SDL_CreateRenderer segfaults on windows with dummy videodriver after updating to SDL 2.24.0

[ankith26]

Till 2.0.22 SDL_CreateRenderer didn't segfault in the same configuration. This is a bug noticed downstream in the pygame unit tests running on CI (hence the dummy driver)

This bug is not reproducible locally when using a non-dummy driver, or using the dummy driver and setting SDL_RENDER_DRIVER to software.

(I know this is a bit of a weird usecase, considering that it does not make sense to be using a renderer when there's no actual window, but I just reported the bug because it shouldn't segfault)

[ankith26]

I tested this locally by compiling debug SDL binaries. This is fixed on main (probably) as a side affect of #6112 (IDK whether it was indented or not) but is still broken on the 2.24.x branch at the time of writing of this comment. This regression (probably) came from #5778

I'd appreciate it if a fix (if it's not much of a hassle) was backported to the 2.24.x branch so we can get to using that earlier.

Anyways, here is the stack trace (using VSC). The internal function at the root of this segfault was entirely replaced in that PR

[sezero]

So, the following for release-2.24.x branch, then?

diff --git a/src/video/windows/SDL_windowswindow.c b/src/video/windows/SDL_windowswindow.c
index eb536bb..3fd3290 100644
--- a/src/video/windows/SDL_windowswindow.c
+++ b/src/video/windows/SDL_windowswindow.c
@@ -1408,10 +1408,16 @@ WIN_SetWindowOpacity(_THIS, SDL_Window * window, float opacity)
 void
 WIN_GetDrawableSize(const SDL_Window *window, int *w, int *h)
 {
-    const SDL_WindowData *data = ((SDL_WindowData *)window->driverdata);
-    HWND hwnd = data->hwnd;
+    HWND hwnd;
     RECT rect;
 
+    if (!window || !(window->driverdata)) {
+        *w = 0;
+        *h = 0;
+        return;
+    }
+
+    hwnd = ((SDL_WindowData *)window->driverdata)->hwnd;
     if (GetClientRect(hwnd, &rect)) {
         *w = rect.right;
         *h = rect.bottom;

[ankith26]

Yes, I think this should do the trick. I'm not familiar with SDL internals so I can't say for sure whether this is the only thing that needs updating or whether there are more places like this that need fixing for this issue

[sezero]

@slouken: Is the patch above good for release-2.24.x branch?
Can there be any other places that need touching?

[slouken]

Yes, that's fine. I don't know of any other places that would need touching.

[sezero]

OK, the patch is in the release-2.24.x branch.