[fix] align remote mouse coordinates with the displayed video area

apps/desktop/src/gui/rendering/slint_video_presenter.cpp

@@ -154,10 +154,8 @@ struct SlintVideoPresenter::Impl {
             const auto window_size = (*stream)->window().size();
             const float scale_factor =
                 std::max(1.0f, (*stream)->window().scale_factor());
-            const int top_inset =
-                !surface.fullscreen && surface.tab_count > 1
-                    ? static_cast<int>(std::lround(30.0f * scale_factor))
-                    : 0;
+            const int top_inset = static_cast<int>(std::lround(
+                (*stream)->get_video_area_top() * scale_factor));
             const bool rounded_window = (*stream)->get_custom_titlebar() &&
                                         !surface.fullscreen &&
                                         !(*stream)->window().is_maximized();
@@ -335,8 +333,7 @@ SlintVideoPresenter::PresentResult SlintVideoPresenter::Present(
       dynamic_cast<MacMetalVideoRenderer*>(&impl_->renderer);
   if (metal_renderer && metal_renderer->IsActive()) {
     const SurfaceState surface = impl_->CurrentState();
-    const double top_inset =
-        !surface.fullscreen && surface.tab_count > 1 ? 30.0 : 0.0;
+    const double top_inset = (*impl_->stream)->get_video_area_top();
     const auto outcome = metal_renderer->RenderLatest(
         frame.remote_id, top_inset, !surface.fullscreen);
     if (outcome.result == VideoRenderer::RenderResult::rendered) {

apps/desktop/src/gui/rendering/slint_video_presenter.h

@@ -28,7 +28,6 @@ public:
   struct SurfaceState {
     std::string selected_stream;
     bool fullscreen = false;
-    size_t tab_count = 0;
   };
 
   struct Frame {
@@ -78,4 +77,4 @@ private:
 
 }  // namespace crossdesk
 
-#endif
+#endif

apps/desktop/src/gui/rendering/video_geometry.h

@@ -0,0 +1,51 @@
+/*
+ * @Author: DI JUNKUN
+ * @Date: 2026-09-22
+ * Copyright (c) 2026 by DI JUNKUN, All Rights Reserved.
+ */
+
+#ifndef _VIDEO_GEOMETRY_H_
+#define _VIDEO_GEOMETRY_H_
+
+#include <algorithm>
+#include <cmath>
+#include <optional>
+
+namespace crossdesk {
+
+struct VideoPoint {
+  float x;
+  float y;
+};
+
+// Area dimensions and pointer positions must use the same coordinate space.
+// Slint supplies both in logical pixels; remote input is resolution
+// independent.
+inline std::optional<VideoPoint> MapVideoPoint(float x, float y,
+                                               float area_width,
+                                               float area_height,
+                                               int video_width,
+                                               int video_height) {
+  if (!std::isfinite(x) || !std::isfinite(y) || !std::isfinite(area_width) ||
+      !std::isfinite(area_height) || area_width <= 0 || area_height <= 0 ||
+      video_width <= 0 || video_height <= 0) {
+    return std::nullopt;
+  }
+
+  const double scale =
+      std::min(static_cast<double>(area_width) / video_width,
+               static_cast<double>(area_height) / video_height);
+  const double width = video_width * scale;
+  const double height = video_height * scale;
+  const double left = (area_width - width) * 0.5;
+  const double top = (area_height - height) * 0.5;
+  if (x < left || x > left + width || y < top || y > top + height) {
+    return std::nullopt;
+  }
+  return VideoPoint{static_cast<float>((x - left) / width),
+                    static_cast<float>((y - top) / height)};
+}
+
+}  // namespace crossdesk
+
+#endif