ci(js-sdk): install Playwright Chromium without --with-deps - #1699
Conversation
The `pretest` hook ran `npx playwright install --with-deps chromium`, which shells out to apt on Linux and to a DISM Media Foundation enable on Windows on every invocation, even when the cached browser is restored. That dominated the Node legs: 20m57s of a 23m21s step on ubuntu-22.04 (18m59s of it downloading 18.4 MB of CJK/Cyrillic fonts at 16 kB/s) and 4m31s per run on windows-latest, while vitest itself takes ~2m20s. Every shared library Chromium needs is already on the runner images, and the headless `browser` project renders none of the fonts `--with-deps` adds, so drop it. Moving the install from `pretest` into a Node-only workflow step also keeps it out of every local `pnpm test`. Co-authored-by: Mish Ushakov <mishushakov@users.noreply.github.com>
…ming-e227 Co-authored-by: Mish Ushakov <mishushakov@users.noreply.github.com>
|
Package ArtifactsBuilt from 38c5de1. Download artifacts from this workflow run. JS SDK ( npm install ./e2b-2.41.1-cursor-author-pr-claiming-e227.0.tgzCLI ( npm install ./e2b-cli-2.16.3-cursor-author-pr-claiming-e227.0.tgzPython SDK ( pip install ./e2b-2.41.0+cursor.author.pr.claiming.e227-py3-none-any.whl |
| # `--with-deps` is deliberately omitted. It shells out to apt on Linux and | ||
| # to a DISM Media Foundation enable on Windows on every run, cache hit or | ||
| # not, which cost ~21m and ~4m30s respectively. The runner images already | ||
| # ship every shared library Chromium needs; the only packages | ||
| # `--with-deps` pulled in were CJK/Cyrillic fonts that the headless | ||
| # `browser` project never renders. Without it this step is a no-op once | ||
| # the cache above hits. |
|
@cursor check |
|
Addressed the review feedback: removed the 7-line comment above the Pushed as   |
Co-authored-by: Mish Ushakov <mishushakov@users.noreply.github.com>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 818bc91d90
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
No TASTE.md surface here; the change itself verifies clean, with one worthwhile follow-up
The diff is one workflow step plus one scripts entry — no src/ file, no export, no docstring — so the SDK design principles have nothing to bite on. I spent the run verifying the change instead.
What checked out
In this PR's own CI (run 32279418801, both Node legs green):
| leg | Install Playwright Chromium |
Run Node tests |
browser project |
|---|---|---|---|
| node / ubuntu-22.04 | 2s (Cache hit for: playwright-Linux-1.55.1) |
2m22s | browser (chromium) run.test.tsx (1 test) 840ms; 101 files passed, 1 skipped |
| node / windows-latest | 5s | 2m29s | passed in 3.6s; 101 files passed, 1 skipped |
Baselines from runs on concurrent branches that don't carry this change: Run Node tests (which still included the pretest hook) took 8m30s and 3m43s on ubuntu, 5m23s and 5m57s on windows. So the saving is real even when apt behaves — roughly 1m20s to 6m on Linux and about 3m on Windows, the latter being the deterministic DISM cost. The 20m57s in the description is the tail case rather than the expected win; worth keeping straight before someone cites it as the norm.
Locally, on a bare Ubuntu 24.04 VM with an empty ~/.cache/ms-playwright and playwright install-deps never run:
pnpm run playwright:installtook 7s cold and 0s warm (description says 6.5s / 0.78s).- The
browserproject then passes, which is a second independent confirmation that the apt fonts and libs weren't load-bearing — the Windows leg passing without the Media Foundation enable is the first. - With no browser present,
pnpm exec vitest run --project browserexits 1 and prints Playwright's own boxed remedy. Losing the auto-install fails loudly instead of silently skipping the suite, which was the main risk I wanted to rule out. pnpm run check-depsproduces byte-identical output on base and HEAD (both exit 1 on the same pre-existing findings), andplaywrightis not among the four flagged unused devDependencies, so the knip claim holds. Nothing in CI gates knip, for what it's worth.
On completeness of the pretest removal: js_sdk_tests.yml is the only workflow that mentions Playwright, and its only callers are sdk_tests.yml, release.yml and release-candidate.yml, so every CI path that runs the browser project now installs the browser explicitly. test:bun, test:deno and test:cf pass explicit --project lists that exclude browser, and the logs confirm those legs skip the new step. No pretest reference is left anywhere in the tree.
