[juliusmarminge]

Problem

On Windows, 17 server tests fail on main because of three test-side assumptions; production is correct and the same files pass on Linux. Attachment tests search JSON-serialized payloads for raw paths, or expect raw paths where production JSON-quotes them, so Windows backslashes never match. The Claude tool-search settings test assumes the POSIX worktree-to-repository-root lookup of .claude/settings.local.json, while production deliberately reads it from cwd on Windows. A Codex test hard-codes the /-separated suffix providers/codex/<id>/shadow.

Reproduced on main 9e85538535 with the Windows test lane: run 41868132030, 17 failed, 311 passed, 13 skipped.

Change

Test-only; no production file changes.

• ClaudeAdapterV2.test.ts attachments: a small sdkMessageText helper reads the text blocks of the SDK message, and the five path assertions check that text, where production writes the raw path, instead of JSON.stringify of the whole content.
• userInputAttachments.test.ts: the four path assertions expect JSON.stringify(path), the form production emits.
• ClaudeAdapterV2.test.ts tool search: branches on yield* HostProcess.Platform. Off Windows the original assertions are unchanged; on Windows it asserts that the repository-root local file is not read from a worktree or a subdirectory and that a local file in cwd is.
• CodexDriver.test.ts: the expected suffix is built with NodePath.join.

Scope and approval

Fixes #30950

Verification

Windows (.github/workflows/windows-tests.yml, package=apps/server, the three changed files):

Run	Commit	Result
Before, 41868132030	9e85538535	17 failed, 311 passed, 13 skipped
After, 41868139348	8ca906a282	328 passed, 13 skipped

The 17 before-failures are the issue's inventory: 12 ClaudeAdapterV2 attachments, 3 userInputAttachments, 1 tool-search settings, 1 Codex managed catalog. The 13 skips are the existing Windows skips in these files.

Linux:

• vp test run on the three files: 341 passed, before and after.
• Backslash temp directory (TMPDIR containing \): main fails the 15 JSON-path tests; after the change they pass. Four CodexDriver npm-update tests still fail under that simulation on both sides because POSIX production normalizes \ to /; that does not happen on Windows and is not changed here.
• Tool-search test with HostProcess.Platform provided as win32 (temporary, not committed): fails at the root-local assertion on main, passes after.
• vp run check:changed: passed.

No UI or runtime surface changes, so there are no captures.

AI-assisted; verified with the tests and tools listed above.

🤖 Generated with Claude Code

[macroscopeapp]

Approvability

Verdict: Approved at 8ca906a

Macroscope's review found this PR approvable — This PR adjusts Windows-specific server test expectations and is confined to ignored test files. No production behavior change or material operational exposure is identified.

Notes:

• No code objects were reviewed. Approvability was decided on eligibility alone.

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

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Codex	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	21.5 KiB	21.5 KiB	0 B (0.0%)	24.4 KiB	✅
Codex	Live turn messages	1	1	0 (0.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Claude	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	21.9 KiB	21.9 KiB	0 B (0.0%)	24.4 KiB	✅
Claude	Live turn messages	1	1	0 (0.0%)	8	✅

Baseline: 9e85538 · PR result: 8ca906a · 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: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

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

[coderabbitai]

No actionable comments were generated in the recent review.

Walkthrough

The server tests now check attachment paths at the correct serialization depth, account for platform-specific Claude settings lookup, and build the Codex shadow-home path with NodePath.join.

Changes: Windows server test fixes

Layer / File(s)	Summary
Attachment path assertions apps/server/src/orchestration-v2/Adapters/ClaudeAdapterV2.test.ts, apps/server/src/provider/userInputAttachments.test.ts	Claude adapter tests extract message text before checking attachment paths. Attachment tests expect paths encoded with JSON.stringify.
Platform-specific path expectations apps/server/src/orchestration-v2/Adapters/ClaudeAdapterV2.test.ts, apps/server/src/provider/Drivers/CodexDriver.test.ts	Claude settings assertions distinguish Windows lookup from other platforms. The Codex shadow-home expectation uses NodePath.join.

Estimated code review effort: 2 (Simple) | ~10 minutes

Change: Other · Severity of issue fixed: Low

Suggested reviewers: maria-rcks, t3dotgg

Merge Risk: Minimal · up to 8ca90

The test-only changes correctly account for Windows path and settings behavior without weakening the covered checks.

Pre-merge checks: 4

Check name	Status	Explanation
Title check		The title names the main change: test-side fixes that stop Windows path and settings assumptions from breaking server tests. It follows the conventional commit style of the template. The changeset tou…
Description check		The description follows the template sections: Problem, Change, Scope and approval, and Verification. The Problem section states the failing Windows tests and the cause. The Change section lists the t…
Linked Issues check		Issue [#30950] requires three test-only corrections. The diff extracts Claude SDK text blocks before checking attachment paths, uses JSON.stringify(path) for serialized attachment references, branches…
Out of Scope Changes check		All changed lines support issue [#30950]. The sdkMessageText helper supports the requested attachment assertions, and the platform-specific settings assertions preserve the requested settings-selectio…

Full details: Title check

Explanation

The title names the main change: test-side fixes that stop Windows path and settings assumptions from breaking server tests. It follows the conventional commit style of the template. The changeset touches only test files, and the title says so by referring to server tests.

Full details: Description check

Explanation

The description follows the template sections: Problem, Change, Scope and approval, and Verification. The Problem section states the failing Windows tests and the cause. The Change section lists the test-only edits in each of the three files. Scope links issue #30950. Verification gives before and after Windows CI runs, Linux results, and a backslash-TMPDIR simulation. The author states that no UI changes apply. The template's Scope section asks for explicit maintainer approval or an explanation for a small, obvious fix. The author gives only an issue link and no approval comment or explanation of why the fix qualifies without one. This is a small gap, but the description is otherwise mostly complete.

Full details: Linked Issues check

Explanation

Issue [#30950] requires three test-only corrections. The diff extracts Claude SDK text blocks before checking attachment paths, uses JSON.stringify(path) for serialized attachment references, branches settings assertions on HostProcess.Platform while preserving POSIX checks and testing Windows cwd behavior, and builds the Codex suffix with NodePath.join. The diff changes only the three requested test files. No tests are skipped or weakened.

Full details: Out of Scope Changes check

Explanation

All changed lines support issue [#30950]. The sdkMessageText helper supports the requested attachment assertions, and the platform-specific settings assertions preserve the requested settings-selection coverage. The Codex path change addresses the requested separator assumption. No unrelated production, UI, or runtime changes appear in the reviewed diff.

Finishing Touches

Generate docstrings

• Commit to this branch
• Create a new PR

Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Recent review info

Run configuration

• Source mode: Auto
• Repository: pingdotgg/t3code
• Branch: fix/issue-30950
• File: .coderabbit.config.ts
• Commit: 8ca906a
• Layers: Repository TS config
• Review profile: CHILL
• Plan: Team
• Run ID: 8ebb4f8e-67e4-439b-8472-f41b838c0f72

Commits

Reviewing files that changed from the base of the PR and between 9e85538 and 8ca906a.

Files selected for processing: 3

• apps/server/src/orchestration-v2/Adapters/ClaudeAdapterV2.test.ts
• apps/server/src/provider/Drivers/CodexDriver.test.ts
• apps/server/src/provider/userInputAttachments.test.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 4 remain after this review.

[Yash-Singh1]

Superseded by #30983