The pinch clamped its lower bound to `ceil(fillPx)`. Rounding a fill height up
by the fraction of a pixel that 24 hours don't divide the viewport into makes
the timeline up to one pixel per hour taller than the viewport those hours are
supposed to fill — roughly 24px of leftover scroll, enough to bounce off
Android's overscroll stretch. `FitDay` resolves to the unrounded height and sits
still, so the two ways of reaching "the whole day on one screen" disagreed.
That inconsistency is what #290 is about.
Clamps to `fillPx` itself. The whole-pixel rounding exists so the hour gutter's
24 stacked boxes share a grid with the lines and blocks drawn at the fractional
height, and it still applies everywhere the pinch is free to move. The floor is
the one height where matching `FitDay` matters more — and because it is the
clamp *result* rather than a bound the gesture is merely held against, the pinch
lands on it exactly, so the next frame reads it back unchanged and the focal
anchor gets no correction to apply.
Week and Day both measure their viewport inside a `BoxWithConstraints` below the
all-day strip, so the strip appearing only changes the height both paths agree
on — that half of the report needed no change.
Tests: the two cases that encoded the old rounding are updated (the dead-space
invariant still holds, now exactly rather than by a pixel), plus one for the
reported symptom and one asserting the pinch floor equals what `FitDay`
resolves to at the same viewport.
Closes#290
Co-authored-by: Jean-Luc Makiola <business@jeanlucmakiola.de>
Reviewed-on: https://codeberg.org/jlmakiola/calendula/pulls/315
Jean-Luc Makiolaandmakiolaj
bdcf2b3823