Code Review

Overview

This is a minimal, well-scoped fix for a real mobile UX bug: the geocode-bar's drag handle (with pointer-events: auto) sits at the same y-coordinate as the guidance pill's Stop button after navigation starts, silently stealing taps. Adding hideGeocodeBar() before startGuidance() is the correct and natural fix — the bar has nothing more to offer once guidance is active.

Code Quality

Correctness ✅
The fix is placed exactly right — hiding the bar before startGuidance() means the UI is clean when the guidance pill appears. hideGeocodeBar() → snapTo('hidden') is idempotent, so calling it when the bar is already hidden (edge case) is safe.

Scope ✅
The PR description's reasoning about why the other entry points (sheet-result__nav-btn, long-press without the bar) don't exhibit the same overlap is sound. The narrow fix is appropriate.

Minor: comment length ⚠️
CLAUDE.md says "multi-line comment blocks — one short line max". The 4-line block is informative, but it could be collapsed to a single line and still convey the essential non-obvious constraint:

// Bar's drag-handle overlaps the guidance pill's Stop button at this y-coord (#175).
hideGeocodeBar();

The issue number is enough for future readers to trace the full backstory. Not a blocker, but worth tightening.

Test Coverage

• Automated suite (633 tests) passes — appropriate, since this interaction is a CSS/DOM positioning issue that's not realistically unit-testable.
• Manual test plan covers the happy path, the alternate flow ("Go to location" → bar opens → Navigate), and the regression case (drag-handle still works when guidance is inactive). That's the right set.

No concerns on

• Performance (no hot path touched)
• Security (no user-facing data flows changed)
• TypeScript / lint (already green per test plan)

Verdict

Approve with nit. The logic is correct, the scope is tight, and the PR description thoroughly explains the root cause. Only suggestion is collapsing the comment to a single line per project style. Ready to merge once the manual checklist items are ticked off.

Code Review

Summary: Minimal, well-targeted bug fix. Adds hideGeocodeBar() before startGuidance() in the geocode-bar's Navigate click handler to clear the z-index overlap that was swallowing taps on the guidance pill's Stop button.

Correctness ✅

• The fix is in exactly the right place — the barNavBtn click handler is the only path that starts guidance while the geocode-bar can still be visible. The PR description's scope analysis is accurate: the dropdown's .sheet-result__nav-btn path skips the bar entirely, so it correctly receives no change.
• hideGeocodeBar() is synchronous (snapTo('hidden')), so currentPinLatLng and currentPinLabel are still fully in scope when startGuidance consumes them — no ordering hazard.
• Consistent with the existing pattern: hideGeocodeBar() is already called in every other dismiss path (lines 658, 688, 710, 740).

One minor observation

If startGuidance fails (e.g. Valhalla returns no route), the geocode-bar will have already been dismissed and the user loses the reverse-geocoded pin context. This is the same behaviour as the other dismiss paths, so it's consistent — but worth a conscious choice. The alternative would be to call hideGeocodeBar() only on success (inside the guidance-started callback). Given that the toast message already signals failure and the user can long-press again to reinstate the pin, the current approach seems fine; just flagging it in case you hadn't considered it.

Code style ✅

The comment explains the non-obvious why (z-index/layout collision, issue reference) rather than the what — exactly right per project conventions.

Tests ✅

No unit-testable logic changed; the fix is a DOM event handler interaction. The test plan's manual checklist covers the affected flows, and the existing 633-test suite passes.

Verdict: Looks good to merge. The fix is correct, minimal, and consistent with the rest of the file.