[cknitt]

A PR run takes about 12.5 minutes (median of 12 recent runs). The run waits for the slowest build-compiler leg: ARM 557 s, Windows 550 s, macOS 459 s. This PR shortens all three.

• Website build only for new API docs. The website build (about 3.5 min on ARM) ran on every PR. The "docs changed" check always found a change: the copied .gitkeep showed up as untracked, and the website's v13 branch has older docs than master. Now an empty cache entry, keyed on the website commit and the hash of scripts/res/apiDocs, records each successful build. A PR whose docs match an earlier build skips the website checkout, Node setup and build. The first PR after the website branch moves still builds once.
• Format check only on Linux. It gives the same result everywhere, but took 25 s on Windows and 18 s on macOS.
• Faster polling in the canonical rewatch tests. The canonical suite is the slowest part of the rewatch step on every platform (Windows 224 s, macOS 221 s, ARM 120 s). Its ~150 waits polled once a second. They now poll every 0.1 s, with the timeouts still in seconds. Locally, watch tests went from 6.9 s to 3.7 s, 4.1 s to 0.5 s, and 16.3 s to 10.1 s.

Together this should bring a PR run to about 10.5 minutes. The real numbers will show up in this PR's CI.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Kbf3gG8vKzkJGrU7AZ1CXQ

[github-actions]

Changelog

No changelog entry.

Semantic Diff

3 commits · Full page

Build the website in CI only for API docs it hasn't built yet

_{adc28e5 · cknitt + Claude · Nov 14, 2026, 09:16 UTC}

The website build ran on every pull request: the check for changed docs always found the copied .gitkeep and the website branch's older docs. An empty cache entry keyed on the website commit and the docs hash now records each build, so the ARM job skips about 3.5 minutes when the docs match an earlier build.

Check formatting only on Linux in CI

_{adc3052 · cknitt + Claude · Nov 14, 2026, 09:16 UTC}

The format check gives the same result on every platform but took 25 s on Windows and 18 s on macOS, both on the critical path.

Poll every 0.1 s in the build system's canonical tests

_{4f18f1e · cknitt + Claude · Nov 14, 2026, 09:16 UTC}

The canonical suite is the slowest part of each platform's rewatch step, and its waits polled once a second, so each of its ~150 waits took up to a second too long. Timeouts stay in seconds. Locally, the passing watch tests took 3.7 s instead of 6.9 s, 0.5 s instead of 4.1 s and so on.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.06%. Comparing base (1efa212) to head (4f18f1e).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #9302      +/-   ##
==========================================
- Coverage   82.06%   82.06%   -0.01%     
==========================================
  Files         455      455              
  Lines       65399    65399              
==========================================
- Hits        53672    53670       -2     
- Misses      11727    11729       +2     

see 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[github-actions]

✅ Compiler benchmarks

Allocated words

In total: 640,952,947 → 640,952,815 (±0.00%).

Generated projects: 920,897,791 → 920,897,145 (±0.00%).

None of the 265 benchmarks changed by 0.1% or more.

Instructions

In total: 9,209,713,740 → 9,209,713,740 (±0.00%).

None of the 33 benchmarks changed by 0.1% or more.

Output size

In total: 25,843,406 → 25,843,406 (±0.00%).

None of the 13 benchmarks changed by 0.1% or more.

Pull request 4f18f1e compared with base 1efa212 in this run.

[pkg-pr-new]

Open in StackBlitz

rescript

npm i https://pkg.pr.new/rescript@9302

@rescript/belt

npm i https://pkg.pr.new/@rescript/belt@9302

@rescript/darwin-arm64

npm i https://pkg.pr.new/@rescript/darwin-arm64@9302

@rescript/linux-arm64

npm i https://pkg.pr.new/@rescript/linux-arm64@9302

@rescript/linux-x64

npm i https://pkg.pr.new/@rescript/linux-x64@9302

@rescript/runtime

npm i https://pkg.pr.new/@rescript/runtime@9302

@rescript/win32-x64

npm i https://pkg.pr.new/@rescript/win32-x64@9302

commit: 4f18f1e

[github-actions]

Developer playground preview: https://rescript-lang.github.io/rescript/dev-playground/?version=pr-9302

[cknitt]

@codex review

[chatgpt-codex-connector]

Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-14T10:06:56.305027Z	4f18f1e	Manual request

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Keep it up!

Reviewed commit: 4f18f1eec4

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".