fix(test): prepush hook test builds PATH with a POSIX-only separator

[luckywenapere]

The follow-up I offered in #2543. Independent of that PR — touches one test file
and no shipping code, so it can land in either order.

The bug

test/redact-prepush-hook.test.ts shadows git with a stub by prepending a temp
dir to PATH, built as:

PATH: `${stubDir}:${process.env.PATH}`

On Windows the separator is ;, so : produces a single unparseable entry, the
stub is never found, and the real git runs. The diff then succeeds,
gitStrict never throws, and the hook exits 0 where the test expects 1.

It surfaces as a wrong assertion (Expected: 1, Received: 0) rather than as a
portability problem, which is what made it hard to place — I initially assumed it
was a regression from my own changes in #2543 and only found the real cause after
reproducing it against a pristine checkout.

The fix

A prependPath helper mirroring the one already in
test/gstack-brain-context-load.test.ts, which handles both platform details:

• path.delimiter rather than a hardcoded :
• a case-insensitive lookup of the existing env key — Windows commonly spells it
  Path, and adding a second PATH alongside an inherited Path leaves which
  one wins up to the spawn implementation

On POSIX the helper resolves to { PATH: binDir + ":" + process.env.PATH } —
byte-identical to the expression it replaces, so behaviour there is
unchanged.

Why the test is also gated to POSIX

Fixing the separator alone does not make it pass on Windows, and it can't:

• The premise is that a signal-killed child yields spawnSync status === null.
  Windows has no equivalent — a force-killed process reports a non-zero exit
  code — so the branch the test names is unreachable there.
• The stub is a #!/bin/sh file named git, which Windows will not execute at
  all: process creation resolves commands through PATHEXT (.exe/.cmd/.bat) and
  ignores the shebang.

A Windows variant would have to assert the non-zero-exit path instead — a
different branch than the test name claims — so it's gated with
test.skipIf(process.platform === "win32"), matching
test/session-runner-timeout.test.ts and test/setup-emoji-font.test.ts.

The separator fix is still worth landing on its own: it's a latent trap for the
next test that copies this idiom, and the current form fails silently by
running the real binary rather than loudly.

Verification

Windows, bun test test/redact-prepush-hook.test.ts:

	before	after
pass	14	14
skip	0	1
fail	1	0

Three consecutive clean runs after the change.

Two caveats, since I can only verify one platform here:

• POSIX is unverified by me. The test should still run and pass there, by the
  byte-identical argument above, but it's worth a CI run.
• I saw the neighbouring new branch (zero remote sha) test fail once at
  ~6.8s under heavy parallel load, then pass 3/3 afterwards, and it also passes
  on a pristine checkout. I believe it's load-related flakiness unrelated to this
  change, but flagging it rather than leaving it unmentioned in case you've seen
  it in CI too.

---

🤖 Generated with Claude Code

Disclosing the tooling, as in #2543. The before/after numbers above were measured
by running the suite, not inferred, including against a pristine checkout to
confirm the failure pre-dated my changes.

[trunk-io]

Merging to main in this repository is managed by Trunk.

• To merge this pull request, check the box to the left or comment /trunk merge below.

After your PR is submitted to the merge queue, this comment will be automatically updated with its status. If the PR fails, failure details will also be posted here

[garrytan]

Thank you — this was absorbed on main (credited in the v1.6x CHANGELOG entries; roster in PR #2604). Closing.