Shipping without a changeset matches precedent: ci(...)-scoped PRs #1662, #1646, #1632, #1622 and #1588 all merged with zero .changeset/ files (#1639 is the lone exception and it was itself about changesets). The already-resolved P1 from chatgpt-codex-connector doesn't apply — npm never runs a dependency's pretest, so renaming a dev script is invisible to consumers.
Worth acting on
The browser project never launches full Chromium, so this install can be a third of its size — details and measurements in the inline comment.
The documented way to run tests now fails on a fresh clone. AGENTS.md / CLAUDE.md tells contributors to run tests with pnpm run test; the root script is pnpm test --recursive --if-present, which reaches js-sdk's vitest run and therefore the browser project. CONTRIBUTING.md (2 lines) and DEV.md (3 lines) say nothing about tests, so this PR description is the only record of the new step. When it fails, Playwright's message says to run pnpm exec playwright install, which fetches all three browsers rather than the chromium-only script added here. One line in packages/js-sdk/README.md or AGENTS.md would close the gap — flagging it rather than insisting, since dropping the workflow comment suggests the minimalism is deliberate.
Coordination with #1609 (test(js-sdk): run the full unit test suite in a browser, open since July): it edits the same pretest line, moving it to a pretest:browser hook behind a dedicated test:browser script and matrix leg, which keeps the self-provisioning this PR gives up. Measured with git merge-tree: #1609 already conflicts with main in three files (js_sdk_tests.yml, tests/sandbox/network.test.ts, pnpm-lock.yaml), and this PR adds a fourth in packages/js-sdk/package.json.
Nits
- The description is stale in two spots: the workflow step no longer lands "with a comment recording why
--with-depsis omitted" (removed in818bc91at @mishushakov's request), and #1698 was already closed at 17:02 UTC, before the request to close it was written. - The one failing check is
JS SDK - cloudflareontests/sandbox/network.test.ts > maskRequestHost(cat: /tmp/headers.txt: No such file or directory). That leg skips all three Playwright steps, and the earlier run on this same branch — differing only by the removed YAML comment — was fully green, so it is unrelated flake.
Sent by Cursor Automation: /check SDK complies with TASTE.md
| "generate:mcp": "json2ts -i ./../../spec/mcp-server.json -o src/sandbox/mcp.d.ts --unreachableDefinitions --style.singleQuote --no-style.semi", | ||
| "check-deps": "knip", | ||
| "pretest": "npx playwright install --with-deps chromium", | ||
| "playwright:install": "playwright install chromium", |
There was a problem hiding this comment.
playwright install chromium fetches full Chromium (594MB) and the headless shell (321MB), but the browser project only ever launches the shell: vitest.config.mts sets headless: true, and moving ~/.cache/ms-playwright/chromium-1193 out of the way leaves the project passing (1 passed). The missing-browser error names it too — Executable doesn't exist at .../chromium_headless_shell-1193/chrome-linux/headless_shell.
Measured on a clean cache here:
| command | cold time | on disk |
|---|---|---|
playwright install chromium |
7s | 920MB |
playwright install --only-shell chromium |
2s | 326MB |
The browser project passes with only the shell installed, so --only-shell would cut ~594MB from the download and shrink the actions/cache artifact about 3x, which speeds up restore on a hit as well (the restore step was 3s on this run's ubuntu leg).
The trade-off: headed local debugging (--browser.headless=false) needs full Chromium, and a contributor would have to re-run without the flag. If you'd rather keep that available with one command, leaving this as-is is defensible — the CI win here is already the bulk of it.
There was a problem hiding this comment.
SDK test coverage: zero delta, and the browser project earns its keep
The diff contains no file under src/ or tests/ in either SDK, so the coverage delta is exactly zero by construction — a base/head A/B would only measure noise (repeat runs of the same commit move ±2 lines in template/buildApi.ts). Absolute numbers on 818bc91d9, E2B_API_KEY present so the live suites really ran:
| package | statements | branches | lines | tests |
|---|---|---|---|---|
js-sdk (unit+connectionConfig+template) |
82.06% (2214/2698) | 73.26% (1337/1825) | 81.96% (2164/2640) | 631 pass / 1 fail / 32 skip |
js-sdk browser project alone |
16.58% (448/2701) | 10.73% (196/1825) | 16.60% (439/2644) | 1 pass |
| python-sdk | 69.44% (10304/14838) | 46.69% (1328/2844) | — | 950 pass / 2 fail / 57 skip |
| cli | 17.40% (247/1419) | 21.72% (161/741) | 17.05% (236/1384) | 109 pass |
All 3 failures are the pre-existing httpbin-template environmental ones, identical on main. python splits into 83.58% hand-written (5268/6303) and 59.00% generated (5036/8535); the cli's 17% is an artifact of its tests driving the built CLI as a subprocess, which v8 cannot instrument.
