ci(tests): run the tests on the maintainers' PRs, per platform label, with one results comment by mihailt · Pull Request #969 · wonderwhy-er/DesktopCommanderMCP

mihailt · 2026-10-30T09:31:49Z

The tests now run in CI for the maintainers' PRs, on each platform the PR is labelled with, and one PR comment shows each platform's results and coverage. This builds on #966, the test runner.

What this adds

Problem	Fix
No CI runs the tests on a PR; only a release runs them, on Linux.	`.github/workflows/tests.yml` runs them on each labelled platform: `linux`, `windows`, `macos`.
A maintainer can't run one platform or one commit on demand.	A `/test windows` comment, or `gh workflow run tests.yml -f commit=… -f platforms=…`, runs it.
Timed tests can't run on a shared machine.	`-f tests=timed` runs each timed test in a job of its own, per platform.
Results are only in the run's log.	`test/ci-comment.js` keeps one PR comment: a row per platform, its coverage and report links.

Only the logins in MAINTAINERS (test/helpers/ci-platforms.js) start a run.
A run from a comment or on demand checks out the commit by SHA, with a read-only token and no secret.
/test and gh workflow run work only once the workflow is on main.
CI runs Node 22 with npm 11; Linux sets kernel.apparmor_restrict_unprivileged_userns=0, so the PDF tests run Chrome with its own sandbox.
On Windows with Node 22, 3 thread-pool repro scripts fail until fix(windows): start Node 22 with a 16-thread libuv pool #968 is in.
Stack 2 of 3, on test(runner): run the tests through node:test, with coverage and standard reports #966 (framework/1-test-runner); framework/3-ai-review builds on it.

Where to look

.github/workflows/tests.yml: the triggers, the platforms job, and the unit job per platform.
test/helpers/ci-platforms.js: MAINTAINERS, and which platforms and tests each event runs.
test/ci-comment.js: the PR comment, built from each platform's results.
test/helpers/github-api.js, test/helpers/workflow-file.js: the GitHub calls, and the workflow lines the tests read.
test/README.md: its CI section.

How to verify

npm run build && node test/run-all-tests.js test-ci-platforms.js test-ci-comment.js
gh workflow run tests.yml -f commit=<sha> -f platforms=linux   # once the workflow is on main

Tests and results

test-ci-platforms.js: which events start a run, for whom, on which platforms, and the workflow's own lines.
test-ci-comment.js: the comment's tables, from made-up results of each platform.
Most of this PR's changes came test first: the commit before each added its test, which failed there.
At 9db23303, these tests and the coverage check are being run on Windows, macOS and Linux.

🤖 Generated with Claude Code

coderabbitai · 2026-10-30T09:31:57Z

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

github-actions · 2026-10-30T09:46:58Z

Tests and coverage for `c4a7727`

Platform	Tests	Ran	Minutes	Commit	Run
Linux	✅ 1 of 1 files passed	`files:test-interact-output-cleaning.js`	1	`c4a7727`	run
Windows	✅ 226 of 226 files passed; Docker: GitHub's Windows runner has no Docker for Linux containers; 9 checks skipped	`everything`	30	`a383442`	run
macOS	❌ no results	`everything`	15	`a383442`	run

Platform	Lines	Branches	Functions	Report (HTML, lcov, JSON, Cobertura)	Results (JUnit, JSON, TAP)
Linux	19.05 %	49.82 %	15.66 %	coverage-ubuntu-latest	results-ubuntu-latest
Windows	66.46 %	79.74 %	81.05 %	coverage-windows-latest	results-windows-latest
macOS	–	–	–	–	–
Combined	–	–	–	needs Linux, Windows and macOS in one run	–

Lines by area

Area	Linux	Windows
`install scripts`	0 % (0/2074)	0 % (0/2074)
`src/*`	35 % (1990/5681)	90 % (5115/5681)
`src/handlers`	28.8 % (299/1038)	82.9 % (860/1038)
`src/npm-scripts`	0 % (0/227)	24.2 % (55/227)
`src/remote-device`	0 % (0/5656)	89.1 % (5038/5656)
`src/tools`	30.9 % (1666/5386)	79.4 % (4279/5386)
`src/ui`	2.1 % (168/8127)	35.2 % (2863/8127)
`src/utils`	36.7 % (2656/7229)	73.9 % (5341/7229)
`vendor/telemetry-contract`	0 % (0/164)	59.1 % (97/164)
total	19.1 % (6779/35582)	66.5 % (23648/35582)

