[Bug 57199] New: window_surface not always destroyed

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=57199

Bug ID: 57199
           Summary: window_surface not always destroyed
           Product: Wine
           Version: unspecified
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: win32u
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

Created attachment 77093
  --> https://bugs.winehq.org/attachment.cgi?id=77093
DPI unaware app with winex11 not destroying window surfaces

With the latest Wine master (tested 03d985369dc9afe943ee3072478a2567b3fe4077,
but haven't run a bisection, so can't say when this was introduced), there are
scenarios where the previous window_surface for a HWND is not destroyed when a
new one is created for that same HWND (e.g., resizing). This creates memory
leaks which have different effects depending on the driver (e.g., winewayland
eventually runs out of SHM memory).

I am seeing this issue with applications that are not DPI aware when the system
dpi != 96. I am attaching two logs, one for winex11 (under xwayland) and one
for winewayland, from the same application (Crayon Physics Deluxe) while
interactively resizing the window, in a prefix with dpi=192 configured
(matching the 2x monitor scale). Note the absence of any window surface
destruction calls in the logs, whereas multiple new window surfaces (for the
same window) are being created during the resize.

I have slightly changed the {WAYLAND,X11DRV}_CreateWindowSurface TRACE() so
that it prints `*surface` (instead of `surface`) in the logs.