Since the percentages cannot say anything about this PR, I spent the run on the question they cannot answer: does the suite this PR reprovisions still run, and is it worth running?
It still runs, and it fails loudly when it can't
pretestreally was live, and not for the reason it looks like: pnpm does not honour pre/post hooks by default,/workspace/.npmrcturns them on withenable-pre-post-scripts=true. Probed it on pnpm 10.33.3 —pretest→test→posttestall fire. So dropping the hook is a real behaviour change, exactly as the description says.pnpm run playwright:installresolves (playwrightis a direct devDep) and installs Chromium plus the headless shell in 8s.- With Chromium present and
playwright install-depsnever run on this bare Ubuntu 24.04 VM,--project browserpasses in 1.8s. Dropping--with-depsholds up. - With Chromium absent the project fails: exit 1,
Executable doesn't exist at …/chromium_headless_shell-1193/…. Not a silent skip — which is the property that actually matters for coverage, because it means CI cannot lose this suite quietly. - CI wiring is complete. js-sdk's
pnpm testhas exactly one call site in.github/(this workflow, line 110), andrelease.yml,release-candidate.ymlandsdk_tests.ymlall reach it viauses: ./.github/workflows/js_sdk_tests.yml, so they inherit the new step;cli_tests.yml'spnpm testruns inpackages/cli. Theif: matrix.runtime == 'node'gate is right, sincetest:bun/test:deno/test:cfpass explicit--projectlists that excludebrowser.
Its 6 unique lines badly understate its value
Measured rather than argued: the browser project executes 439 lines, but only 6 lines and 5 branch arms that the three node projects don't — utils.ts:45-46, api/metadata.ts:25, and three module-level import statements. On line coverage alone it looks deletable.
It isn't. utils.ts:45-46 is the typeof window !== 'undefined' arm of getRuntime(), and this project is the only thing in the repo that reaches it (zero hits everywhere else). Mutating runtime: 'browser' → 'unknown' there fails the browser test with SandboxError: 2: [unknown] Failed to fetch, because connectionConfig.ts:471 then selects https://sandbox.<domain>, which CORS-blocks in a browser — the exact case that line's comment exists for. One live test is the only guard on that choice, and every other test in the repo is blind to it. Keeping it cheap to run is worth doing.
Two follow-ups inline: a latent browser-upload defect that this coverage gap is hiding, and the local prerequisite now being undocumented. Neither is a blocker for this PR.
Sent by Cursor Automation: /coverage SDK Test Coverage Report
| "generate:mcp": "json2ts -i ./../../spec/mcp-server.json -o src/sandbox/mcp.d.ts --unreachableDefinitions --style.singleQuote --no-style.semi", | ||
| "check-deps": "knip", | ||
| "pretest": "npx playwright install --with-deps chromium", | ||
| "playwright:install": "playwright install chromium", |
There was a problem hiding this comment.
The gap in the suite this script serves — and a real defect hiding in it. Follow-up material, not something to fix here.
The browser project is the SDK's only source of runtime === 'browser', but its single test does commands.run + files.read only, so the places the SDK actually behaves differently in a browser are never executed there. Branch counts from the node run:
utils.ts:233cond-expr [0, 11]— theruntime === 'browser'arm oftoUploadBody's gzip path (buffer to a Blob instead of streaming) has zero hits.utils.ts:77if [0, 58]—dynamicImport's browser guard is called 58 times and throws never.- In the browser run,
if 224,cond-expr 233andif 238are all[0, 0]: the one browser test never callstoUploadBodyat all.
tests/utils.test.ts is already written as the contract test for precisely this — line 9 is const streams = runtime !== 'browser' and five assertions read expect(streamed).toBe(streams) — but the project's include is tests/runtimes/browser/**/*.{test,spec}.tsx, so it only ever runs under node, where streams is permanently true and the test.skipIf(!streams) never skips. That is the gap SDK-292 is aiming at.
