[PATCH v4 0/2] MR7157: winex11: Avoid setting RDW_ERASE on expose events.

[Rémi Bernon (＠rbernon)]

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=57665

---

The issue is that The Medium launcher uses a dialog window procedure, and implements its background by doing a StretchBlt call on WM_PAINT in the dialog procedure call, which happens before the window message loop.

It then itself calls InvalidateRect(hwnd, NULL, 0), which queues a WM_PAINT as well but with only the RDW_INVALIDATE flag.

Next, when the window message loop is executed, the WM_PAINT message is being processed as it should, but as we've invalidate the window with RDW_ERASE ourselves from the expose event, the WM_NCPAINT handler erases the entire window, clearing the pixels that the launcher has just painted.

This regressed since the window surface clipping region logic changed, as the expose event handling was previously not calling `NtUserRedrawWindow` on windows with a surface and without a clipping region. The clipping region was also previously not always set, or set later, and we're setting it more consistently since 51b16963f6e0e8df43118deac63f640aee4698b7, even when it matches the window rect, for compatibility with some old wineandroid logic (where I believe it was used to clip window surfaces to their proper sizes).

--
  v4: winex11: Avoid setting RDW_ERASE on expose events.
      explorer: Paint the desktop even without RDW_ERASE.

https://gitlab.winehq.org/wine/wine/-/merge_requests/7157

[PATCH v4 1/2] explorer: Paint the desktop even without RDW_ERASE.

[Rémi Bernon]

From: Rémi Bernon

---
 programs/explorer/desktop.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/programs/explorer/desktop.c b/programs/explorer/desktop.c
index e0b57a4b37f..ce70c5dd79d 100644
--- a/programs/explorer/desktop.c
+++ b/programs/explorer/desktop.c
@@ -842,7 +842,7 @@ static LRESULT WINAPI desktop_wnd_proc( HWND hwnd, UINT message, WPARAM wp, LPAR
             BeginPaint( hwnd, &ps );
             if (!using_root)
             {
-                if (ps.fErase) PaintDesktop( ps.hdc );
+                PaintDesktop( ps.hdc );
                 draw_launchers( ps.hdc, ps.rcPaint );
             }
             EndPaint( hwnd, &ps );

[PATCH v4 2/2] winex11: Avoid setting RDW_ERASE on expose events.

[Rémi Bernon]

From: Rémi Bernon

The Medium launcher uses a dialog window procedure, and implements its
background by doing a StretchBlt call on WM_PAINT in the dialog
procedure call, which happens before the window message loop

It then itself calls InvalidateRect(hwnd, NULL, 0), which queues a
WM_PAINT as well but with only the RDW_INVALIDATE flag.

Next, when the window message loop is executed, the WM_PAINT message is
being processed as it should, but as we've invalidated the window with
RDW_ERASE from the expose event, the WM_NCPAINT handler erases the
entire window, clearing the pixels that the launcher has just painted.

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=57665
---
 dlls/winex11.drv/event.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/dlls/winex11.drv/event.c b/dlls/winex11.drv/event.c
index f9aa4b3f66b..cc58efed502 100644
--- a/dlls/winex11.drv/event.c
+++ b/dlls/winex11.drv/event.c
@@ -978,7 +978,7 @@ static BOOL X11DRV_Expose( HWND hwnd, XEvent *xev )
     RECT rect, abs_rect;
     POINT pos;
     struct x11drv_win_data *data;
-    UINT flags = RDW_INVALIDATE | RDW_ERASE | RDW_FRAME | RDW_ALLCHILDREN;
+    UINT flags = RDW_INVALIDATE | RDW_FRAME | RDW_ALLCHILDREN;

TRACE( "win %p (%lx) %d,%d %dx%d\n",
            hwnd, event->window, event->x, event->y, event->width, event->height );

[Rémi Bernon (＠rbernon)]

Rebased the changes, is there something wrong with this?

[Paul Gofman (＠gofman)]

FWIW it looks like now always-called RedrawWindow() with RDW_ALLCHILDREN breaks more than that, specifically Fallout New Vegas launcher (which is a dialog window with bogus static controls config) now starts without the text in static controls visible, due to those being overwritten with bigger control with background. This behaviour of the dialog is Windows nehaviour as well (redrawing it with RDW_ALLCHILDREN | RDW_INVALIDATE) on Windows yields the same result, but Windows doesn't call such a RedrawWindow window for it (and so didn't Wine before the mentioned changes). I am now trying to come up with some patch which would avoid such a redraw (it preliminary seems to me that it is coming from expose_window_surface_rect logic which takes into account surface rect which is bigger than actual window dimensions), but if you have some immediate ideas...