fix(ct): Don't bloat metainfo.json file when # or ? is present in id #39074

mikstime · 2026-02-01T08:44:03Z

Proposal

Remove the ? or # part from dependencies in the metainfo.json file.

Description

@playwright/experimental-ct-core generates a metainfo.json file used for caching and other optimizations performed by Playwright under the hood.

The current implementation relies on a Vite plugin to generate this file. However, Vite prohibits the usage of # and ? in file names, meaning there is no reason to store them in the metainfo.json file.

Both # and ? are reserved for compatibility with browser URLs.

-- patak-dev

We experienced two major issues with the current behavior:

Our metainfo.json file exceeded 512MB, which is the hard limit for the JSON.parse() method, resulting in unexpected crashes
This data was serialized and deserialized each time tests ran, resulting in a 4x slowdown of tests in our project

80% performance improvement with proposed fix

Example

Metainfo.json file is reduced from 1.4M to 4.0K in provided example

1.4M    current/playwright/.cache/metainfo.json
4.0K    patched/playwright/.cache/metainfo.json

https://github.com/mikstime/playwright

dgozman · 2026-02-03T14:18:27Z

@mikstime Thank you for the PR! Is there any way you can share a repro, or even better - turn that into a test? That would help us to review and prevent future regressions.

mikstime · 2026-02-03T16:18:05Z

@mikstime Thank you for the PR! Is there any way you can share a repro, or even better - turn that into a test? That would help us to review and prevent future regressions.

Here is an example https://github.com/mikstime/playwright

dgozman · 2026-02-04T15:15:35Z

@mikstime Thank you for the PR! Is there any way you can share a repro, or even better - turn that into a test? That would help us to review and prevent future regressions.

Here is an example https://github.com/mikstime/playwright

Wonderful! Could you please add a test to tests/playwright-test/playwright.ct-build.spec.ts that follows this example? You can run it with npm run ttest playwright.ct-build.spec:<line>.

mikstime · 2026-02-04T17:40:04Z

@microsoft-github-policy-service agree

mikstime · 2026-02-04T17:42:10Z

@mikstime Thank you for the PR! Is there any way you can share a repro, or even better - turn that into a test? That would help us to review and prevent future regressions.

Here is an example https://github.com/mikstime/playwright

Wonderful! Could you please add a test to tests/playwright-test/playwright.ct-build.spec.ts that follows this example? You can run it with npm run ttest playwright.ct-build.spec:<line>.

Sure. Thank you for the instructions.

Running 1 test using 1 worker

  ✓  1 [playwright-test] › tests/playwright-test/playwright.ct-build.spec.ts:740:5 › should remove ? or # from deps in metainfo.json (3.1s)

  1 passed (3.8s)

dgozman · 2026-02-06T11:19:02Z

packages/playwright-ct-core/src/vitePlugin.ts

  if (!path.isAbsolute(id))
    return;
-  const normalizedId = path.normalize(id);
+  const normalizedId = path.normalize(id).split(/\?|#/)[0];


I think we should strip ? and # before the line 292, just in case there was something like /.. in the query params or hash.

It would be great to add a comment explaining/having an example of an import with a query param.

Fixed. I’ve stripped ? and # before line 292 and added a comment with an example of an import that includes query parameters.
For context: even without stripping, path.isAbsolute would still treat such paths as absolute. According to the Node.js documentation, paths like /baz/.. are considered absolute, and this also applies when query params or hashes are present, e.g.:

path.isAbsolute('C:/foo/..'); // true path.isAbsolute('C:\\foo\\..'); // true path.isAbsolute('/baz/..'); // true path.isAbsolute('/baz/../..'); // true

This is also true:

> p.isAbsolute('/dir/file#x=../../b.png') true > p.isAbsolute('/dir/file?x=../../b.png') true

That said, removing ? and # up front makes the intent clearer and avoids any edge cases involving traversal in query or hash parts.

dgozman · 2026-02-06T11:26:03Z

tests/playwright-test/playwright.ct-build.spec.ts

  expect(result.passed).toBe(1);
 });
+
+test('should remove ? or # from deps in metainfo.json', async ({ runInlineTest }, testInfo) => {


Thank you for the test, this really helped!

Just to understand the scenario better, do you have a custom loader/plugin/something else that handles query params? What's the actual usecase of yours, not the toy example we have here?

Can we have a test that resembles real life? IIUC, this should manifest in watch not triggering when you change a styles file that was imported as './style.css?inline'. At least, that's the impression I get from reading vite docs. If so, we can add a simple test next to should watch component when editing util.

Just to understand the scenario better, do you have a custom loader/plugin/something else that handles query params? What's the actual usecase of yours, not the toy example we have here?

We use a Vite plugin that transforms source code and appends query parameters to module imports. These query params are used to alter compilation behavior based on the user agent.

If so, we can add a simple test next to should watch component when editing util.