So I ran it in the browser to see what it would say: copied to tests/runtimes/browser/, two import paths fixed, no other edits. It reports 1 failed / 15 passed / 1 skipped:
FAIL toUploadBody leaves an async-iterable foreign stream alone
AssertionError: expected '[object ReadableStream]' to be 'hello'
The premise in that test ("the platform accepts any async iterable as a body") and in foreignPlatformObjects.ts:54 is true of undici, not of browsers — BodyInit has no async-iterable member. Verified directly in HeadlessChrome 140:
| in a browser | result |
|---|---|
new Response(nativeStream).text() |
'hello' |
new Response(foreignAsyncIterableStream).text() |
'[object ReadableStream]' |
Symbol.asyncIterator in nativeStream |
true (Chrome ≥ 124) |
toDispatchableStream's Symbol.asyncIterator in stream clause is supposed to recognize a stream the platform will accept, so in a browser it waves a foreign/polyfilled stream through and toBlob buffers the stringified tag. A browser caller doing files.write(path, polyfilledStream) uploads the literal [object ReadableStream] instead of the data — the failure mode utils.ts:139-141 says the function exists to prevent.
I don't have a clean one-liner to offer. Gating the clause on runtime !== 'browser' fixes this case and keeps the node suite at 17/17, but it turns toUploadBody does not re-wrap a native stream when the global class was replaced red in the browser, because Chromium's native streams are async-iterable and that same clause is what rescues the replaced-global case there. The two contracts the file encodes need a nativeness probe that isn't async-iterability. Scratch files deleted; nothing committed.
|
|
||
| - name: Install Playwright Chromium | ||
| if: matrix.runtime == 'node' | ||
| run: pnpm run playwright:install |
There was a problem hiding this comment.
CI is wired correctly — I checked every pnpm test call site in .github/, and the reusable-workflow callers inherit this step. The loose end is local, and it's a documentation gap rather than a disagreement with the design: keeping the install off every contributor's pnpm test is the right call.
Nothing in the repo mentions playwright:install except the package.json line that defines it. Meanwhile AGENTS.md/CLAUDE.md tell contributors and coding agents to run tests with pnpm run test, and the root script is pnpm test --recursive --if-present, which reaches js-sdk's test → vitest run → the browser project. With enable-pre-post-scripts=true in .npmrc that path used to self-provision; now a fresh clone gets Playwright's own error, whose remediation hint is the generic npx playwright install (all three browsers) rather than this repo's pnpm run playwright:install.
Mirroring the description's local snippet into CONTRIBUTING.md or DEV.md would close it.
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
There was a problem hiding this comment.
Approved — this is a straightforward CI tooling change that drops --with-deps from the Playwright install to avoid the apt mirror stalls described in the PR. One nit was flagged inline about a latent Windows Media-Foundation dependency risk on a future cache miss; it's speculative, CI-only, and doesn't affect published SDK behavior, so it doesn't block approval.
Extended reasoning...
Overview
This PR touches only .github/workflows/js_sdk_tests.yml and packages/js-sdk/package.json. It replaces the pretest hook (npx playwright install --with-deps chromium) with an explicit playwright:install script (playwright install chromium, no --with-deps) invoked as its own CI step gated on matrix.runtime == 'node'. The change is well-motivated by the PR's own timing data showing the apt-based --with-deps install stalling for ~21 minutes on a slow Ubuntu mirror, fetching only CJK/Cyrillic fonts that the single headless browser test never renders. It also moves the install out of pretest so it no longer runs on every local pnpm test invocation for contributors who don't touch the browser project.
Security risks
None. This is CI/dev-tooling only — it does not change any published SDK code path, and playwright install chromium without --with-deps is a narrower operation (fewer packages fetched) than what it replaces, not a broader one.
Level of scrutiny
Low. This is a CI workflow and devDependency script change with no effect on runtime behavior of the published e2b package. The PR author (via the linked #1698 clone) did verify the change locally: Chromium launches and drives a real browser test without the dropped packages, and format/lint/typecheck/check-deps all pass for packages/js-sdk.
Other factors
The one prior human review comment ("remove the comment") was addressed in the latest commit (818bc91), and no changeset is expected here per project convention (the original hook in #977 also shipped without one, and this is inert for consumers). The inline nit about a possible Windows Media-Foundation regression on a future Playwright-version/cache-eviction cache miss is worth having on record, but it's conditional on a future event rather than a current failure, and the fix (re-adding --with-deps or the DISM step on the Windows leg only) is cheap to apply later if it materializes.
| "generate:mcp": "json2ts -i ./../../spec/mcp-server.json -o src/sandbox/mcp.d.ts --unreachableDefinitions --style.singleQuote --no-style.semi", | ||
| "check-deps": "knip", | ||
| "pretest": "npx playwright install --with-deps chromium", | ||
| "playwright:install": "playwright install chromium", |
There was a problem hiding this comment.
