[marcus7989]

[K-Marcus] The Windows Codex Desktop restart test used a short RUNNER~1 path while the restart script normalizes it to the long runneradmin form. Its fake process/launcher therefore failed with Wrong launcher even on the unchanged base.

This PR normalizes the fixture package path identically and includes the existing fixture log, stdout/stderr and exit status in a failure. It changes only rust/src/codex_accounts/codex_desktop.rs; production code and all safety/session assertions remain unchanged. The one-off workflow was removed after Paul's counter-review.

Validation at exact head c8e11559ed78df845a1ee732e511025b458d1320:

• W307 native Windows full suite: 2,036 library + 1 CLI passed, 0 failed, 1 pre-existing ignored; doc tests contain 0 cases, 0 filtered tests. Rust 1.98.0, locked MSVC x64, static CRT. The workflow is on a separate CI branch; the artifact manifest binds this exact source to the run. No registered PR-head check is claimed.
• Earlier original/fixed proof: all four pinned baseline sources reproduce the path error and pass with this same Rust-file fix.
• Format, whitespace, byte-identical production prefix and separate CI YAML lint passed.

Review: Paul's old-head counter-review had no blocker; his workflow removal request is fixed. Automated Codex review requested on the new head, still pending. Keep Draft; no merge approval. Planned order #21 → #20 → #19 → #18. No devices, release or .16 pin change.

Additional repository-required native Desktop-shell unit suite on unchanged cumulative source ef4aace55a59ed61fc66ab9b7d341129b82afa69: run 40908176267, 431 passed / 0 failed / 0 ignored / 0 filtered. Combined backend, CLI and Desktop-shell total: 2,481 PASS / 0 FAIL / 1 pre-existing ignored, same exact cumulative source. Both Rust manifests are verified; no app/device/customer-flow test is claimed.

The current-head Codex request has now had at least 20 minutes without a bot review, inline finding or bot reaction. Review remains pending; no automated no-findings or merge readiness is claimed.

[marcus7989]

[K-Marcus] @DreamyTalesPAN please counter-review #21 (test path fix) and #20 (Windows test timeouts). Plan: merge order #21 → #20 → #19 → #18 into vibetv; the .16 pin stays unchanged until a .17 candidate exists. Only review needed, no build from your side.

[DreamyTalesPAN]

Counter-review of #21 at head 7b8a834deab1eac699e4e90784be68d286ba2996 (read-only review, nothing built or pushed on my side).

Verdict: correct and minimal for the code change. No blocker. One non-blocking request on the workflow file.

• Product code untouched. In rust/src/codex_accounts/codex_desktop.rs the #[cfg(test)] module starts at line 314; both hunks (@@ -347 and @@ -389) are below it. The only other file is a new workflow.
• Fix is correct. The restart script already resolves the package root with [System.IO.Path]::GetFullPath(...) (lines 155/158), so the mock Get-AppxPackage and the expected launcher must use the same spelling. Applying GetFullPath to $fixturePackage in the fixture mirrors the product behavior. It does not mask a product bug: the product keeps its normalization and the test now matches it.
• No assertion weakened. Stop/launch, standalone-CLI safety (987654320 not stopped, /T rejected), session backup/clear and unrelated-file assertions are all unchanged. The only other change adds exit code, stdout, stderr and the fixture log to the failure message, which is diagnostic only.
• CI evidence checked. Run 40896491998 (desktop, SUCCESS) is on exactly this head. The guard job shows SKIPPED because the PR is a draft.
• Non-blocking: .github/workflows/codex-desktop-test.yml (+101 lines) is a one-off proof harness. It pins four historical SHAs and triggers only on pushes to codex/w296-desktop-restart-test. Once merged into vibetv it can never fire again and is dead weight. Since you plan to merge #21 into vibetv, please drop the workflow file from the merge (or from this PR) so only the 8-line codex_desktop.rs change lands. The red-to-green proof stays in run 40896491998.

Not checked: a real-Windows run on my side; I relied on the linked GitHub-hosted run.

[marcus7989]

[K-Marcus] @codex review

W307 applied Paul's counter-review at https://github.com/VibeTV-shop/vibetv/issues/284#issuecomment-6501070522. Exact new head c8e11559ed78df845a1ee732e511025b458d1320 removes only .github/workflows/codex-desktop-test.yml from 7b8a834. The PR now changes only rust/src/codex_accounts/codex_desktop.rs; its fixture-path normalization, diagnostics and all assertions are unchanged. Product code before #[cfg(test)] is byte-identical to the base.

The old native red/green proof remains at run 40896491998. A separate temporary CI branch will test this exact source and the cumulative #21 → #20 → #19 → #18 source without carrying an ephemeral workflow in this PR. New-run evidence will follow; no new CI success or merge approval claimed yet. Format and whitespace checks passed. No device action.

[marcus7989]

[K-Marcus] W307 native CI finished GREEN at exact source c8e11559ed78df845a1ee732e511025b458d1320: 2,036 library + 1 CLI passed, 0 failed, 1 pre-existing ignored, no filtered tests. Run/job. The separate workflow's artifact manifest records this source and workflow 90c559a41b1f5d974a2462f67e383fc6e9dfefd1. The PR contains only the Rust test fix; no one-off YAML. Codex request remains pending; 0 review threads at readback. No merge approval.