[miguel-heygen]

What

chunkFrames.injectorBudget.test.ts no longer leaves a no-op video frame injector behind for the tests that run after it in the same bun process.

Why

The test replaced the whole @hyperframes/engine module with mock.module, swapping createVideoFrameInjector for async () => {}, and then imported chunkFrames.js under that mock. Its afterAll put the engine module back, but chunkFrames.js was already loaded and kept the no-op. Run in one process with bun test src/services/distributed, every later chunk render drew no video frames:

• a locally read video stayed on its first frame;
• a video read by URL did not show at all.

So renderChunk's "a video the chunk does not show" test failed with two different chunk hashes. CI runs each bun test file in its own process (scripts/run-test-lane.mjs), so CI never hit this, and the renderer itself is unchanged in behaviour. Only a hand-run of the folder did.

How

chunkFrames.ts builds injectors through a small mutable chunkFrameInjectorDependencies object, the same seam renderChunk.ts already has (distributedCaptureSessionDependencies). The budget test swaps it inside try/finally, so the real injector is back before the next test. No module mock.

Tests

• New in chunkFrames.injectorBudget.test.ts: "leaves the engine's own injector in place for every later test". It runs after the budget test and checks that the factory builds the engine's real injector, which returns null for an empty lookup.

fails without the fix: on main the new test gets [AsyncFunction] (the leftover no-op) instead of null. Running chunkFrames.injectorBudget.test.ts and renderChunk.test.ts in one process fails renderChunk's "a video the chunk does not show" 3 of 3 on main; with this change it passes 2 of 2.

Break checks:

• Drop the finally restore and the new test fails.
• Make the factory call createVideoFrameInjector directly and the budget test fails.

Run on Linux:

• The budget test, 3 runs, green.
• chunkFrames.test.ts and renderChunk.test.ts, each on their own.
• The whole services/distributed folder in one process: renderChunk now passes. The three files there that use vitest's vi.mock (plan.masterBus, renderChunk.stall, renderChunk.verifyInWorker) still report load errors under plain bun test, the same as on main; CI runs them under vitest.
• bun run typecheck and the comment ratchet.

Size

About 30 lines, on its own: a test-isolation fix in one file and its seam. No open PR in this repo touches these files.

[github-actions]

Edit accuracy: accurate 267 (base branch 267), smooth 109 of those

The gate passes.
Smoothness is reported in the artifact, not gated. A case fails only if it fails 2 of 3 runs.

Quarantined, measured but not gated (8)

• seqplainkeys-none-center-r0-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
• seqplainkeys-none-center-r0-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): fail / pass / pass, passes
• seqplainkeys-none-center-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
• seqplainkeys-none-pct-r0-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): pass, passes
• seqplainkeys-none-pct-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
• seqplainkeys-none-px-r0-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): fail / pass / pass, passes
• seqplainkeys-none-px-r30-nested-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run
• seqplainkeys-none-px-r30-root-z100 (fixed by studio: undo/redo on a keyframed plain element sometimes leaves the box 230-400px off (edit-accuracy seqplainkeys-none-*) #6989): not run

[heygengenesis]

Verdict: APPROVE · 0 blocker · 0 important · 0 nits · head 3939a725

Scope

• Read both files. The seam has the same shape as distributedCaptureSessionDependencies. The try/finally restore puts the real injector back before the next test.
• The new test asserts null for a null lookup. That holds: the real createVideoFrameInjector returns null first thing (videoFrameInjector.ts:180). On main, the module mock is only undone in afterAll, so the test gets the no-op and fails. The "fails without the fix" line holds.
• renderChunk.verifyInWorker.test.ts uses vi.mock on engine. It spreads the original module and leaves createVideoFrameInjector alone. vi.mock is hoisted, so the object-captured binding sees the same value as before.
• CI at this head: CI, Comments, CodeQL, PR captures and Player perf pass. regression and Windows render were still running when I posted.
• No prior reviews.

— genesis-v2