Sure. I’ve added two tests:

One that directly covers the query-param import case (e.g. style.css?inline), matching the real-world scenario described above.

Another that exercises the same behavior through a story file, since playwright compiles these cases differently.

dgozman

This looks great, thank you for following through!

dgozman · 2026-02-11T14:33:29Z

@mikstime It seems like this broke our component tests. Could you please take a look? You can run npm run ct for all component tests, or cd tests/components/ct-svelte && npm install && npx playwright test to see what's going on with specific tests. Please let me know if you stumble upon any issue.

mikstime · 2026-02-11T15:18:57Z

@mikstime It seems like this broke our component tests. Could you please take a look? You can run npm run ct for all component tests, or cd tests/components/ct-svelte && npm install && npx playwright test to see what's going on with specific tests. Please let me know if you stumble upon any issue.

Sorry about that — looks like I dropped a small but important guard during the rebase:

  if (deps.has(normalizedId))
    return;

Must be working now. npm run ct is passing now (18 passed (4.3m))

github-actions · 2026-02-11T17:17:39Z

Test results for "tests 1"

4 failed
❌ [playwright-test] › reporter-html.spec.ts:304 › merged › should not include image diff with non-images @macos-latest-node20
❌ [playwright-test] › ui-mode-test-network-tab.spec.ts:371 › should not preserve selection across test runs @macos-latest-node20
❌ [playwright-test] › update-aria-snapshot.spec.ts:299 › should update missing snapshots in tsx @macos-latest-node20
❌ [playwright-test] › update-aria-snapshot.spec.ts:669 › update-source-method › should overwrite source when specified in the config @macos-latest-node20

5 flaky

⚠️ [chromium-library] › library/popup.spec.ts:258 › should not throw when click closes popup `@chromium-ubuntu-22.04-arm-node20`
⚠️ [firefox-library] › library/inspector/cli-codegen-1.spec.ts:1082 › cli codegen › should not throw csp directive violation errors `@firefox-ubuntu-22.04-node20`
⚠️ [firefox-page] › page/page-wait-for-function.spec.ts:104 › should work with strict CSP policy `@firefox-ubuntu-22.04-node20`
⚠️ [playwright-test] › ui-mode-trace.spec.ts:384 › should reveal errors in the sourcetab `@macos-latest-node20`
⚠️ [playwright-test] › ui-mode-trace.spec.ts:812 › should update state on subsequent run `@macos-latest-node20`

38565 passed, 843 skipped

Merge workflow run.

mikstime · 2026-02-11T17:34:14Z

Test results for "tests 1"

4 failed ❌ [playwright-test] › reporter-html.spec.ts:304 › merged › should not include image diff with non-images @macos-latest-node20 ❌ [playwright-test] › ui-mode-test-network-tab.spec.ts:371 › should not preserve selection across test runs @macos-latest-node20 ❌ [playwright-test] › update-aria-snapshot.spec.ts:299 › should update missing snapshots in tsx @macos-latest-node20 ❌ [playwright-test] › update-aria-snapshot.spec.ts:669 › update-source-method › should overwrite source when specified in the config @macos-latest-node20

5 flaky
38565 passed, 843 skipped

Merge workflow run.

Can't reproduce locally on macOS + node 20. Can it be a fluke?

Tests passing locally

mikstime@mikstime-osx playwright-1 % node -v
v20.13.1
mikstime@mikstime-osx playwright-1 % npm run ttest update-aria-snapshot.spec.ts:299
mikstime@mikstime-osx playwright-1 % npm run ttest update-aria-snapshot.spec.ts:299

> playwright-internal@1.59.0-next ttest
> node ./tests/playwright-test/stable-test-runner/node_modules/@playwright/test/cli test --config=tests/playwright-test/playwright.config.ts update-aria-snapshot.spec.ts:299


Running 1 test using 1 worker

  ✓  1 [playwright-test] › tests/playwright-test/update-aria-snapshot.spec.ts:299:5 › should update missing snapshots in tsx (10.3s)

  1 passed (11.0s)
mikstime@mikstime-osx playwright-1 % npm run ttest update-aria-snapshot.spec.ts:669

> playwright-internal@1.59.0-next ttest
> node ./tests/playwright-test/stable-test-runner/node_modules/@playwright/test/cli test --config=tests/playwright-test/playwright.config.ts update-aria-snapshot.spec.ts:669


Running 1 test using 1 worker

  ✓  1 [playwright-test] › tests/playwright-test/update-aria-snapshot.spec.ts:669:7 › update-source-method › should overwrite source when specified in the config (8.9s)

  1 passed (9.6s)
mikstime@mikstime-osx playwright-1 % npm run ttest reporter-html.spec.ts:304       

> playwright-internal@1.59.0-next ttest
> node ./tests/playwright-test/stable-test-runner/node_modules/@playwright/test/cli test --config=tests/playwright-test/playwright.config.ts reporter-html.spec.ts:304


