Fix the crash that could occur when the window is closed while web contents drag is in progress.

Fix the crash that could occur when the window is closed while web contents drag is in progress.

We now check if the contents view window is gone after system drag-and-drop returns. Bail out immediately if so. With this fix, we do not need the delay-close-at-drag workaround any more.

This fix applies towards to both Windows and Aura. We do not have problem on GTK since it does not run nested message loop.

On Mac, the system drag-and-drop is invoked from the Cocoa view that could outlive the WebContentsViewMac due to Cocoa retains count (see comment in WebContentsViewMac::~WebContentsViewMac). However, we do need to ensure that WebDragSource stops using web contents when it is gone.

BUG=145363
TEST=Manual test by following the steps in the bug

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=159366

[jianli, 2012-09-20 23:54:22]

Ben, could you please review both Windows and Aura changes?

Tony, could you please review Mac changes?

Darin, could you please approve it as content owner?

[tony, 2012-09-21 21:11:11]

Avi would be a better Mac reviewer than me.

http://codereview.chromium.org/10966023/diff/1/content/browser/web_contents/w...
File content/browser/web_contents/web_contents_drag_win.cc (right):

http://codereview.chromium.org/10966023/diff/1/content/browser/web_contents/w...
content/browser/web_contents/web_contents_drag_win.cc:209: // instance is still
alive bacuse the task holds a reference count to it.
Nit: bacuse -> because

[Ben Goodger (Google), 2012-09-27 16:05:20]

I am OK with this if tony (who I think wrote this stuff originally) is.

LGTM

http://codereview.chromium.org/10966023/diff/11001/content/browser/web_conten...
File content/browser/web_contents/web_contents_drag_win.h (right):

http://codereview.chromium.org/10966023/diff/11001/content/browser/web_conten...
content/browser/web_contents/web_contents_drag_win.h:71: // This could happens
when the window gets destroyed when the drag is still in
happen

http://codereview.chromium.org/10966023/diff/11001/content/browser/web_conten...
content/browser/web_contents/web_contents_drag_win.h:73: // because the instance
has been disposed.
destroyed.

[tony, 2012-09-27 20:41:04]

LGTM.

It would be nice if we could come up with some way to test this since it seems
subtle, but that would be a pretty big task.

[jianli, 2012-09-28 17:52:52]

Avi or Jam, could one of you approve the patch as the content owner? Thanks.

[jam, 2012-09-28 19:20:51]

On 2012/09/28 17:52:52, jianli wrote:
> Avi or Jam, could one of you approve the patch as the content owner? Thanks.

ben is an owner in content/browser/web_contents. lgtm for the rest

[jam, 2012-09-28 19:21:01]

https://codereview.chromium.org/10966023/diff/16001/content/browser/web_conte...
File content/browser/web_contents/web_contents_view_aura.cc (right):

https://codereview.chromium.org/10966023/diff/16001/content/browser/web_conte...
content/browser/web_contents/web_contents_view_aura.cc:101: if
(content::NOTIFICATION_WEB_CONTENTS_DISCONNECTED != type)
nit: type != content::...

[commit-bot: I haz the power, 2012-09-28 20:24:57]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/jianli@chromium.org/10966023/12020

[commit-bot: I haz the power, 2012-09-28 23:55:25]

Change committed as 159366