Implement BrowserList::AllBrowsersClosed for Linux.

Implement BrowserList::AllBrowsersClosed for Linux.

This fixes leaks caused by not calling gtk_widget_destroy on
dialog windows when we exit.

While I'm here, fix a GList leak on Linux toolkit views.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=27130

[mattm, 2009-09-23 23:40:02]

Are there any suppressions that can be removed with this?

http://codereview.chromium.org/223015/diff/1/2
File chrome/browser/browser_list_gtk.cc (right):

http://codereview.chromium.org/223015/diff/1/2#newcode14
Line 14: if (GTK_WIDGET_VISIBLE(GTK_WIDGET(window)))
Why is it only destroyed if it's visible?

[tony, 2009-09-24 00:24:23]

I don't think there are suppressions this fixes since I don't think the tests
open any dialogs.

http://codereview.chromium.org/223015/diff/1/2
File chrome/browser/browser_list_gtk.cc (right):

http://codereview.chromium.org/223015/diff/1/2#newcode14
Line 14: if (GTK_WIDGET_VISIBLE(GTK_WIDGET(window)))
On 2009/09/23 23:40:02, mattm wrote:
> Why is it only destroyed if it's visible?

There appear to be top level windows that we didn't explicitly create that hang
around.  For example, all the drop down menus on the options page seem to have a
top level window associated with them and the tooltip has a top level window
associated with it.  I'm not 100% sure that VISIBLE is the right check to use,
but it seems that anything we gtk_widget_show should be torn down.  If I destroy
everything, I get "assertion `GTK_IS_WIDGET (widget)' failed".

[tony, 2009-09-24 18:04:19]

I also tried filtering by GdkWindowTypeHint and only destroying
GDK_WINDOW_TYPE_HINT_NORMAL and GDK_WINDOW_TYPE_HINT_DIALOG windows, but I still
got a crash.

[Evan Stade, 2009-09-24 18:10:28]

what if you use the GTK_IS_WIDGET check?

[tony, 2009-09-24 19:01:18]

On 2009/09/24 18:10:28, Evan Stade wrote:
> what if you use the GTK_IS_WIDGET check?

Still crashes.  I thought GTK_WIDGET already did that check anyway.

[Evan Stade, 2009-09-24 20:08:28]

alright, I'm ok with this change.

[mattm, 2009-09-24 20:09:37]

On 2009/09/24 20:08:28, Evan Stade wrote:
> alright, I'm ok with this change.

I'm ok with it, but maybe add a comment?