test: resolve the Windows temp directory to its long name

[juliusmarminge]

GitHub's Windows runners set TEMP to the 8.3 short form
(C:\Users\RUNNER~1...), and every path that goes through git or realpath
comes back in the long form, so equality checks between a temp directory and
its canonical form fail. A shared setup file, run by every package's vitest
config, points TEMP/TMP at the long form once so os.tmpdir() and everything
built on it agree with the canonical spelling.

bootstrap.test opens the platform null device instead of /dev/null and skips
the FIFO case, since Windows has neither.

Part of the Windows test-suite stack rooted at #9564; the manual Windows lane comes from #9538.

Model: Claude Fable 5 in Claude Code.

🤖 Generated with Claude Code

Note

Resolve Windows temp directory to long name in test setup

• Adds longTempDir.ts, which on win32 resolves the canonical long form of the temp directory and assigns it to TEMP and TMP; non-Windows hosts and resolution failures are left unchanged
• Wires this setup module into the root, desktop, web, and client-runtime Vite test configs via an absolute path or package subpath export in package.json
• Updates bootstrap.test.ts to select the platform-specific null device (NUL vs /dev/null) and skips the FIFO timeout test on Windows
• Behavioral Change: Windows test runs now see long-form TEMP/TMP paths; if resolution fails the environment is untouched

^{Macroscope summarized 5032af7.}

---

Note

Low Risk
Test-only environment and platform-guard changes; production runtime behavior is unchanged aside from optional Vitest setup on Windows.

Overview
Adds shared Vitest setup longTempDir.ts that on win32 rewrites TEMP and TMP to the canonical long path from realpathSync, so os.tmpdir() matches git/realpath spelling on GitHub Windows runners (8.3 short names). The module is exported from @t3tools/shared and registered in the root, desktop, web, and client-runtime Vite test configs (server inherits it via mergeConfig with its existing git setup).

bootstrap.test.ts uses NUL vs /dev/null for null-device FD tests and skipIf(windowsHost) on the FIFO timeout case because Windows lacks mkfifo.

^{Reviewed by Cursor Bugbot for commit 5032af7. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ No successful main baseline artifact is available yet. This run establishes the initial measurement.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	—	13.6 KiB	—	15.1 KiB	✅
Codex	Thread snapshot wire	—	7.0 KiB	—	7.3 KiB	✅
Codex	Live turn WebSocket wire	—	6.6 KiB	—	7.8 KiB	✅
Codex	Live turn WebSocket decoded	—	57.1 KiB	—	66.4 KiB	✅
Codex	Live turn messages	—	10	—	21	✅
Claude	Total thread wire	—	13.5 KiB	—	15.1 KiB	✅
Claude	Thread snapshot wire	—	7.0 KiB	—	7.3 KiB	✅
Claude	Live turn WebSocket wire	—	6.5 KiB	—	7.8 KiB	✅
Claude	Live turn WebSocket decoded	—	57.8 KiB	—	66.4 KiB	✅
Claude	Live turn messages	—	9	—	21	✅

Baseline: unavailable · PR result: 5032af7 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 113.8 KiB
• Claude decoded thread snapshot: 114.5 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This is a focused, test-only Windows portability change that canonicalizes temporary directories and avoids unsupported POSIX test primitives, with no product-path behavior changes. The new shared test setup also adds a file-level suppression for an Effect static-analysis diagnostic, warranting explicit human review.

No code changes detected at 5032af7. Prior analysis still applies.

^{You can add or adjust custom eligibility rules. Learn more.}

[ylcn91]

Same failure on macOS, from the other side: os.tmpdir() is /var/folders/… and anything canonicalised comes back as /private/var/folders/…. On a Mac from today's main:

FAIL  src/entrypoint.test.ts > isEntrypoint > matches through a symlinked entrypoint, as npm and npx install it
AssertionError: expected false to be true

FAIL  src/provider/Layers/CursorProvider.test.ts > Cursor skills > discovers recursive project skills with project precedence
-     "path": "/var/folders/…/T/cursor-skills-workspace-s4PxXA/.cursor/skills/internal/SKILL.md",
+     "path": "/private/var/folders/…/T/cursor-skills-workspace-s4PxXA/.cursor/skills/internal/SKILL.md",

#9400 and #9401 fix those two in place by resolving the temp dir before asserting (#9389, #9388). Your setup file would cover them as well: with TMPDIR pointed at realpathSync.native(os.tmpdir()) and the tests untouched, both files pass (30/30), same as with the two patches applied.

So if you would rather drop the win32 guard in longTempDir.ts and set TMPDIR on darwin too, I am happy to close #9400/#9401 in favour of this one, or to fold that change into here, whichever you prefer.