release / govulncheck (push) Successful in 15s
release / web (push) Successful in 1m11s
release / go (push) Successful in 1m29s
release / android (push) Successful in 4m46s
release / Build signed APK (releases and dev) (push) Successful in 4m58s
release / integration (push) Successful in 15m34s
release / Attach APK to the Release (tag releases only) (push) Skipped
release / Build + push container image (push) Successful in 14s
release / Verify release artifacts (tag releases only) (push) Skipped
The drag-offset-to-row arithmetic lived inline in queueReorderDrag's onDragEnd lambda, which no JVM test can reach. Web's copy, offsetToDelta, has been extracted and tested since the start, and the Android comment says it mirrors web, but only one side could be held to that. - QueueDragMath.kt adds queueDragDelta (web's offsetToDelta) and queueDragTarget (delta plus the clamp to the queue). Kotlin's roundToInt breaks ties toward positive infinity, the same as JS Math.round, so the web cases carry over exactly, including half a row up staying put. - queueDragTarget returns the start index for an empty queue instead of letting coerceIn(0, -1) throw. - QueueDragMathTest mirrors queue-row-math.test.ts one case at a time, plus clamping past either end, a sub-half-row drag, an unmeasured row, and the empty queue. No behaviour change for a non-empty queue. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>