…el, with one PR comment .github/workflows/tests.yml runs npm run test:coverage on Node 22 with npm 11, for PRs opened by the maintainers in MAINTAINERS (test/helpers/ci-platforms.js): - the labels linux, windows and macos pick the platforms; a push runs every labelled one, adding a label runs its platform at once; - a maintainer's /test linux windows comment runs those platforms on the PR's head, and gh workflow run tests.yml runs any commit on demand, with a tests value (empty for everything, or test, integration, repro, docker, timed, changed:A..B, files:a.js,b.js); tests=timed runs one job per timed test; - one PR comment, with no prose: a results row and a coverage row per platform, and the lines by area. It fails closed: a job that failed, an unreadable results.json, or no test run is never a pass. A comment's or on-demand run tests unmerged code with a read-only token, no secret and no npm cache; the comment job runs main's script. test/README.md says how. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…g first) The repro's first install took 522 s on GitHub's Windows from its own empty npm cache. test-ci-platforms.js checks a step after npm ci, before the tests, names the job's npm cache (npm config get cache) as REPRO_NPM_CACHE. Fails until the workflow does so. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

… CI files name no later step - A step after npm ci names the job's npm cache (npm config get cache) as REPRO_NPM_CACHE, so the release-packages repro's installs read from it instead of an empty cache: its first install took 522 s on GitHub's Windows, and once ran the repro into its 10-minute guard. - tests.yml, test/README.md and helpers/ci-platforms.js named the AI review, which the next pull request adds; startedBy's on-demand branch is reached only by the review. check-coverage-report.js's header names the CI scripts this PR adds to TEST_TOOLING. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…g first) GitHub sends opened, then a labeled event for each label on a new PR. With opened in tests.yml's types, opened ran every labelled platform and each labeled event ran its platform again: each platform twice (#233, 30 Oct: 7 runs at once, Windows twice). test-ci-platforms.js replays those events against the workflow's own types and counts each platform's runs; the review label's event must pick no platform. Fails until opened starts nothing. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

GitHub sends opened, then a labeled event for each label a new PR has. With opened among tests.yml's pull_request types, opening a PR ran every labelled platform, and each labeled event ran its platform again: each platform twice (#233: 7 runs at once, Windows twice). opened is out of the types, and platformsFor no longer treats it as a push; a push and a reopen still run every labelled platform, and the review label's event picks no platform. test/README.md says so. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

… line and its job (failing first) "207 of 211 files passed" on Windows said nothing of which 4. The comment gets a table of the failed files per platform: each linked to the job it ran in, with its first error line (a ❌ or ✗ line, else one naming an Error, else its last line; never a ✓ line that mentions an error), at most 20 per platform and the rest counted. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…t error line and its job ci-comment.js keeps, for each failed file, unreadable results part and stopped job, its first error line (firstError: a ❌ or ✗ line, else one naming an Error, never a ✓ line, else the last line) and its job's link (the job its results artifact came from). renderComment lists them per platform under the results, at most 20 each and the rest counted. The error shows in a code span, so only backticks, pipes and line breaks are taken out ("BLOCKED >30000ms" keeps its >); the test now says so. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…hows the three platforms combined (failing first) - test-ci-platforms.js: tests.yml's on-demand tests_from input is checked as a SHA and resolved like commit, through the environment only. The test job takes test/ from it before the install. A cross run writes no PR comment, and the heading names both commits. - The same test: the comment gets a Combined row under the platforms and a Combined column by area. Only one run's own three coverage-final.json files make it. With fewer, or one that can't be read, the row has no figure and says why. - test-ci-comment.js: TESTS_FROM names the tests' commit in the heading, and the three platforms' reports give the Combined figure. Fails until tests.yml, ci-comment.js and ci-platforms.js do so. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…ined row and column - tests.yml's on-demand tests_from runs that commit's test/ on commit's code, as test/cross-test.js does locally. It's checked as a SHA and resolved like commit. The test job takes test/ from it before anything installs. Its rows go into the job summary only, never a PR's comment. - ci-comment.js combines this run's three coverage-final.json files (test/combined-coverage.js). renderComment shows the figure under the platforms and as a column by area. Without the three, or with one that can't be read, the row has no figure and says why. mergeComment never carries an earlier run's figure over. - test/README.md says how. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

