Blow away BackgroundContents when RenderView goes away.

Blow away BackgroundContents when RenderView goes away.

Reopening, because this CL broke NoticeNotificationChanges on XP for some reason, so I'm going to undo my (now superfluous) changes to flush the queue in WaitForResourceChange().

BUG=65189
TEST=TaskManagerBrowserTest.KillBGContents

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=71458

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=71688

[Andrew T Wilson (Slow), 2011-01-12 23:58:14]

Please take a look.

[rafaelw, 2011-01-14 01:04:45]

code lg. just question about testing.

http://codereview.chromium.org/6226002/diff/5001/chrome/browser/task_manager/...
File chrome/browser/task_manager/task_manager_browsertest.cc (right):

http://codereview.chromium.org/6226002/diff/5001/chrome/browser/task_manager/...
chrome/browser/task_manager/task_manager_browsertest.cc:70: // a task to make
sure we aren't triggered by a transient change.
I'm not familiar with the TaskManager or this test, so forgive me if the
question is silly:

I assume that this is safe because the remove & add that results from a BC
navigation always happens in a single event cycle? even so, this seems a bit
brittle. is there a safer way to approach this?

[Andrew T Wilson (Slow), 2011-01-14 01:38:29]

On 2011/01/14 01:04:45, rafaelw wrote:
> code lg. just question about testing.
> 
>
http://codereview.chromium.org/6226002/diff/5001/chrome/browser/task_manager/...
> File chrome/browser/task_manager/task_manager_browsertest.cc (right):
> 
>
http://codereview.chromium.org/6226002/diff/5001/chrome/browser/task_manager/...
> chrome/browser/task_manager/task_manager_browsertest.cc:70: // a task to make
> sure we aren't triggered by a transient change.
> I'm not familiar with the TaskManager or this test, so forgive me if the
> question is silly:
> 
> I assume that this is safe because the remove & add that results from a BC
> navigation always happens in a single event cycle? even so, this seems a bit
> brittle. is there a safer way to approach this?

Exactly, it's safe because the remove & add that results from a BC navigation
happens in the processing of a single event.

What happens is the task manager code wants to update the title of a single
item. The way it does this is by removing the old item and re-adding an item
with the correct title. You can see an example of this in
task_manager_resource_providers.cc:369.

Basically, the test code was producing false-positives previously - it would
incorrectly trigger WaitForResourceChange() even though we hadn't truly add or
removed a resource, so it's actually less fragile now/more correct than before.

[rafaelw, 2011-01-14 01:53:38]

ok. lgtm