[PATCH 0/1] MR10598: winex11: Use unsigned long for monitor indices generation.

[Rémi Bernon (＠rbernon)]

Avoiding invisible padding in the structure that could make memcmp
to constantly fail when comparing old / new values.

-- 
https://gitlab.winehq.org/wine/wine/-/merge_requests/10598

[PATCH 1/1] winex11: Use unsigned long for monitor indices generation.

[Rémi Bernon]

From: Rémi Bernon <[e-mail]>

Avoiding invisible padding in the structure that could make memcmp
to constantly fail when comparing old / new values.
---
 dlls/winex11.drv/window.c   | 4 ++--
 dlls/winex11.drv/x11drv.h   | 4 ++--
 dlls/winex11.drv/xinerama.c | 2 +-
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/dlls/winex11.drv/window.c b/dlls/winex11.drv/window.c
index 33d8ceee438..a83359b2d74 100644
--- a/dlls/winex11.drv/window.c
+++ b/dlls/winex11.drv/window.c
@@ -1318,7 +1318,7 @@ static void window_set_net_wm_fullscreen_monitors( struct x11drv_win_data *data,
  * windows spanning multiple monitors */
 static void update_net_wm_fullscreen_monitors( struct x11drv_win_data *data )
 {
-    struct monitor_indices monitors;
+    struct monitor_indices monitors = {0};
 
     /* If the current display device handler cannot detect dynamic device changes, do not use
      * _NET_WM_FULLSCREEN_MONITORS because xinerama_get_fullscreen_monitors() may report wrong
@@ -1736,7 +1736,7 @@ static UINT window_update_client_config( struct x11drv_win_data *data )
 {
     static const UINT fullscreen_mask = (1 << NET_WM_STATE_MAXIMIZED) | (1 << NET_WM_STATE_FULLSCREEN);
     RECT rect, old_rect = data->rects.window, new_rect;
-    unsigned int old_generation, generation;
+    unsigned long old_generation, generation;
     long old_monitors[4], monitors[4];
     UINT flags;
 
diff --git a/dlls/winex11.drv/x11drv.h b/dlls/winex11.drv/x11drv.h
index 9f38ceba0df..e1fe1b55620 100644
--- a/dlls/winex11.drv/x11drv.h
+++ b/dlls/winex11.drv/x11drv.h
@@ -635,7 +635,7 @@ enum x11drv_net_wm_state
 
 struct monitor_indices
 {
-    unsigned int generation;
+    unsigned long generation;
     long indices[4];
 };
 
@@ -763,7 +763,7 @@ extern POINT virtual_screen_to_root( INT x, INT y );
 extern POINT root_to_virtual_screen( INT x, INT y );
 extern RECT get_host_primary_monitor_rect(void);
 extern RECT get_work_area( const RECT *monitor_rect );
-extern BOOL xinerama_get_fullscreen_monitors( const RECT *rect, unsigned int *generation, long *indices );
+extern BOOL xinerama_get_fullscreen_monitors( const RECT *rect, unsigned long *generation, long *indices );
 extern void xinerama_init( unsigned int width, unsigned int height );
 extern void init_recursive_mutex( pthread_mutex_t *mutex );
 extern void init_icm_profile(void);
diff --git a/dlls/winex11.drv/xinerama.c b/dlls/winex11.drv/xinerama.c
index 81f739285e0..d6478fcd03a 100644
--- a/dlls/winex11.drv/xinerama.c
+++ b/dlls/winex11.drv/xinerama.c
@@ -126,7 +126,7 @@ static inline int query_screens(void)
 
 /* Get xinerama monitor indices required for _NET_WM_FULLSCREEN_MONITORS. Return FALSE if rect is
  * not fullscreen */
-BOOL xinerama_get_fullscreen_monitors( const RECT *rect, unsigned int *generation, long *indices )
+BOOL xinerama_get_fullscreen_monitors( const RECT *rect, unsigned long *generation, long *indices )
 {
     RECT window_rect, intersected_rect, monitor_rect;
     BOOL ret = FALSE;
-- 
GitLab

https://gitlab.winehq.org/wine/wine/-/merge_requests/10598

[Zhiyi Zhang (＠zhiyi)]

Let's compare `generation` separately and only use memcmp() for `indices`. Changing `generation` to `unsigned long` works, but someday someone might add a new field and the comparison will break silently again.

-- 
https://gitlab.winehq.org/wine/wine/-/merge_requests/10598#note_135456

[Rémi Bernon (＠rbernon)]

On Wed Apr  8 10:41:11 2026 +0000, Zhiyi Zhang wrote:
> Let's compare `generation` separately and only use memcmp() for
> `indices`. Changing `generation` to `unsigned long` works, but someday
> someone might add a new field and the comparison will break silently again.
> It's never going to be fully future proof, adding new fields we might also forget to add them to the comparison. Other state structures use memcmp directly, and I think it is better to use a common pattern. We'll just need to be mindful of alignment, which could be indicated with a comment.

-- 
https://gitlab.winehq.org/wine/wine/-/merge_requests/10598#note_135474