sadpandajoe commented on PR #41351: URL: https://github.com/apache/superset/pull/41351#issuecomment-4999653879
Closing this one. The PR targeted a latch in the overflow recalculation: when the item set changes, the reset path zeroes the overflowed items, so the existing `index === -1 && overflowedItems.length > 0` guard can't fire on the pass that follows, and a transient "everything fits" measurement could settle to `-1` with no dependency change left to correct it. That path is real in the code, but for it to bite, `getBoundingClientRect()` has to report a container wider than its settled width. gBCR forces synchronous layout, so it returns settled geometry rather than mid-reflow geometry — and the resize detector on the container plus the `ResizeObserver` on the children already re-run the recalculation whenever geometry genuinely changes. The behaviour that prompted this never reproduced on master, and #38193 already covers the flicker case it was traced to. The regression suite here doesn't close that gap either. Against master's component it fails only where it forces `getBoundingClientRect` to return a transient value, or where it asserts on the `requestAnimationFrame` machinery this PR itself introduces; the two tests that measure honestly pass on master unchanged. So it demonstrates that master doesn't recover from a synthetic transient, not that there's a browser-reproducible defect. On that basis the added ref/rAF coordination isn't carrying its weight in a shared core component. Happy to revisit if a reproduction on master turns up. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
