fix(jsdom): support Blob on jsdom 28+ and 30.1 [backport to v4]

[rifflock]

Description

Written by an AI agent operated by @rifflock, who has reviewed it.

Refs #11336

Backports #11295 and #11379 to v4.

On Vitest 4 with jsdom 28+, any Blob or File that goes through URL.createObjectURL, a Request body or FormData reads back as the string "undefined", because jsdom 28 renamed _buffer to _bytes. Tests that check upload contents still pass, because they compare against the wrong value. jsdom 30.1 also stopped storing the impl under an own Symbol(impl), so the lookup throws Cannot read properties of undefined (reading '_buffer') (#11336).

One change beyond the backport: the bindings are resolved inside the jsdom copy the environment imported (dirname(require.resolve('jsdom/package.json'))). A bare jsdom/lib/... subpath is resolved from Vitest's own location and through NODE_PATH, which pnpm points at its store, so it can land in a different jsdom whose implForWrapper returns null for these wrappers. This repo hits it: v4 tests run on jsdom 27.4 while 29.1.1 is hoisted at the root.

main has the same bug. The reproduction is in the comments below, and the same change on top of main is on rifflock:fix/jsdom-bindings-resolution.

Please don't delete this checklist! Before submitting the PR, please make sure you do the following:

• It's really useful if your PR references an issue where it is discussed ahead of time. If the feature is substantial or introduces breaking changes without a discussion, PR might be closed.
• Ideally, include a test that fails without this PR but passes with it.
• Please, don't make changes to pnpm-lock.yaml unless you introduce a new test example.
• Please check Allow edits by maintainers to make review process faster. Note that this option is not available for repositories that are owned by Github organizations.

Tests

• Run the tests with pnpm test:ci.

test/core/test/environments/jsdom.spec.ts was run against jsdom 27.4.0, 28.1.0, 30.0.1 and 30.1.1. With the patch all 72 tests pass on every version. Without it, 28.1.0 and 30.0.1 fail the two new byte-level tests (6 failures) and 30.1.1 fails 12. pnpm lint and pnpm typecheck pass. The full pnpm test:ci was not run.

Documentation

• If you introduce new functionality, document it. You can run documentation with pnpm run docs command.

No new functionality.

Changesets

• Changes in changelog are generated from PR name. Please, make sure that it explains your changes in an understandable manner. Please, prefix changeset messages with feat:, fix:, perf:, docs:, or chore:.

🤖 Generated with Claude Code

[netlify]

✅ Deploy Preview for vitest-dev ready!

Built without sensitive environment variables

Name	Link
🔨 Latest commit	ef5ef7b
🔍 Latest deploy log	https://app.netlify.com/projects/vitest-dev/deploys/6abae8a59664c70008c69f02
😎 Deploy Preview	https://deploy-preview-11384--vitest-dev.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.
🤖 Make changes	Run an agent on this branch

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[hi-ogawa]

One change beyond the backport: the bindings are resolved inside the jsdom copy the environment imported (dirname(require.resolve('jsdom/package.json'))). A bare jsdom/lib/generated/idl/utils.js that doesn't exist in that copy (jsdom < 28.1) makes Node fall through to a different jsdom higher in node_modules, and that copy's implForWrapper returns null for these wrappers. This repo hits it: v4 tests run on jsdom 27.4 while 29.1.1 is hoisted at the root. main uses the same bare-subpath lookup; I haven't reproduced it there yet, but can send that fix separately.

Can you try reproduce that on main? main and v4 should have the same fix if we'd backport.

[rifflock]

Written by an AI agent operated by @rifflock, who has reviewed it.

Yes — main has the same bug, though not with the versions it currently installs.

createBlobImplGetter requires the bindings by bare specifier, so Node resolves them independently of the copy import('jsdom') returned. In a pnpm workspace that search also covers NODE_PATH, which pnpm points at node_modules/.pnpm/node_modules. When the two land on different copies, implForWrapper doesn't recognise the wrapper and returns undefined.

Repro on main (ad208cc29): point packages/vitest/node_modules/jsdom at the version under test, leave the hoisted node_modules/.pnpm/node_modules/jsdom at 30.1.1, then run test/unit/test/environments/jsdom.spec.ts:

env's jsdom	main	with the fix
27.4.0	4 failed / 22 passed	26 passed
28.0.0	4 failed / 22 passed	26 passed
28.1.0	26 passed	26 passed
30.0.1	26 passed	26 passed
30.1.1	26 passed	26 passed

The four are Fetch API accepts other APIs, FormData > keeps Blob bytes when used as a Request body, URL.createObjectUrl works properly and URL.createObjectURL keeps the Blob bytes, each TypeError: Vitest cannot read the bytes of a jsdom Blob.

One caveat on which installs hit it: pnpm --filter vitest add -D jsdom@27.4.0 on main hoists 27.4.0 into .pnpm/node_modules too, so both lookups agree and the suite stays green. It only shows when the hoisted copy differs from the one the environment imported.

The same change on top of main is on rifflock:fix/jsdom-bindings-resolution (a0a25f1d0) — lint and typecheck pass, full pnpm test:ci not run. @rifflock can open it as a PR against main whenever you'd like; it isn't open yet because the policy allows one pull request at a time and this is it.

[rifflock]

I'm re-running tests to make sure we're all good. Let me know if you want that PR to main, or if there's anything else I can do to be helpful.

[hi-ogawa]

Let me know if you want that PR to main

pls

[github-actions]

Hello @rifflock. Thank you for taking the time to contribute!

Unfortunately, the team accepts pull requests only from maintainers and approved contributors, so this pull request was closed automatically. We are sorry about that, it is not a judgement of your work. The number of pull requests grew beyond what the team can review, and this policy gives maintainers the space to triage and prioritize issues at their own pace.

If there is no issue for this change yet, please open one to discuss it with the team first. Your changes are not lost: a maintainer can reopen this pull request if the team decides to go forward with it. See our pull request policy for more context.

[rifflock]

Written by an AI agent operated by @rifflock, who has reviewed it.

The same change on top of main is ready on rifflock:fix/jsdom-bindings-resolution (3ea312e91). My earlier comment cited a0a25f1d0, which was amended away before the push — the branch is the reliable reference.

I haven't opened it as a PR because this one was closed automatically today, and a new PR from me would be closed the same way: I'm not on APPROVED_CONTRIBUTORS. Either /approve-user rifflock or reopening this PR unblocks it, whenever it suits you. If you'd rather not, the branch can stay as it is and a maintainer is welcome to take the commit.