4 files changed, 33 insertions, 1 deletions

diff --git a/video/out/vo.h b/video/out/vo.h
index 1b7a239798..5deb99a6f1 100644
--- a/video/out/vo.h
+++ b/video/out/vo.h
@@ -48,10 +48,11 @@ enum {
     // Special thing for encode mode (vo_driver.initially_blocked).
     // Part of VO_EVENTS_USER to make vo_is_ready_for_frame() work properly.
     VO_EVENT_INITIAL_UNBLOCK            = 1 << 7,
+    VO_EVENT_FOCUS                      = 1 << 8,
 
     // Set of events the player core may be interested in.
     VO_EVENTS_USER = VO_EVENT_RESIZE | VO_EVENT_WIN_STATE | VO_EVENT_DPI |
-                     VO_EVENT_INITIAL_UNBLOCK,
+                     VO_EVENT_INITIAL_UNBLOCK | VO_EVENT_FOCUS,
 };
 
 enum mp_voctrl {
@@ -99,6 +100,8 @@ enum mp_voctrl {
     VOCTRL_GET_UNFS_WINDOW_SIZE,        // int[2] (w/h)
     VOCTRL_SET_UNFS_WINDOW_SIZE,        // int[2] (w/h)
 
+    VOCTRL_GET_FOCUSED,                 // bool*
+
     // char *** (NULL terminated array compatible with CONF_TYPE_STRING_LIST)
     // names for displays the window is on
     VOCTRL_GET_DISPLAY_NAMES,
diff --git a/video/out/wayland_common.c b/video/out/wayland_common.c
index a79c8aa8ca..c6c0e342c9 100644
--- a/video/out/wayland_common.c
+++ b/video/out/wayland_common.c
@@ -447,11 +447,15 @@ static void keyboard_handle_enter(void *data, struct wl_keyboard *wl_keyboard,
                                   uint32_t serial, struct wl_surface *surface,
                                   struct wl_array *keys)
 {
+    struct vo_wayland_state *wl = data;
+    wl->has_keyboard_input = true;
 }
 
 static void keyboard_handle_leave(void *data, struct wl_keyboard *wl_keyboard,
                                   uint32_t serial, struct wl_surface *surface)
 {
+    struct vo_wayland_state *wl = data;
+    wl->has_keyboard_input = false;
 }
 
 static bool create_input(struct vo_wayland_state *wl)
@@ -951,6 +955,7 @@ static void handle_toplevel_config(void *data, struct xdg_toplevel *toplevel,
 
     bool is_maximized = false;
     bool is_fullscreen = false;
+    bool is_activated = false;
     enum xdg_toplevel_state *state;
     wl_array_for_each(state, states) {
         switch (*state) {
@@ -961,6 +966,7 @@ static void handle_toplevel_config(void *data, struct xdg_toplevel *toplevel,
             wl->pending_vo_events |= VO_EVENT_LIVE_RESIZING;
             break;
         case XDG_TOPLEVEL_STATE_ACTIVATED:
+            is_activated = true;
             /*
              * If we get an ACTIVATED state, we know it cannot be
              * minimized, but it may not have been minimized
@@ -992,6 +998,16 @@ static void handle_toplevel_config(void *data, struct xdg_toplevel *toplevel,
         m_config_cache_write_opt(wl->vo_opts_cache, &vo_opts->window_maximized);
     }
 
+    if (wl->activated != is_activated) {
+        wl->activated = is_activated;
+        if ((!wl->focused && wl->activated && wl->has_keyboard_input) ||
+            (wl->focused && !wl->activated))
+        {
+            wl->focused = !wl->focused;
+            wl->pending_vo_events |= VO_EVENT_FOCUS;
+        }
+    }
+
     if (!(wl->pending_vo_events & VO_EVENT_LIVE_RESIZING))
         vo_query_and_reset_events(wl->vo, VO_EVENT_LIVE_RESIZING);
 
@@ -1544,6 +1560,10 @@ int vo_wayland_control(struct vo *vo, int *events, int request, void *arg)
         }
         return VO_TRUE;
     }
+    case VOCTRL_GET_FOCUSED: {
+        *(bool *)arg = wl->focused;
+        return VO_TRUE;
+    }
     case VOCTRL_GET_DISPLAY_NAMES: {
         *(char ***)arg = get_displays_spanned(wl);
         return VO_TRUE;
diff --git a/video/out/wayland_common.h b/video/out/wayland_common.h
index 65f7cbf3e5..7eec7be2d5 100644
--- a/video/out/wayland_common.h
+++ b/video/out/wayland_common.h
@@ -76,6 +76,9 @@ struct vo_wayland_state {
     bool frame_wait;
     bool state_change;
     bool toplevel_configured;
+    bool activated;
+    bool has_keyboard_input;
+    bool focused;
     int wakeup_pipe[2];
     int pending_vo_events;
     int mouse_x;
diff --git a/video/out/x11_common.c b/video/out/x11_common.c
index 461ff378d0..92b3539aca 100644
--- a/video/out/x11_common.c
+++ b/video/out/x11_common.c
@@ -1128,11 +1128,13 @@ void vo_x11_check_events(struct vo *vo)
         case FocusIn:
             x11->has_focus = true;
             vo_update_cursor(vo);
+            x11->pending_vo_events |= VO_EVENT_FOCUS;
             break;
         case FocusOut:
             release_all_keys(vo);
             x11->has_focus = false;
             vo_update_cursor(vo);
+            x11->pending_vo_events |= VO_EVENT_FOCUS;
             break;
         case KeyRelease:
             release_all_keys(vo);
@@ -1878,6 +1880,10 @@ int vo_x11_control(struct vo *vo, int *events, int request, void *arg)
         }
         return VO_TRUE;
     }
+    case VOCTRL_GET_FOCUSED: {
+        *(bool *)arg = x11->has_focus;
+        return VO_TRUE;
+    }
     case VOCTRL_GET_DISPLAY_NAMES: {
         if (!x11->pseudo_mapped)
             return VO_FALSE;