Running 2 tests using 2 workers

  ✓  1 [playwright-test] › tests/playwright-test/reporter-html.spec.ts:304:5 › created › should not include image diff with non-images (2.7s)
  ✓  2 [playwright-test] › tests/playwright-test/reporter-html.spec.ts:304:5 › merged › should not include image diff with non-images (3.1s)

  2 passed (5.7s)
mikstime@mikstime-osx playwright-1 % npm run ttest ui-mode-test-network-tab.spec.ts:371

> playwright-internal@1.59.0-next ttest
> node ./tests/playwright-test/stable-test-runner/node_modules/@playwright/test/cli test --config=tests/playwright-test/playwright.config.ts ui-mode-test-network-tab.spec.ts:371


Running 1 test using 1 worker

  ✓  1 [playwright-test] › tests/playwright-test/ui-mode-test-network-tab.spec.ts:371:5 › should not preserve selection across test runs (7.4s)

  1 passed (8.2s)

github-actions · 2026-02-11T17:53:28Z

Test results for "MCP"

10 failed
❌ [chrome] › mcp/autowait.spec.ts:19 › racy navigation destroys context @mcp-windows-latest
❌ [chromium] › mcp/roots.spec.ts:47 › check that trace is saved in workspace @mcp-windows-latest
❌ [firefox] › mcp/cli-cookies.spec.ts:59 › cookie-delete removes cookie @mcp-windows-latest
❌ [firefox] › mcp/cli-parsing.spec.ts:54 › wrong argument type @mcp-windows-latest
❌ [firefox] › mcp/cli-route.spec.ts:85 › unroute removes all routes @mcp-windows-latest
❌ [chrome] › mcp/cli-parsing.spec.ts:63 › should preserve leading zeros in string arguments @mcp-macos-15
❌ [chrome] › mcp/cli-save-as.spec.ts:19 › screenshot @mcp-macos-15
❌ [chrome] › mcp/cli-save-as.spec.ts:26 › screenshot @mcp-macos-15
❌ [chrome] › mcp/test-debug.spec.ts:234 › test_debug / evaluate x 2 @mcp-macos-15
❌ [chromium] › mcp/launch.spec.ts:21 › test reopen browser @mcp-macos-15

4769 passed, 135 skipped

Merge workflow run.

dgozman

All the related tests are passing, merging in. Thank you for the fix!

This comment has been minimized.

Sign in to view

mikstime force-pushed the patch-1 branch from ea4114f to fce9086 Compare February 4, 2026 17:37

dgozman reviewed Feb 6, 2026

View reviewed changes

mikstime force-pushed the patch-1 branch from fce9086 to 2bf3351 Compare February 10, 2026 19:17

mikstime added 2 commits February 10, 2026 22:19

fix(ct): Reduce metainfo.json file when # or ? is present in id

2c6829d

fix(ct): Reduce metainfo.json file when # or ? is present in id

cacad5a

mikstime force-pushed the patch-1 branch from 2bf3351 to cacad5a Compare February 10, 2026 19:20

Merge branch 'microsoft:main' into patch-1

0ab1ac5

dgozman approved these changes Feb 11, 2026

View reviewed changes

This comment has been minimized.

Sign in to view

fix(ct): Prevent infinite loop

84b05d9

dgozman approved these changes Feb 11, 2026

View reviewed changes

dgozman merged commit a7bcfe3 into microsoft:main Feb 11, 2026
30 of 33 checks passed

fix(ct): Don't bloat metainfo.json file when # or ? is present in id #39074

fix(ct): Don't bloat metainfo.json file when # or ? is present in id #39074

Uh oh!

Conversation

mikstime commented Feb 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Proposal

Description

Example

Uh oh!

dgozman commented Feb 3, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

mikstime commented Feb 3, 2026

Uh oh!

dgozman commented Feb 4, 2026

Uh oh!

mikstime commented Feb 4, 2026

Uh oh!

mikstime commented Feb 4, 2026

Uh oh!

dgozman Feb 6, 2026

Choose a reason for hiding this comment

Uh oh!

mikstime Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

dgozman Feb 6, 2026

Choose a reason for hiding this comment

Uh oh!

mikstime Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

dgozman left a comment

Choose a reason for hiding this comment

Uh oh!

This comment has been minimized.

dgozman commented Feb 11, 2026

Uh oh!

This comment has been minimized.

mikstime commented Feb 11, 2026

Uh oh!

github-actions bot commented Feb 11, 2026

Test results for "tests 1"

Uh oh!

mikstime commented Feb 11, 2026

Test results for "tests 1"

Uh oh!

github-actions bot commented Feb 11, 2026

Test results for "MCP"

Uh oh!

dgozman left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

mikstime commented Feb 1, 2026 •

edited

Loading