mihailt added linux CI: run the tests on Linux windows CI: run the tests on Windows macos CI: run the tests on macOS labels Oct 30, 2026

mihailt force-pushed the framework/2-ci branch from 9db2330 to 9d71b39 Compare October 30, 2026 10:21

mihailt added linux CI: run the tests on Linux and removed linux CI: run the tests on Linux labels Oct 30, 2026

mihailt changed the title ~~ci: run the tests on the maintainers' PRs, per platform label, with one results comment~~ ci(tests): run the tests on the maintainers' PRs, per platform label, with one results comment Oct 30, 2026

mihailt force-pushed the framework/2-ci branch from e9678ae to 8b93272 Compare October 30, 2026 21:49

This was referenced Nov 2, 2026

test(coverage): the three platforms combined in the PR comment, and type-only files out of the count #973

Closed

test(runner): run one commit's tests on another commit's code, with coverage #972

Closed

mihailt force-pushed the test/node-test-runner branch from b51284b to e704cff Compare November 2, 2026 06:38

mihailt force-pushed the framework/2-ci branch 3 times, most recently from 5abbf52 to 84fdcb2 Compare November 2, 2026 14:33

mihailt force-pushed the test/node-test-runner branch from d538f4f to cdd5316 Compare November 3, 2026 22:47

mihailt force-pushed the framework/2-ci branch from 84fdcb2 to c8d5bcb Compare November 3, 2026 22:48

mihailt mentioned this pull request Nov 3, 2026

fix(windows): start Node 22 with a 16-thread libuv pool #968

Open

mihailt force-pushed the test/node-test-runner branch from cdd5316 to 8bc554e Compare November 4, 2026 04:12

mihailt force-pushed the framework/2-ci branch from c8d5bcb to ff59a4d Compare November 4, 2026 04:12

mihailt and others added 9 commits November 4, 2026 15:41

mihailt force-pushed the test/node-test-runner branch from 8bc554e to 4ba8b57 Compare November 4, 2026 13:58

mihailt force-pushed the framework/2-ci branch from ff59a4d to e490ec9 Compare November 4, 2026 14:11

mihailt mentioned this pull request Nov 4, 2026

test(runner): run the tests through node:test, with coverage and standard reports #966

Draft

mihailt added 3 commits November 4, 2026 18:09

test(tests): require history for changed commit ranges (failing first)

c73da2f

ci(tests): fetch history for changed commit ranges

a383442

test(process): disable piped Bash input echo without filtering stderr

c4a7727

Navigation Menu

Repository navigation

ci(tests): run the tests on the maintainers' PRs, per platform label, with one results comment - #969

ci(tests): run the tests on the maintainers' PRs, per platform label, with one results comment#969
mihailt wants to merge 12 commits into
test/node-test-runnerfrom
framework/2-ci

mihailt commented Oct 30, 2026

Uh oh!

coderabbitai Bot commented Oct 30, 2026 •

edited

Loading

Draft PR not reviewed

Uh oh!

github-actions Bot commented Oct 30, 2026 •

edited

Loading

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

mihailt commented Oct 30, 2026

What this adds

Where to look

How to verify

Uh oh!

coderabbitai Bot commented Oct 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Draft PR not reviewed

Uh oh!

github-actions Bot commented Oct 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Tests and coverage for c4a7727

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

coderabbitai Bot commented Oct 30, 2026 •

edited

Loading

github-actions Bot commented Oct 30, 2026 •

edited

Loading

Tests and coverage for `c4a7727`