Keep Codex cost fixtures on the local day

[Finesssee]

Summary

Three cost scanner tests fail on CircleCI when a build runs between 00:00 and 01:00 UTC. This PR fixes the test fixtures; production code is unchanged.

Failing builds, all with the same three tests (2157 passed, 3 failed):

Build	Branch	Started (UTC)
977	port/micro-0.67.0-codex-direct-fork-baselines	2026-10-01 00:00:43
978	port/micro-0.69.0-kimi-stale-cli-guidance	2026-10-01 00:37:17
979	fix/reset-timer-overflow (#720)	2026-10-01 00:39:54

• codex_source_recovery_keeps_appended_duplicate_unpriced_after_cache_reload (tests.rs:1584)
• paginated_continuation_raises_inherited_baseline_from_total_last (paginated.rs:169)
• paginated_history_base_equal_parent_keeps_true_fork_subtraction (paginated.rs:230)

Cause

The Codex session fixtures stamped their events at now - 1h but filed the session under today's local date folder. In the first hour after local midnight that hour falls on yesterday:

• the recovery test read today's day bucket while the rows were filed under yesterday;
• the fork fixtures landed in two date folders (the fork fixture in today's, the continuation in yesterday's), which flipped the folder scan order the paginated tests depend on.

Fix

• New test helper recent_codex_fixture_time(): now - 1h, or the start of the local day when that hour reaches back into yesterday.
• write_codex_session_fixture and write_codex_session_fixture_with_inputs take both the event time and the date folder from it.
• The recovery test appends its row 30 minutes after the fixture time and reads that row's local day.
• Both paginated tests use the helper for their base time.
• recent_codex_fixture_time_stays_on_the_local_day pins the helper at fixed times around midnight in UTC+7 and UTC (the CI runner).

Validation

Run in a worktree on this branch (base main 7695471), through the local build gate:

Check	Result
cargo +1.98.0 fmt --all --check	pass
cargo +1.98.0 clippy --workspace --all-targets -- -D warnings	pass
cargo test -p codexbar cost_scanner	90 passed, 0 failed
cargo test -p codexbar	2161 passed, 0 failed, 1 ignored
cargo test -p codexbar-desktop-tauri	459 passed, 0 failed (bootstrap_payload_exposes_every_provider_variant filtered; it needs #711)

Affected areas

• rust/src/cost_scanner/tests.rs, rust/src/cost_scanner/tests/paginated.rs (tests only)

UI proof

Not applicable: test-only change, no UI or runtime behaviour changes.

Summary by CodeRabbit

• Tests
  • Updated cost-scanner test fixtures to account for local dates and timezone differences, including cases near midnight. Coverage also checks date handling when continuing paginated results and reloading cached data. These updates make automated checks more consistent across time zones and reduce the chance of misleading failures around day boundaries.

[coderabbitai]

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: 2b21c916-8b6e-4bdd-af88-2426593988bd

📥 Commits

Reviewing files that changed from the base of the PR and between 7695471 and 400f805.

📒 Files selected for processing (2)

• rust/src/cost_scanner/tests.rs
• rust/src/cost_scanner/tests/paginated.rs

Included review availability: This review used your included allowance. Your plan provides up to 4 included reviews per hour; 3 remain after this review.

---

📝 Walkthrough

Walkthrough

The cost scanner test fixtures now use timezone-aware timestamps that align with their local-date directories. Cache-reload and paginated tests use the shared fixture-time helper.

Changes

Fixture timestamps

Layer / File(s)	Summary
Recent fixture time and fixture writers rust/src/cost_scanner/tests.rs	A helper selects a recent timestamp that remains on the current local date. Codex fixture writers use that timestamp for both the directory date and event timestamp.
Cost scanner test timestamps rust/src/cost_scanner/tests.rs, rust/src/cost_scanner/tests/paginated.rs	Cache-reload and paginated tests use the shared fixture time. The cache-reload test derives its expected day from the appended event timestamp.

Priority: ⬇️ Low

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

Change: Bug fix

Merge Risk: ⚪ Minimal · up to 400f8

The test-only changes address midnight fixture failures without changing production behavior. No merge-blocking issue is identified; merge after normal checks pass.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 77.78% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 9 functions across 2 files.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: keeping Codex cost-scanner fixtures aligned with the local calendar day.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

• Fix all pre-merge checks with AI

✨ Finishing Touches 💡 1

📝 Generate docstrings 💡

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

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

Autopilot is currently an internal CodeRabbit preview.

---

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.}

[Finesssee]

Adversarial validation passed at afe1b747c816d0e1535e5c234f18545abf48701c

Scope: keep Codex cost fixtures on the local day (#721, GAP row FIX-MIDNIGHT), validated as merged into release/v0.70.0 (merge afe1b74 = merge of 400f805).

Attacks:

• The claim is a CI flake fix: fixtures stamped now - 1h could land on yesterday's local date folder when the CI runner runs in the 00:00–01:00 local window, so tests reading "today's" bucket failed in that hour. Verified in the merged tree: recent_codex_fixture_time() clamps the event time to now.date_naive().and_time(NaiveTime::MIN) when now - 1h reaches back into yesterday, and recent_codex_fixture_time_stays_on_the_local_day covers 00:40, 01:00, 00:00, the 08:30 normal case, and a UTC CI runner at 00:45 — the exact failure window.
• Test honesty: the helper is used by write_codex_session_fixture, write_codex_session_fixture_with_inputs, and the recovery test's appended row (appended_time = fixture time + 30 min so both rows share one local day). The paginated fork tests also switched from Utc::now() - 1h to the helper. That covers the three originally-flaking tests (codex_source_recovery_keeps_appended_duplicate_unpriced_after_cache_reload, plus both paginated fork tests).
• Looked for shortcuts: the helper is a pure function of now (testable via recent_fixture_time_at), so the clamp logic itself is directly tested instead of being buried in fixture-writing code. The production scanner is untouched — this is test-only, matching the PR body and the "n/a (test-only)" proof column.
• Windows-specific check: the clamp uses Local::now() and Local.from_local_datetime(...).earliest() (DST-safe); on a UTC CI runner the local day equals the UTC day, so the clamp reduces to start-of-day, which the dedicated test pins.

No defects found. READY for the un-draft rule.