element/scrollarea: add optional draggable scrollbar

[UncleJ4ck]

Redone as a real element in the tree, per your suggestion.

showScrollbar(true) adds a track and a draggable thumb. The scroll offset is applied to every child, so the content now lives in an inner layer that scrolls, and the scrollbar sits on top as a sibling that stays put. The strip takes mouse input:

• drag the thumb to scroll
• click the track to jump there and keep dragging
• thumb brightens on hover

Vertical and horizontal are both handled. The paintOver() hook from the first version is gone.

Tests (headless, hard assertions): scroll clamp, thumb size/position vs scroll, a tiny-viewport regression for a std::clamp abort that the first version hit, and a drag that asserts the scroll moves proportionally then stops on release.

[vaxerski]

would it not be better to make the scrollbar actually part of the layout tree so it can receive input events and stuff?

[UncleJ4ck]

actually that's a good idea

[UncleJ4ck]

redid it as a real element in the tree. the thumb is draggable now, click on the track jumps and keeps dragging, and it brightens on hover. content goes into an inner scrolled layer so the bar can sit on top and take input. dropped the paintOver hook.

[vaxerski]

it can still disappear, as long as its absolutely sized and not in a layout it wont reserve space.

[UncleJ4ck]

yep that's how it's set up. the strip is HT_POSITION_ABSOLUTE anchored to the right/bottom edge, content sits in a separate inner layer at full size, so the bar overlays it and reserves no space. and it's removed entirely when there's nothing to scroll, so it disappears when not needed.

[vaxerski]

can you add a scroll area to the controls test? You could put the entire thing in a vertical scroll area for testing

[UncleJ4ck]

done. the controls test already had everything in a vertical scroll area, so i turned the bar on there. it shows when the controls overflow the window and hides when they fit, and you can drag it.

[vaxerski]

very jittery when scrolling down continuously, constantly jitters and snaps up and down.

[UncleJ4ck]

good catch. the snap was the new inner content layer getting laid out twice per reposition: positionChildren(self) recurses through it and writes the content's natural height, then the grow pass overwrites that with the unconstrained height. maxScroll() measured the overflow live off whatever was currently in the content's box, so a wheel event landing while it was in the natural state read a smaller overflow and yanked the clamp down, then the next frame put it back. that back and forth was the jitter.

fix is to measure the overflow once per reposition while the content is in its grown state (recalcMaxScroll()) and have maxScroll() return the cached value, so the off-frame wheel/clamp path can't sample the transient box.

reproduced it under hyprland on the controls scroll area: max flipped 140 <-> 100 continuously while scrolling and the offset snapped with it, now it holds steady. added a regression test (scrollAreaMaxScrollIsStableAcrossContentRelayout) that fails without the cache. pushed to the same branch.

[vaxerski]

Still here, less jittery, but bugged. FWIW, I am using continuous scroll for this (touchpad)

I scroll max down, and let go, then it snaps to the top. I start scrolling, it snaps to the bottom.

[UncleJ4ck]

fixed. the scroll offset only lived in a one-shot pass over the content layer's children, so when a content element scheduled its own reflow the window relaid that layer with no offset and it jumped to the top while currentScroll stayed at the bottom. moved the offset onto the layer's own position (setAbsolutePosition) so any reposition keeps it scrolled. regression test added, pushed to the same branch.

[vaxerski]

thanks. All good now. In the future, if you wanna work on it a bit more:

• smooth scrolling for continuous scroll sources
• scrollbar hiding (fading out) on no movement automatically
• we seem to have a similar problem to the one in the polkit agent where text flickers between "hello" and "hel..." constantly in the spinbox.