Replace ancient crufty AppCacheThread with much improved MessageLoopProxy usage.

Replace ancient crufty AppCacheThread with much improved MessageLoopProxy usage.
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=100726

[michaeln, 2011-09-09 21:34:05]

Hi Dirk, I still have to update more unittests for the new Intialize signature,
but this first snapshot is the crux of it.

[michaeln, 2011-09-09 22:05:53]

ptal... looks like we got lucky on how widely the Initialize signature change
would ripple (not very). I've started some tryjobs too and i'll be sure to start
layout bots too.

[Dirk Pranke, 2011-09-09 22:40:19]

LGTM. Definitely better than my patch, and it's nice that we can get rid of
those globals.

http://codereview.chromium.org/7863009/diff/16/webkit/appcache/appcache_stora...
File webkit/appcache/appcache_storage_impl.h (right):

http://codereview.chromium.org/7863009/diff/16/webkit/appcache/appcache_stora...
webkit/appcache/appcache_storage_impl.h:126: bool is_incognito_;
Nit: any particular reason you moved this up? Because cache_directory_ isn't
used when in incognito mode, and you wanted the two to be together?

[michaeln, 2011-09-09 23:25:36]

thnx!

looks like i have to tweek something to make the
chrome_appcache_service_unittest happy... its just tripping on one of the
onthread dchecks

if i have to change anything interesting, i'll ask you to take another look

http://codereview.chromium.org/7863009/diff/16/webkit/appcache/appcache_stora...
File webkit/appcache/appcache_storage_impl.h (right):

http://codereview.chromium.org/7863009/diff/16/webkit/appcache/appcache_stora...
webkit/appcache/appcache_storage_impl.h:126: bool is_incognito_;
On 2011/09/09 22:40:19, Dirk Pranke wrote:
> Nit: any particular reason you moved this up?

The two i smushed together, cache_directory_ and is_incognito_, both have to do
with where stuff is stored. The (now two) message_loop ptrs don't have anything
to do with that. That cache_thread_ member probably shouldn't have been put
where it was to start with (but there was no compelling reason to change it
since it wasn't broke).

[michaeln, 2011-09-09 23:32:46]

> if i have to change anything interesting, i'll ask you to take another look

nothing interesting, the unittest uses the same thread for the IO and DB
threads, so this DCHECK that's executed on the db thread isn't valid in that
environment.

  DCHECK(!io_thread_->BelongsToCurrentThread())

it's phrased as NOT because there's no need for a db_thread ref in this class
(other than perhaps to make this assertion in the positive).

i'll just remove the DCHECK