🟡 Removing --with-deps from the Playwright install (packages/js-sdk/package.json:38) also drops install_media_pack.ps1's Install-WindowsFeature Server-Media-Foundation step on Windows, which Chromium's launch-time host validation on Windows Server hard-fails without. This is currently masked because the actions/cache-backed Playwright cache still holds a validation marker from before this change, but on the next cache miss (Playwright version bump, cache eviction, or new runner image) the node/windows-latest browser vitest project could fail outright; consider keeping --with-deps (or an explicit Media Foundation install) on the Windows leg.
Extended reasoning...
What the bug is
packages/js-sdk/package.json:38 replaces playwright install --with-deps chromium with plain playwright install chromium, and the new step in .github/workflows/js_sdk_tests.yml runs it unconditionally for both the ubuntu-22.04 and windows-latest legs of the node matrix entry. On Linux, --with-deps just apt-installs shared libraries and fonts, and the PR's own verification shows those werent load-bearing (already the newest version). On Windows, however, --with-deps does something functionally different: it runs install_media_pack.ps1, which calls Install-WindowsFeature Server-Media-Foundation on Windows Server. That feature supplies mf.dll, mfplat.dll, msmpeg2vdec.dll, evr.dll, and avrt.dll, which Chromiums Windows build links against.
The code path that triggers it
Playwrights browser-launch path (playwright-core/lib/server/registry/index.js, _validateHostRequirementsForExecutableIfNeeded -> validateDependenciesWindows in dependencies.js) inspects the installed Chromium binarys DLL imports and throws a hard Host system is missing dependencies! error (pointing the user at Install-WindowsFeature Server-Media-Foundation) if those DLLs arent resolvable. This is not a soft warning about missing codecs — it prevents chromium.launch() from succeeding at all, which would fail the browser vitest project (packages/js-sdk/vitest.config.mts) that pnpm test runs unconditionally on the node matrix legs, including Windows.
Why nothing currently prevents it
The validation is skipped only when a DEPENDENCIES_VALIDATED marker file already exists inside the executables cache directory and is less than 30 days old. That directory is exactly the one keyed and persisted by the workflows actions/cache step (playwright-${{ runner.os }}-${{ steps.playwright-version.outputs.version }}). Because earlier CI runs (pre-PR) used --with-deps and therefore both enabled the Windows feature and produced a valid marker, todays cache hits skip revalidation and the leg passes green. That is precisely why this regression is invisible in the PRs own CI run.
Step-by-step proof of the failure condition
- Playwright is bumped (or the cache is evicted, or GitHub rolls a new
windows-latestimage) -> the cache key changes -> cache miss onwindows-latest. pnpm run playwright:installrunsplaywright install chromiumfresh, with no--with-deps, soinstall_media_pack.ps1never runs andServer-Media-Foundationis never enabled on this ephemeral runner.pnpm teststarts thebrowservitest project, which callschromium.launch().- Playwright validates host requirements, finds no recent
DEPENDENCIES_VALIDATEDmarker (fresh extraction), and inspects the DLL imports. - If
Server-Media-Foundationis not part of the runners base image, the missing DLLs causevalidateDependenciesWindowsto throw,browserfails, and the requirednode/windows-latestleg (which gatesSDK Tests Statusand the release publish step) goes red — for reasons unrelated to any actual product change.
The PRs own timing table is suggestive here: it attributes 4m31s on windows-latest to a "DISM Media Foundation enable" step, which is consistent with the feature not being preinstalled on that runner image (though, as one reviewer pointed out, Install-WindowsFeature can also be slow even as a no-op, so this data point alone isnt fully conclusive).
