fix(ci): pin chrome-headless-shell + clamp PSNR checkpoint to a valid frame

Two narrow fixes to keep the regression suite green and reproducible.
Stale baselines from the sub-composition refactor (PR #918) are being
regenerated separately in PR #925; this PR is just the structural
fixes that PR can't make on its own.

1. **Pin `chrome-headless-shell` in `Dockerfile.test`** to
   `148.0.7778.167` instead of `@stable`. `@stable` is a moving tag;
   every Chrome stable promotion shifts pixel output enough to fail
   PSNR on the golden baselines, so the regression suite silently
   broke whenever Docker.test rebuilt against a freshly-promoted
   stable. Pinning to the version `@stable` currently resolves to
   (matching what main's regenerated baselines were captured under)
   makes Chrome bumps an explicit, batched-with-baseline-regen
   action. Comment on the `RUN` line spells out the bump procedure.

2. **Clamp the last PSNR checkpoint to a frame the video stream
   actually contains.** `runTestSuite` samples 100 checkpoints across
   `min(rendered, snapshot)` container duration. Container duration
   includes audio padding past the last video frame — many-cuts is
   5.654s container vs 5.6s of video at 30fps = 168 frames. At i=99
   the raw container duration mapped to time 5.59746s → frame index
   168 (round(5.59746 × 30)), one past the last frame the stream
   contains. ffmpeg's `psnr` filter emits no `average:` line for a
   non-existent frame, so the harness crashed with `Unable to parse
   PSNR output at 5.59746s` — pre-existing on plain `origin/main`,
   which PR #918 admin-merged through on shard-2. Miguel's regen via
   `--update` didn't catch it because `--update` only writes the
   snapshot; it doesn't validate. Subtracting one frame interval
   from the sampling duration guarantees the last checkpoint always
   lands on a real frame.

Verified locally inside `Dockerfile.test`:

  bun run --cwd packages/producer docker:build:test
  bun run --cwd packages/producer docker:test many-cuts   #  green
  bun run --cwd packages/producer docker:test style-3-prod \
    style-5-prod sub-composition-video                    #  green
This commit is contained in:
James
2026-05-17 22:41:55 +00:00
parent 6c533c0b0f
commit 070d40fecb
2 changed files with 16 additions and 8 deletions
+6 -1
View File
@@ -53,7 +53,12 @@ ENV CONTAINER=true
# Install chrome-headless-shell for deterministic BeginFrame rendering.
# This lightweight Chrome binary supports HeadlessExperimental.beginFrame.
# Install to ~/.cache/puppeteer/ where resolveHeadlessShellPath() looks.
RUN npx --yes @puppeteer/browsers install chrome-headless-shell@stable \
#
# Pinned to a specific build (NOT @stable) so the regression-test golden
# baselines in packages/producer/tests/*/output/output.mp4 stay reproducible.
# Each Chrome stable bump shifts pixel output enough to fail PSNR. Bump this
# version together with regenerating baselines via `docker:test:update`.
RUN npx --yes @puppeteer/browsers install chrome-headless-shell@148.0.7778.167 \
--path /root/.cache/puppeteer \
&& find /root/.cache/puppeteer/chrome-headless-shell -name "chrome-headless-shell" -type f \
&& echo "chrome-headless-shell installed"
+10 -7
View File
@@ -1079,16 +1079,19 @@ async function runTestSuite(
videoMetadata.durationSeconds,
snapshotMetadata.durationSeconds,
);
const fps = fpsToNumber(suite.meta.renderConfig.fps);
// Container duration includes audio padding past the last video frame
// (e.g. many-cuts: 5.654s container vs 5.6s of video). At i=99 the
// raw container duration maps to a frame index past nb_frames, and
// ffmpeg's PSNR filter emits no `average:` line for a non-existent
// frame. Subtract one frame interval so the last checkpoint always
// lands on a frame the video stream actually contains.
const sampleDuration = Math.max(0, videoDuration - 1 / fps);
const minPsnrForMode = resolveMinPsnrForMode(options.mode, suite.meta.minPsnr);
for (let i = 0; i < 100; i++) {
const time = (videoDuration * i) / 100;
const psnr = psnrAtCheckpoint(
renderedOutputPath,
snapshotVideoPath,
time,
fpsToNumber(suite.meta.renderConfig.fps),
);
const time = (sampleDuration * i) / 100;
const psnr = psnrAtCheckpoint(renderedOutputPath, snapshotVideoPath, time, fps);
visualCheckpoints.push({
time,
psnr,