[Yury Semikhatsky (yury-s)]

Summary

• Create the socket directory (CLI daemon, dashboard, browser.bind) and the transform cache directory with mode 0700, and refuse an existing one that the current user does not own. On a shared /tmp, another user could create these predictable directories first.
• When the transform cache directory is refused, transpile without the disk cache.

Fixes #43698

[github-actions]

Test results for "MCP"

5189 passed, 614 skipped

---

Merge workflow run.

[github-actions]

Test results for "tests 1"

1 failed
❌ [playwright-test] › reporter-html.spec.ts:3723 › created › speedboard › next/prev buttons should follow speed @macos-latest-node22

23574 passed, 708 skipped

---

Merge workflow run.

[github-actions]

🟡 Uncertain whether this PR caused the failure

Hi, I'm the Playwright bot. I checked the latest CI report, the diff, and refreshed test history for c1c0e66a21bf8743be165bbde19c4a70ad3fd9dc.

Details

There is one failing test. I found no direct link to the PR's changes, but no matching historical failure that would establish a pre-existing flake either.

Uncertain

[playwright-test] › reporter-html.spec.ts:3723 › created › speedboard › next/prev buttons should follow speed — macOS failure.

• Failure: after selecting three, the previous link stayed visible for 75 seconds (:3746). The test assumes its 100/200/300 ms waits make three the slowest test. Timing variation could explain the result, but the timeout alone does not establish a flake.
• Diff: the PR changes socket-directory permissions and compilation-cache access, not Speedboard sorting or navigation. This test runs an inline suite through the changed compilation cache, so an indirect effect cannot be ruled out.
• History: after updating the aggregated results, this exact test passed in all 619 other macOS run attempts across 618 SHAs, in the September 30–October 27 window. No matching failure or retry-rescued failure appeared on another PR/SHA, including other platforms.

A repeated comparison of this test on the PR head and its base, using the same macOS runner setup, is needed to distinguish a PR effect from an unconfirmed timing failure.

_{Triaged by the Playwright bot - agent run}

[Pavel Feldman (pavelfeldman)]

I would make this opt-in with a documented environment variable pointing to the location.

[Yury Semikhatsky (yury-s)]

Replaced by #43732 (sockets) and #43733 (transform cache).