Avoids releasing capture prematurily when dragging tabs from one to another browser window

chrome/browser/ui/views/tabs/tab_drag_controller.cc

Index: chrome/browser/ui/views/tabs/tab_drag_controller.cc
diff --git a/chrome/browser/ui/views/tabs/tab_drag_controller.cc b/chrome/browser/ui/views/tabs/tab_drag_controller.cc
index 741a77fa8638741377a44a8b4dbde868e5839354..a2f456df427aa5f1d337723d93be823fed21578a 100644
--- a/chrome/browser/ui/views/tabs/tab_drag_controller.cc
+++ b/chrome/browser/ui/views/tabs/tab_drag_controller.cc
@@ -694,9 +694,12 @@ TabDragController::DragBrowserToNewTabStrip(
     // that to effect the position of any windows.
     SetWindowPositionManaged(browser_widget->GetNativeView(), false);
 
+#if !defined(OS_LINUX)

[Elliot Glaysher, 2014/02/26 20:33:17]

Isn't this code used on chromeos, too?

[varkha, 2014/02/26 20:59:35]

It is, but I did not see any visual impact from skipping the call to Hide() on
Chrome OS. I suspect that the reason for having it here was not relevant on
Chrome OS.

     // EndMoveLoop is going to snap the window back to its original location.
-    // Hide it so users don't see this.
+    // Hide it so users don't see this. Hiding a window in Linux aura causes
+    // it to lose capture so skip it.
     browser_widget->Hide();

[varkha, 2014/02/26 20:30:18]

Not calling Hide on Linux seems to be lesser evil (not even sure if the concern
was valid on Linux).

+#endif
     browser_widget->EndMoveLoop();
 
     // Ideally we would always swap the tabs now, but on non-ash it seems that
@@ -1211,15 +1214,15 @@ void TabDragController::RunMoveLoop(const gfx::Vector2d& drag_offset) {
   if (end_run_loop_behavior_ == END_RUN_LOOP_CONTINUE_DRAGGING) {
     end_run_loop_behavior_ = END_RUN_LOOP_STOP_DRAGGING;
     if (tab_strip_to_attach_to_after_exit_) {
+      tab_strip_to_attach_to_after_exit_->GetWidget()->Activate();

[varkha, 2014/02/26 20:30:18]

Activating before calling Detach (which may close the browser frame) helps to
not lose capture prematurely.

[Elliot Glaysher, 2014/02/26 20:33:17]

What's the effect on Windows here?

[varkha, 2014/02/26 20:59:35]

I didn't have a chance to test this on Windows yet (TBD). I hope that this is
benign since I am moving it earlier.

+      // Activate may trigger a focus loss, destroying us.
+      if (!ref)
+        return;
       gfx::Point point_in_screen(GetCursorScreenPoint());
       Detach(DONT_RELEASE_CAPTURE);
       Attach(tab_strip_to_attach_to_after_exit_, point_in_screen);
       // Move the tabs into position.
       MoveAttached(point_in_screen);
-      attached_tabstrip_->GetWidget()->Activate();
-      // Activate may trigger a focus loss, destroying us.
-      if (!ref)
-        return;
       tab_strip_to_attach_to_after_exit_ = NULL;
     }
     DCHECK(attached_tabstrip_);