On Thu Dec 18 10:57:54 2025 +0000, Rémi Bernon wrote:
This cannot work based only on the SWP flags signal edges. There are various cases where the host window needs to be kept unmapped until it can actually be shown (offscreen, layered windows, ...) and we then need to check the Win32 window state to decide too. Maybe the whole thing could be changed and maybe tracking SWP flags to latch a visibility flag somewhere could be better, but this doesn't seem like a reasonable change during regression fixes. I also definitely don't know and wouldn't bet on any guarantee of what can actually be done with user32. If it's a regression then probably a revert is an appropriate solution, not a pile of other hacks on top.
-- https://gitlab.winehq.org/wine/wine/-/merge_requests/9793#note_125794