PPLT-6073: raise the PDF regression tolerance and close the PDF coverage gaps

[RaghavsBrowserStack]

Fixes the two CI failures seen on the 1.32.10 release PR (#2430). Test-only — no product code changes.

Regression (Track P)

jack sparrow resume.pdf diffed by 648 pixels against a 600 budget, in box [369, 1447, 692, 1448] at a channel delta of 4.

That box is not the circular photo clip the 600 was measured against — it is the hairline rule under the LANGUAGES heading, whose two pixel rows are the only non-white rows across those 324 columns. A second, previously unmeasured source of run-to-run jitter on a fixed Chromium build (the manifest browser and the run browser matched, so the goldens are not stale).

Budget raised to 2000 — roughly 2x the 918-pixel worst case of both sources landing at once, and still 0.1% of the page. A real rendering regression moves glyphs or layout and blows straight past it, and a size change is never tolerated at all. The second source is now documented alongside the first.

@percy/core coverage

All 1322 specs passed; the job failed nyc's 100% threshold on two uncovered spots.

• pdf-rasterize.js:128 — the asInputError rethrow when a rendered page comes back under Percy's 10px floor. fitScale only clamps the upper bound, so a 4x4pt MediaBox rasterizes to 8x8 at scale 2 and trips it. The new test also pins that this surfaces as a 400, not a 500.
• pdf-snapshot.js:67-68 — the decoded-size check. The existing oversize test trips the encoded cap first, leaving the decoded one unreachable. The cap rounds up to a whole base64 quantum, so 52428801 bytes encodes to exactly MAX_PDF_BASE64_CHARS and clears it — the two-byte window the second check exists to close.

Verification

Both new specs pass locally. Because nyc returns no data for a spec-filtered run, coverage was confirmed by mutation instead — each target line was broken in turn and the corresponding spec failed:

Mutation	Result
pdf-rasterize.js:128 — throw asInputError(error) → throw error	Expected 500 to be 400
pdf-snapshot.js:67 — if (buffer.length > MAX_PDF_BYTES) → if (false && …)	fell through to the %PDF- magic check

Regression has already passed green with this change.

Merge order

CI runs on the merge result (refs/pull/N/merge), so once this lands on master, re-running #2430 will pick the fix up automatically — no change needed on the release branch.

🤖 Generated with Claude Code

[coderabbitai]

Warning

Review limit reached

Next included review available in 42 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

This review ran on the open-source allowance, not this organization's plan, because the pull request author doesn't have an assigned seat. Waiting won't change this — ask an organization admin to assign them a seat, or add seats in Billing if every seat is already assigned, then retry.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

Configuration used: Workspace UI

Review profile: CHILL

Plan: Advanced

Run ID: 400bab14-c326-48b6-bde7-a73accbf0094

📥 Commits

Reviewing files that changed from the base of the PR and between b90cd9f and 96d0cc6.

📒 Files selected for processing (2)

• packages/core/test/pdf-snapshot.test.js
• test/regression/pdf-render.test.js

---

_{Comment @coderabbitai help to get the list of available commands.}