feat(test-runner): experimental esbuild transpilation

[Pavel Feldman (pavelfeldman)]

Summary

• PLAYWRIGHT_EXPERIMENTAL_ESBUILD=1 transpiles TS/JS with esbuild instead of Babel; the loader, tsconfig resolution and compilation cache are unchanged.
• esbuild is an optional peer dependency (>=0.21.5) that users install themselves.
• Class fields follow useDefineForClassFields the same way as Babel.
• Known differences from Babel: .css imports are not stripped, assert import attributes are unsupported in ESM, PW_TEST_SOURCE_TRANSFORM is ignored.
• Depends on feat(test-runner): respect useDefineForClassFields from tsconfig #43262.

esbuild

real 0m4.392s
real 0m4.349s

babel

real 0m10.349s
real 0m10.255s

hot

real 0m2.637s

[github-actions]

Hi, I'm the Playwright bot and I took a first look at the CI failure.

🟢 The failure is a pre-existing flake

The only real failure has occurred on unrelated PRs on the same macOS/Node 22 bot. This PR is clear.

Details

Overall: this PR adds an opt-in esbuild transform. The failing job did not set PLAYWRIGHT_EXPERIMENTAL_ESBUILD, so it used the unchanged default transform path. The failure is an existing polling-output test that observed one fewer retry before the assertion passed.

Pre-existing flake / infra

• [playwright-test] › test-step.spec.ts:1729 › show api calls inside expects (@macos-latest-node22) — the same test failed on the same bot in unrelated feat(webkit): roll to r2331 #41783 (CI report) and fix(fetch): do not crash on late EPIPE after refused body #42116 (CI report). The line number has moved by one, but the project, file, title, and bot are identical. This confirms a pre-existing flake rather than a regression from this PR.

_{Triaged by the Playwright bot - agent run}

[github-actions]

Hi, I'm the Playwright bot and I took a first look at the current CI failure.

🟢 The failure is a pre-existing flake

The same test has failed with the same error on an unrelated PR and has also failed on main. This PR is clear.

Details

Overall: this PR adds an opt-in esbuild transform. The MCP workflow does not set PLAYWRIGHT_EXPERIMENTAL_ESBUILD, so it uses the unchanged default transform path. The failure is a known Windows/Firefox WebMCP registration race.

Pre-existing flake / infra

• [firefox] › mcp/cli-webmcp.spec.ts:58 › webmcp-list and webmcp-call (@mcp-windows-latest-firefox) — webmcp-list found the search tool, but the next command reported No WebMCP tool named "search". The same test and bot failed with that error on unrelated devops(bidi): fix publishing when a single csv report is uploaded #42764 (CI report), failed again on unrelated test(firefox): Alt+key should not type text on macOS #42935 (CI report), and failed on main. This confirms a pre-existing flake rather than a regression from this PR.

_{Triaged by the Playwright bot - agent run}

[github-actions]

Test results for "tests 1"

6 flaky

⚠️ [installation tests] › playwright-cdn.spec.ts:43 › playwright cdn failover should work (https://cdn.playwright.dev) `@package-installations-windows-latest`
⚠️ [chromium-library] › library/beforeunload.spec.ts:130 › should support dismissing the dialog multiple times `@chromium-ubuntu-22.04-arm-node20`
⚠️ [chromium-library] › library/webmcp.spec.ts:216 › should scope tools to their frame `@chromium-ubuntu-22.04-node24`
⚠️ [chromium-library] › library/video.spec.ts:762 › screencast › should work with video+trace `@chromium-ubuntu-22.04-node20`
⚠️ [firefox-library] › library/browsercontext-cookies-third-party.spec.ts:257 › third party 'Partitioned;' cookies `@firefox-ubuntu-22.04-node20`
⚠️ [firefox-library] › library/browsercontext-cookies-third-party.spec.ts:470 › top level 'Partitioned;' cookie and same origin iframe `@firefox-ubuntu-22.04-node20`

53368 passed, 1267 skipped

---

Merge workflow run.

[github-actions]

Test results for "MCP"

9165 passed, 1495 skipped

---

Merge workflow run.