chore: land triage train 2 (#3825 #3826 #3828 #3829) - #3830
Conversation
|
🦞👀 Pull request received. I will update this pull request when review starts. ClawSweeper review completeClawSweeper finished reviewing this revision. The review result is being finalized. |
|
Codex review: needs maintainer review before merge. Reviewed September 21, 2026, 1:17 PM ET / 17:17 UTC (Revision 3). ClawSweeper reviewWhat this changesThe branch improves menu-bar item creation and placement repair, adds Kimi region selection and Cursor Grok Bot percentages, and removes the retired Crof provider. Merge readiness✅ Ready for maintainer review This remains useful, owner-sponsored integration work. No actionable introduced defect was found, and the earlier conflict and regeneration concerns are resolved. Priority: P2 Review scores
Verification
How this fits togetherCodexBar combines provider credentials and usage responses with saved preferences to display usage in the macOS menu bar. These changes affect regional requests, available providers, saved layouts, and status-item placement. flowchart LR
A[Credentials and region] --> B[Provider requests]
B --> C[Usage snapshots]
D[Saved layout preferences] --> E[Compatible layout loading]
C --> F[Menu-bar renderer]
E --> F
G[Saved positions and displays] --> H[Status-item creation]
F --> H
Before mergeNone. Agent review detailsSecurityNone. Review metrics
Technical reviewBest possible solution: Retain the shared provider and layout implementations with their compatibility projections, while keeping recurring placement corruption and Bartender identity persistence tracked separately. Do we have a high-confidence way to reproduce the issue? Yes for the source-level Kimi routing and missing Grok Bot layout capability; current main exposes those gaps. Recurring position corruption and Bartender identity changes have not been reproduced by this review. Is this the best way to solve the issue? Yes: the branch extends existing provider settings, descriptor capabilities, and persistence projections instead of creating competing implementations, and its intentional compatibility changes are documented and owner-approved. AGENTS.md: found and applied where relevant. Codex review notes: model internal, reasoning medium; reviewed against 35d479eafd3f. LabelsLabel changes:
Label justifications:
EvidenceWhat I checked:
Likely related people:
Rating scale
Overall follows the weaker of proof and patch quality. Workflow
History |
…3825) Create zero-length items, attach stable autosave names before registration and normal sizing, and preserve callback-selected widths. Share this factory across startup, provider vending, and merged-item recovery. Repair only matching saved positions beyond the widest attached screen plus 512 points, logging removed keys. Consolidate repair and vending helpers to keep production code net negative. Refs #3355 and #3201. Native AppKit write attribution and Bartender restart identity still require isolated runtime evidence. (cherry picked from commit 3f2ff08)
Keep the existing Codex branch allowlist bound to line 344 after relocating the status-item factory. The anchor text, provider fingerprint, reference count, and justification are unchanged. CI run 35584140370 and a local 41-test gatekeeper run reproduced the stale line-371 anchor. The corrected run passes 70 tests across four suites, make check passes with zero violations, and P2 review is scoped-clean. (cherry picked from commit 8fe8928)
Route Kimi API, web, cookie discovery, settings links, and the Usage Dashboard through the selected region. Keep China as the default and restrict unscoped CLI credentials to its historical host. Reuse JWT decoding and remove redundant forwarding and parsing paths to keep production code smaller. Fixes #3752 (cherry picked from commit b5e4ab1)
Adopt #3657 on the current 74-provider registry, remove Crof-only presentation policy, and retain the existing unknown-provider config path. Regenerate provider manifests, bundled plugin JavaScript, embedded CLI icons, and the documentation index with repository scripts. Verify retired config entries are ignored without rewriting on read and removed on save while preserving all supported provider entries. Co-authored-by: Aanish Bhirud <47579874+baanish@users.noreply.github.com> (cherry picked from commit 9909f0f)
Add descriptor-owned named extra percentages and hide unavailable allowances. Refresh the strip when Grok Bot changes independently of Cursor's other lanes. Keep V3/V2/V1 projections beside V4 so downgrades preserve reset choices and older-version edits still win when upgrading again. Fixes #3309 Refs #3712 Co-authored-by: yicone <yicone@gmail.com> (cherry picked from commit de88f94)
0ee064b to
5b08ff9
Compare
Integrates the four reviewed wave-2 lanes in one PR to avoid repeated changelog conflicts and competing macOS CI runs. The five lane commits retain their original authors, bodies, human co-author trailers, and cherry-pick provenance, with the lane PR number appended to each subject. The seven-commit stack also retains the Kimi regional-subtitle test correction and the existing regeneration commit.
Lane verdicts
Integration
Preserves main's new Highlights section, every Unreleased bullet, and contributor thanks, Kimi dashboard routing alongside HelmCode, and train 1's widget-label switch refactor with only Crof removed. Provider documentation and localized site counts use 77; the capability audit's separate totals are calculated from its audited rows. Architecture anchors and widget fingerprints are checked against the combined sources.
Validation
swift build --jobs 4passed.make checkpassed: zero SwiftLint violations across 2,547 files; all repository checks passed. Its first run caught only the corrected subtitle assertion's 123-character line, fixed by the wrap above.Widget color fingerprints reflect the 77-provider registry, and the architecture gatekeeper passes with the merged provider metadata and anchors.
Existing synthetic UI proof
Cursor Grok Bot layout, captured and inspected in #3829:
Before:

After:

Kimi's synthetic settings renders and all lane verdict evidence remain in the coordinator's local lane reports. The status-item changes make no claim of completed Bartender runtime verification.
Closes #3825
Closes #3826
Closes #3828
Closes #3829
Fixes #3752
Closes #3657
Fixes #3309
Closes #3712