Scenario	Base	Current	Diff
CLI --version	26.4 ms	26.5 ms	+0.0 ms
CLI --help	79.7 ms	79.0 ms	-0.8 ms

[thymikee]

Reviewed at 3d11a8f. This is ready for human review. Checks are green (17/17). Not blocking: the 3-line comment above the literal in https://github.com/callstack/agent-device/blob/3d11a8f/packages/platform-vega/src/inventory.test.ts#L30 explains why the test no longer compares against parseVegaDeviceList(...), but that's history the PR description can carry, not a constraint the literal needs to state, so feel free to drop it or keep it.

I traced the mutation claims statically against runner-io.ts drain()/schedule()/stop() and against parseVegaDeviceList in inventory.ts rather than running the mutated builds, and I only read the PR body's claim about runner-sequence.test.ts (dropped assertions on the cap/low-durationMs accept paths) for context, since it's outside this diff.

[github-actions]

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-09-25 05:43 UTC