test(codex_cli): build session paths component by component so hashes match on Windows

[NickAme03]

Closes #225.

test_message_hashes_remain_stable_after_archiving built the active session
directory with a single join(".codex/sessions/2026/07/22"). On Windows those
forward slashes stay inside the path string, while canonical_session_path
rebuilds the archived path one component at a time and gets native separators.
conversation_hash hashes the string form of the path, so the two sides hashed
different strings and the assertion failed.

Both paths are now built component by component, which produces the same string
on either platform. The comment above them says why, so the next person editing
this fixture does not undo it.

The other nine join calls in this file that embed slashes are left alone. They
only create directories, and Windows accepts forward slashes there, so changing
them would widen the diff without fixing anything.

Verification. 409 passed, 0 failed on both ubuntu-latest and windows-latest,
on the same tree as this branch plus a temporary workflow that is not part of the
change: https://github.com/NickAme03/splitrail/actions/runs/30660532341

The failure this fixes is visible in the run reported in #225, taken on main
before this branch existed: 404 passed and 1 failed, with Windows the only red
job. https://github.com/NickAme03/splitrail/actions/runs/30655673828

This is a test fixture fix. It does not change any counting behaviour, and the
production path stays as it was: WalkDir already yields native separators on
both sides, which is why the numbers splitrail reports were never affected.

One thing this does not do, since it belongs to you and not to a fix: checks.yml
runs only on ubuntu-latest and never runs cargo test, so nothing in CI would
have caught this and nothing will catch the next one. I am happy to add a Windows
test job in a separate pull request if you want it.

[coderabbitai]

📝 Walkthrough

Walkthrough

The Codex CLI tests now cover Codex data-directory resolution, root-aware model configuration cases, and platform-native session paths for stable archive hash comparisons.

Changes

Codex CLI test coverage

Layer / File(s)	Summary
Codex resolution and configuration cases src/analyzers/tests/codex_cli.rs	Tests cover default, custom, and empty Codex homes, plus missing, empty, configured, blank, invalid, and incomplete model configuration files.
Native session path construction src/analyzers/tests/codex_cli.rs	The hash-stability test builds session paths component by component and retains the hash assertions.

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

Possibly related PRs

• Piebald-AI/splitrail#224: Adds the Codex root-resolution and configuration behavior covered by these tests.

Suggested reviewers: jimyag

Poem

I build each path one part at a time,
With native separators in line.
Config cases join the test parade,
Stable hashes no longer fade.
A rabbit checks the suite with glee. 🐇

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Out of Scope Changes check	⚠️ Warning	The pull request also adds Codex home and fallback-model configuration tests that are unrelated to issue #225.	Move the unrelated data-directory and fallback-model configuration tests to a separate pull request or link them to relevant issues.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	The test paths now use component-by-component construction, which addresses the Windows hash mismatch in issue #225.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly describes the test change that builds session paths component by component to fix Windows hash matching.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

🧹 Nitpick comments (2)

src/analyzers/tests/codex_cli.rs (2)

562-572: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick win

Assert precedence over an existing home config.

The test creates only the custom config. It therefore proves that the custom path is read, not that it overrides ~/.codex/config.toml. Write a different model to the home config to cover the precedence rule.

💚 Proposed test strengthening

 fn test_fallback_model_respects_codex_home() {
     let dir = tempfile::tempdir().unwrap();
+    let home_codex = dir.path().join(".codex");
+    std::fs::create_dir(&home_codex).unwrap();
+    std::fs::write(home_codex.join("config.toml"), "model = \"gpt-5-mini\"\n").unwrap();
     let custom = dir.path().join("custom-codex");

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/analyzers/tests/codex_cli.rs` around lines 562 - 572, Strengthen
test_fallback_model_respects_codex_home by creating a different model
configuration in the home-root config.toml before calling
get_fallback_model_with_roots. Keep the custom Codex home configuration set to
"gpt-5.5" and assert it is selected over the conflicting home configuration.

---

514-558: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low value

Simplify the fixture writes and consider a table-driven form.

Each case creates, writes, and drops a File only to replace config.toml. std::fs::write truncates and closes in one call, and Line 566 already uses it. A table of (contents, expected) pairs would also remove the repeated assertion blocks.

♻️ Proposed refactor

-    // Test case 2: empty config file
-    let mut file = std::fs::File::create(&config_path).unwrap();
-    file.write_all(b"").unwrap();
-    drop(file);
-    let model = get_fallback_model_with_roots(None, Some(dir.path()));
-    assert_eq!(model, "gpt-5");
-
-    // Test case 3: model configured in config file
-    let mut file = std::fs::File::create(&config_path).unwrap();
-    file.write_all(b"model = \"gpt-5.5\"\n").unwrap();
-    drop(file);
-    let model = get_fallback_model_with_roots(None, Some(dir.path()));
-    assert_eq!(model, "gpt-5.5");
+    // Contents of config.toml mapped to the expected fallback model.
+    for (contents, expected) in [
+        ("", "gpt-5"),                      // empty file
+        ("model = \"gpt-5.5\"\n", "gpt-5.5"), // configured model
+        ("model = \"\"\n", "gpt-5"),          // empty value
+        ("model = \"   \"\n", "gpt-5"),       // whitespace-only value
+        ("model = \n", "gpt-5"),              // invalid TOML
+        ("other_key = \"value\"\n", "gpt-5"), // model key omitted
+    ] {
+        std::fs::write(&config_path, contents).unwrap();
+        assert_eq!(
+            get_fallback_model_with_roots(None, Some(dir.path())),
+            expected,
+            "unexpected model for config {contents:?}"
+        );
+    }

Keep case 1 before the loop and case 5 after it, because neither uses a config file.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/analyzers/tests/codex_cli.rs` around lines 514 - 558, Simplify the
repeated config-file test setup in get_fallback_model_with_roots tests by
replacing each File::create/write_all/drop sequence with std::fs::write.
Consolidate cases 2–4 and 6–8 into a table-driven loop of contents and expected
models, keeping case 1 before the loop and the None home-directory case after
it. Preserve all existing inputs and assertions.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@src/analyzers/tests/codex_cli.rs`:
- Around line 562-572: Strengthen test_fallback_model_respects_codex_home by
creating a different model configuration in the home-root config.toml before
calling get_fallback_model_with_roots. Keep the custom Codex home configuration
set to "gpt-5.5" and assert it is selected over the conflicting home
configuration.
- Around line 514-558: Simplify the repeated config-file test setup in
get_fallback_model_with_roots tests by replacing each
File::create/write_all/drop sequence with std::fs::write. Consolidate cases 2–4
and 6–8 into a table-driven loop of contents and expected models, keeping case 1
before the loop and the None home-directory case after it. Preserve all existing
inputs and assertions.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: b780dd40-e10f-4053-9a71-22fbd3c8086a

📥 Commits

Reviewing files that changed from the base of the PR and between fa2942e and 2d6c39a.

📒 Files selected for processing (1)

• src/analyzers/tests/codex_cli.rs