Use copies of canned browsing data helpers to populate the cookies tree model.

Use copies of canned browsing data helpers to populate the cookies tree model.

If the fetching of data is cancelled, the helpers might not be ready yet to be reused, triggering an assertion.

BUG=62311
TEST=interactive ui tests:CollectedCookies.*

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=78198

[jochen (gone - plz use gerrit), 2011-03-11 13:02:43]

please review

[Paweł Hajdan Jr., 2011-03-11 14:32:09]

Changes look fine at a glance, but I don't really know this code enough. Could
you find someone else to review it?

Thank you for fixing flakiness!

[jochen (gone - plz use gerrit), 2011-03-11 14:40:15]

bernhard, can you have a look?

[Bernhard Bauer, 2011-03-15 10:39:01]

On 2011/03/11 14:40:15, jochen wrote:
> bernhard, can you have a look?

Does this mean that we possibly show old data in the cookie tree model?

[jochen (gone - plz use gerrit), 2011-03-15 10:44:45]

On 2011/03/15 10:39:01, Bernhard Bauer wrote:
> On 2011/03/11 14:40:15, jochen wrote:
> > bernhard, can you have a look?
> 
> Does this mean that we possibly show old data in the cookie tree model?

that's already now the case. The model is not "live", i.e. once it is created,
it reads the data from all browser helpers and displays the state at the time of
creation.

By cloning the helpers before creating the cookies tree model, no further delay
wrt the freshness of the data is introduced.

The problem this CL is supposed to fix is when you create a cookies tree model,
delete it again, before the async data fetch from the browsing data helpers
completed, and immediately recreate one, then the browsing data helpers trigger
a DCHECK that no new fetch is started while an old one didn't complete yet.

[Bernhard Bauer, 2011-03-15 10:52:32]

On 2011/03/15 10:44:45, jochen wrote:
> On 2011/03/15 10:39:01, Bernhard Bauer wrote:
> > On 2011/03/11 14:40:15, jochen wrote:
> > > bernhard, can you have a look?
> > 
> > Does this mean that we possibly show old data in the cookie tree model?
> 
> that's already now the case. The model is not "live", i.e. once it is created,
> it reads the data from all browser helpers and displays the state at the time
of
> creation.
> 
> By cloning the helpers before creating the cookies tree model, no further
delay
> wrt the freshness of the data is introduced.
> 
> The problem this CL is supposed to fix is when you create a cookies tree
model,
> delete it again, before the async data fetch from the browsing data helpers
> completed, and immediately recreate one, then the browsing data helpers
trigger
> a DCHECK that no new fetch is started while an old one didn't complete yet.

I see. LGTM then.