On Tue Jun 18 13:20:27 2024 +0000, Emil Velikov wrote:
AFAICT this (+ equivalent for the other drivers) changes the functionality, even though I'm not sure how much/if that matters. Perhaps it's worth moving these checks to the WindowPosChanging callback, as a prep commit just incase? On unrelated note: the missing else + memcmp on Android seem a bit odd.
Right, I had some code in win32u to get the previous surface if its rect was large enough, similarly to what the drivers are doing, but then re-ordered it later for some reason.
I'm not completely sure if hidden windows are supposed to have a surface, I suspect not but maybe they do once they had one. In any case the current behavior was only keeping a surface for hidden windows until it needed to be resized. Also, winewayland behaves differently and checked for visibility upfront.