Repository navigation
chore(deps): update runtime dependencies for odiff 4 - #198
Conversation
Update the MCP SDK to 1.32.2, Playwright to 1.65.1, yaml to 2.9.1, odiff-bin to 4.5.0 and pixelmatch to 8.0.0. odiff 4 rejects an empty ignore-region list and fails a pixel diff it cannot write, so pass ignoreRegions only when there are masks and send an unwanted diff image to a scratch file instead of /dev/null. A real odiff test covers both. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub. |
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
Review the following changes in direct dependencies. Learn more about Socket for GitHub.
|
Bugpatrol review: fixture-appNo problem introduced · tested Tested
Not tested
Updated on each push. Tests only what the diff can affect. Does not block the merge. |
Bugpatrol review: dashboardNo problem introduced · tested Tested
Not tested
Updated on each push. Tests only what the diff can affect. Does not block the merge. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a86d599bf1
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| "odiff-bin": "^3.2.1", | ||
| "pixelmatch": "^6.0.0", | ||
| "odiff-bin": "^4.5.0", | ||
| "pixelmatch": "^8.0.0", |
There was a problem hiding this comment.
Keep the cross-check on a shared color metric
When req.threshold is nonzero—packages/agents/src/evaluate.ts passes 0.1—this upgrade makes pixelmatch use OKLab/HyAB while odiff continues using YIQ. The pixelmatch 8 release notes explicitly warn that tuned thresholds and diff counts can change, but diff() still treats a count delta above four as evidence of a bad capture and assumes the pixelmatch fallback produces the same verdict. This therefore creates spurious engine-disagreement reports and host-dependent visual results whenever odiff is unavailable; retain a YIQ-based pixelmatch version or update the cross-check and fallback semantics before upgrading.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Valid. Fixed in the latest commit: pixelmatch is now held at ^7.2.0, the newest release that still uses the YIQ metric odiff uses, so the cross-check and the fallback keep one colour distance. The PR body records why 8 is left.
odiff 4 ships its binary in an optional platform package and throws "could not find the odiff binary" when it is absent. Treat that as an unavailable engine so the comparison falls back to pixelmatch. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Bugpatrol review: cliNo problem introduced · tested Tested
Not tested
Updated on each push. Tests only what the diff can affect. Does not block the merge. |
pixelmatch 8 moves to an OKLab HyAB distance while odiff still uses YIQ, so the cross-check would report engine disagreements on thresholds tuned for YIQ. 7.2.0 is the newest release on the shared metric. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Updates the runtime dependencies before the next release.
@modelcontextprotocol/sdk1.30.1 to 1.32.2playwright1.64.1 to 1.65.1yamlfloor raised to 2.9.1 (already locked there)odiff-bin3.2.1 to 4.5.0pixelmatch6.0.0 to 7.2.0odiff 4 broke two things the unit tests did not cover:
Invalid ignore regions format, so every unmasked diff threw.ignoreRegionsis now passed only when there are masks.Failed to save diff output, so the/dev/nullsink broke every diff withoutdiffOutPath. The unwanted image now goes to a scratch directory that is removed afterwards.A new test runs the real odiff binary with and without a mask.
Left as is:
pngjs,zodandesbuildare current.pixelmatch8 is held back: it switches to an OKLab HyAB colour distance while odiff keeps YIQ, and the cross-check relies on both engines sharing one metric (Codex review comment). Dev tooling majors (typescript7,vitest5,@types/node26) are not runtime dependencies and are out of scope for this release.Local checks:
pnpm check,pnpm knip,pnpm build, the bundle build and--version, and thepackages/difftests pass. Browser tests need Playwright's 1.65 Chromium, so CI runs them.Independent Codex review (kiln) found one issue, fixed in the second commit: odiff 4 moves its binary into an optional platform package, and its "could not find the odiff binary" error did not trigger the pixelmatch fallback. It now does, verified by forcing the platform package lookup to fail.
🤖 Generated with Claude Code