Why this is worth flagging despite the uncertainty
The PRs verification section only exercised the change on Linux/apt; the Windows Media-Foundation path was never tested end-to-end with a real cache miss, so the risk is unverified either way. The blast radius is limited to CI/dev-tooling (no effect on published SDK behavior), and the failure is conditional on a future cache miss rather than immediate — which is why this is a nit rather than blocking. But the fix is cheap and asymmetric: keep --with-deps (or a Install-WindowsFeature Server-Media-Foundation step) gated to the Windows leg only, while keeping the Linux leg on the lighter playwright install chromium. That preserves the PRs actual goal (killing the 20-minute apt stall on Linux) without introducing a latent, hard-to-diagnose failure mode on Windows that only surfaces after the cache masking wears off.
|
@cursor check |
Check complete — PR is readyReview feedback is addressed and CI is green on Review feedback
CI statusAll 29 checks passing, including:
Change summary
Non-blocking nits (no action taken)These came from automation, not from @mishushakov:
Approvals
Nothing further to implement unless you want one of the follow-ups above (e.g. |
2 participants
Supersedes #1698 (claimed via
/sdk claimby @mishushakov). Please close #1698 in favour of this PR — I have no write access to close it myself.This is a straight clone: the commit
f65f602from #1698 is applied here unmodified (original authorship and theCo-authored-by: Mish Ushakovtrailer preserved), withorigin/mainmerged in so the branch is current —mainhad moved one commit ahead (#1693), which touches none of the two files in this PR. The diff againstmainis identical to the original:.github/workflows/js_sdk_tests.ymlandpackages/js-sdk/package.json. Per the claim instructions, nothing was reviewed or changed.The original description follows, verbatim.
Closes SDK-339. Related: SDK-292, which introduced the
browserproject this install serves.Problem
packages/js-sdk/package.jsonhad apretesthook runningnpx playwright install --with-deps chromium.--with-depsshells out to apt on Linux, and to a DISM Media Foundation enable on Windows, on every invocation — regardless of whether the workflow's Playwright browser cache hit. On oneTest JS SDKrun that hook was 90% of the Node leg:Run Node teststotalpretest(--with-deps)vitest run(101 files, 100 passed)pretestDISM Media Foundation enableThe browser cache worked fine (
Cache hit for: playwright-Linux-1.55.1, restored in 3s). The time went to apt:apt-get update1m42s, then 18.4 MB fetched in 18m59s at 16.1 kB/s off a stalling Azure Ubuntu mirror (fonts-wqy-zenheialone stalled 7m49s).The mirror stall is transient; being on that path at all is the structural problem. Every shared library Chromium needs (
libnss3,libgbm1,libdrm2,libcairo2,xvfb, …) was alreadyalready the newest versionon the runner image — the only 9 new packages were CJK/Cyrillic fonts (fonts-wqy-zenhei,fonts-ipafont-gothic,xfonts-*) that the single headlessbrowsertest never renders. For comparison, in the same run the bun (2m44s), deno (2m41s) and cloudflare (2m1s) legs run the same test code with no Playwrightpretest.Change
packages/js-sdk/package.json: replace thepretesthook with an explicitplaywright:installscript (playwright install chromium, no--with-deps)..github/workflows/js_sdk_tests.yml: run it as its own step gated onmatrix.runtime == 'node', right after the existing browser-cache step, with a comment recording why--with-depsis omitted.Moving it out of
pretestalso keeps it off every localpnpm test, including for contributors who never touch the browser project.Usage
CI installs the browser as a distinct, cache-backed step:
Locally, the
browserproject needs Chromium once per Playwright version:Without it, the
browserproject fails with Playwright's own "Executable doesn't exist … runplaywright install" message; the other projects (unit,template,connectionConfig) are unaffected.Verification
Run on this branch with no prior Playwright deps installed on the machine:
pnpm run playwright:install: 6.5s cold (Chromium headless shell + ffmpeg, no apt), 0.78s as a no-op afterwards.pnpm exec vitest run --project browser: 1 passed. Chromium launches and drives a real sandbox without any--with-depspackages, confirming the fonts and libs weren't load-bearing.pnpm build+ fullpnpm test: 101 files, 99 passed / 1 skipped in 2m34s. The one failure istests/sandbox/network.test.ts > injected header is reflected by the httpbin sidecar, which fails with404: template 'httpbin' not found— it needs a prebuilthttpbintemplate that this agent's API key doesn't have, unrelated to this change.pnpm run format,pnpm run lint,pnpm run typecheckclean forpackages/js-sdk(the recursive root scripts fail only inpackages/python-sdk, whereuvisn't installed in this environment).pnpm run check-deps(knip) reports no new findings;playwrightis still resolved as a used devDependency through the new script.No changeset: this touches only dev tooling and CI, with no change to published behavior (the
pretest/playwright:installscripts are inert for consumers of the package). The commit that originally added the hook, #977, likewise shipped without one.