The splash covers the page and swallows raw mouse presses until it has faded
out. Only one spec waited for it, so the tool window resize test sometimes
pressed on the splash instead of the gap (3 of 15 runs locally). The workbench
fixture now waits for it to be removed.
Moving the pointer diagonally from a menu item toward its open submenu crossed sibling items, whose hover timers switched or closed the submenu.
While a submenu is open, a sibling hover is deferred when the pointer is inside the triangle from its previous position to the submenu's near corners, including flipped submenus. Stopping on a sibling still switches after 300 ms, moving away keeps the normal delays, and entering the submenu cancels pending switches. This lives in `MenuList`, so every menu and nested level gets it.
Fixes#2128
On macOS, with an editable element focused and an input method active, WebKit can deliver a tap as pointerup then pointerdown, so drag trackers saw every later pointermove as a held button: one tap plus a pointer move started a phantom drag (file drag label, tab detach, tool window move) and the next tap dropped it. A drag now starts only from a move with the primary button down, a move without it cancels the press, and late button-less moves after a drag started are ignored so the drop is kept. Resizers, dialog drag/resize, pan and the review gutter end instead; the Project view also listens for pointercancel.
Refs #760
Co-authored-by: ntbang0901 <121003585+ntbang0901@users.noreply.github.com>
l0ng-ai
ntbang0901
Walter WangandWalt von Neumann