[PATCH v2 0/2] MR10627: winemac.drv, wintab32: Some fixes for Cocoa-initiated quits

[Tim Clem (＠tclem)]

The EA app was refusing to exit when quit via Cocoa for two reasons:

- the wintab32 window was responding 0 to WM_QUERYENDSESSION
- the app itself closes a window in response to WM_QUERYENDSESSION, and we weren't expecting the corresponding WM_ENDSESSION to fail

--
  v2: winemac.drv: Handle windows becoming invalid between WM_QUERYENDSESSION and WM_ENDSESSION.
 
https://gitlab.winehq.org/wine/wine/-/merge_requests/10627

[PATCH v2 1/2] wintab32: Respond true to WM_QUERYENDSESSION.

[Tim Clem]

From: Tim Clem <[e-mail]>

Otherwise the WineTabletClass window prevents quitting via Cocoa from
macdrv.
---
 dlls/wintab32/wintab32.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/dlls/wintab32/wintab32.c b/dlls/wintab32/wintab32.c
index 15040b9ca9c..8a2af241f74 100644
--- a/dlls/wintab32/wintab32.c
+++ b/dlls/wintab32/wintab32.c
@@ -102,6 +102,7 @@ static LRESULT WINAPI TABLET_WindowProc(HWND hwnd, UINT uMsg, WPARAM wParam,
     switch(uMsg)
     {
         case WM_NCCREATE:
+        case WM_QUERYENDSESSION:
             return TRUE;
 
         case WT_PACKET:
-- 
GitLab


https://gitlab.winehq.org/wine/wine/-/merge_requests/10627

[PATCH v2 2/2] winemac.drv: Handle windows becoming invalid between WM_QUERYENDSESSION and WM_ENDSESSION.

[Tim Clem]

From: Tim Clem <[e-mail]>

We already handle invalid windows at QUERYENDSESSION time in
macdrv_app_quit_request, but some apps (e.g. the EA launcher) close
windows in response to QUERYENDSESSION, and we shouldn't abort
quitting in that case.
---
 dlls/winemac.drv/dllmain.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)

diff --git a/dlls/winemac.drv/dllmain.c b/dlls/winemac.drv/dllmain.c
index eb6efedab17..bbd528d53b9 100644
--- a/dlls/winemac.drv/dllmain.c
+++ b/dlls/winemac.drv/dllmain.c
@@ -141,9 +141,15 @@ static void CALLBACK quit_callback(HWND hwnd, UINT msg, ULONG_PTR data, LRESULT
                 if (!SendMessageCallbackW(qi->wins[i], WM_ENDSESSION, qi->result, qi->flags,
                                           quit_callback, (ULONG_PTR)qi))
                 {
-                    WARN("failed to send WM_ENDSESSION to win %p; error 0x%08lx\n",
-                         qi->wins[i], RtlGetLastWin32Error());
-                    quit_callback(qi->wins[i], WM_ENDSESSION, (ULONG_PTR)qi, 0);
+                    DWORD error = RtlGetLastWin32Error();
+                    BOOL invalid = (error == ERROR_INVALID_WINDOW_HANDLE);
+                    if (invalid)
+                        TRACE("failed to send WM_ENDSESSION to win %p because it's invalid; assuming success\n",
+                            qi->wins[i]);
+                    else
+                        WARN("failed to send WM_ENDSESSION to win %p; error 0x%08lx; assuming refusal\n",
+                            qi->wins[i], error);
+                    quit_callback(qi->wins[i], WM_ENDSESSION, (ULONG_PTR)qi, invalid);
                 }
             }
         }
-- 
GitLab

https://gitlab.winehq.org/wine/wine/-/merge_requests/10627

[Tim Clem (＠tclem)]

Alternatively perhaps the wintab window should become message-only so it doesn't get enumerated with EnumWindows and there aren't repercussions from it not calling DefWindowProc?

-- 
https://gitlab.winehq.org/wine/wine/-/merge_requests/10627#note_135972

[Brendan Shanks (＠bshanks)]

The winemac change looks good.

For wintab it does seem like that should be a message-only window, as far as I can tell it's only used to receive messages from winex11. Also, maybe the window class registration and creation can be moved to `LoadTablet()` instead of `DllMain()` (I assume the EA App links to wintab but doesn't actually call anything from it?)

-- 
https://gitlab.winehq.org/wine/wine/-/merge_requests/10627#note_136228