👷 Split unit-bs CI job per browser

[mormubis]

Motivation

The unit-bs job runs all 5 browsers in a single Karma process. If one browser is unavailable on BrowserStack, the whole job fails and you lose the results from the other 4.

Changes

Each browser now runs as a separate CI job using parallel: matrix. The --browser flag on ci-bs.ts filters the Karma config to a single browser. Resource groups are per-browser so they don't serialize against each other. JUnit reports are written to per-browser directories to avoid collisions. Without --browser, all 5 browsers run as before.

Test instructions

BS_BROWSER=edge yarn test:unit:bs        # Only Edge launches
BS_BROWSER=nonexistent yarn test:unit:bs  # Throws with available ids
yarn test:unit:bs                         # All 5 browsers (backwards compat)

Checklist

• Tested locally
• Tested on staging
• Added unit tests for this change.
• Added e2e/integration tests for this change.
• Updated documentation and/or relevant AGENTS.md file

[cit-pr-commenter-54b7da]

Bundles Sizes Evolution

📦 Bundle Name	Base Size	Local Size	𝚫	𝚫%	Status
Rum	169.51 KiB	169.51 KiB	0 B	0.00%	✅
Rum Profiler	5.97 KiB	5.97 KiB	0 B	0.00%	✅
Rum Recorder	21.23 KiB	21.23 KiB	0 B	0.00%	✅
Logs	54.70 KiB	54.70 KiB	0 B	0.00%	✅
Rum Slim	127.85 KiB	127.85 KiB	0 B	0.00%	✅
Worker	22.99 KiB	22.99 KiB	0 B	0.00%	✅

🚀 CPU Performance

Action Name	Base CPU Time (ms)	Local CPU Time (ms)	𝚫%
RUM - add global context	0.0021	0.0029	+38.10%
RUM - add action	0.0105	0.0135	+28.57%
RUM - add error	0.0097	0.0125	+28.87%
RUM - add timing	0.0004	0.0008	+100.00%
RUM - start view	0.0092	0.0144	+56.52%
RUM - start/stop session replay recording	0.0007	0.0012	+71.43%
Logs - log message	0.0139	0.0279	+100.72%

🧠 Memory Performance

Action Name	Base Memory Consumption	Local Memory Consumption	𝚫
RUM - add global context	38.24 KiB	38.44 KiB	+211 B
RUM - add action	64.64 KiB	64.57 KiB	-67 B
RUM - add timing	36.73 KiB	38.47 KiB	+1.74 KiB
RUM - add error	70.01 KiB	70.96 KiB	+981 B
RUM - start/stop session replay recording	45.58 KiB	41.90 KiB	-3.68 KiB
RUM - start view	483.88 KiB	486.22 KiB	+2.34 KiB
Logs - log message	55.00 KiB	55.82 KiB	+847 B

🔗 RealWorld

[datadog-official]

🎉 All green!

❄️ No new flaky tests detected
🧪 All tests passed

🎯 Code Coverage (details)
• Patch Coverage: 100.00%
• Overall Coverage: 76.96% (+0.00%)

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: a311531 | Docs | Datadog PR Page | Give us feedback!}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a311531a23

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".

[chatgpt-codex-connector]

Restore full-run BrowserStack concurrency

When BS_BROWSER is unset (the documented backward-compatible yarn test:unit:bs / ci-bs.ts test:unit path), filteredConfigurations still contains all five BrowserStack launchers but this now forces concurrency to 1. That makes the full BrowserStack run execute the five browsers serially instead of the previous concurrency: 5, so the fallback/local full run no longer behaves “as before” and takes roughly five times longer; keep 1 only for the single-browser filtered case.

Useful? React with 👍 / 👎.

[thomas-lebeau]

❓ question: ‏would it be possible to use parseArgs here, so we don't need to read it in ci-bs.ts and pass it here as env variable?

[thomas-lebeau]

💬 suggestion: ‏I'd rather not mutate gitlab's build in environment variables.
Instead, and for consistency, I would do like for the e2e: keep the folder name unit-bs as normalized by getTestReportDirectory() and only change the filename of the report:

https://github.com/DataDog/browser-sdk/blob/f7daff3df9c57248f81aa8670cd6bea2b17cdfe8/test/unit/karma.base.conf.js#L77-L79C5

would become:

  junitReporter: {
    outputDir: testReportDirectory,
    outputFile: process.env.BS_BROWSER ? `results-${process.env.BS_BROWSER}.xml` : 'results.xml'
  },

[thomas-lebeau]

Suggested change

	junit: test-report/unit-bs-$BS_BROWSER/*.xml
	junit: test-report/unit-bs/*.xml

[thomas-lebeau]

Suggested change

	- node ./scripts/test/export-test-result.ts unit-bs-$BS_BROWSER
	- node ./scripts/test/export-test-result.ts unit-bs

[thomas-lebeau]

hmm, bs-wrapper.ts also checks that there is only one BS job at a time running.

This can make these job wait longer than necessary: (example)

2026-05-15T12:43:32.881957Z 01O �[32;1mOther build running, waiting...�[0m
2026-05-15T12:44:02.930861Z 01O �[32;1mOther build running, waiting...�[0m
2026-05-15T12:44:32.951962Z 01O �[32;1mOther build running, waiting...�[0m
2026-05-15T12:45:03.482930Z 01O �[32;1mOther build running, waiting...�[0m
2026-05-15T12:45:33.517166Z 01O �[32;1mOther build running, waiting...�[0m
2026-05-15T12:46:03.548156Z 01O �[32;1mOther build running, waiting...�[0m
2026-05-15T12:46:33.586724Z 01O �[32;1mOther build running, waiting...�[0m
2026-05-15T12:47:03.618851Z 01O �[32;1mOther build running, waiting...�[0m
2026-05-15T12:47:33.653588Z 01O �[32;1mOther build running, waiting...�[0m
2026-05-15T12:48:03.683845Z 01O �[32;1mOther build running, waiting...�[0m
2026-05-15T12:48:33.716840Z 01O �[32;1mOther build running, waiting...�[0m
2026-05-15T12:49:03.758851Z 01O �[32;1mStarting BrowserStackLocal...�[0m

I think there is another BS api that can return you the number of seats occupied, maybe we should use that instead instead of just hasRunningBuild