sync: migrate invalidation state from syncable::Directory to InvalidationStorage

sync: migrate invalidation state from syncable::Directory to InvalidationStorage

Deprecates SyncNotifier::SetState, as this is needed only to plumb notification state from the syncable::Directory to notifier.  It is still in use to facilitate migration.

Migration for an existing sync setup works as follows: 
1) normal SyncManager::Init occurs, which calls SetState on notifier.
2) The InvalidationNotifier takes the state received in SetState and echoes it back through the InvalidationStateTracker) which is the new store. It does this in addition to normal stashing, to hand to ChromeInvalidationClient when needed.
3) On a subsequent restart, the SyncNotifierFactory will see a non-empty initial invalidation state in the InvalidationStateTracker, and pass this down on creation to notifier components.
4) SetState calls will now be no-ops, as the invalidation state will be non-empty, and it can only be set once on this code path.

UMA tracks the number of "worthwhile" SetStateDeprecated calls for a rough global migration progress.

At some point in the future, once these bits have served all (statistically speaking) users, the SetStateDeprecated call can be removed entirely and the syncable::Directory schema updated to remove existing and unnecessary columns. I did not remove the column from DirectoryBackingStore::CreateTables, since the code doesn't really support a half-way migration and it's likely cleaner to do it in one fell swoop when we know it is safe.  Open to suggestions.

Built on http://codereview.chromium.org/10451058/

BUG=124140
TEST=

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=139706

[tim (not reviewing), 2012-05-29 17:56:09]

Nicolas / Nilesh, I think having you both review all would be ideal!  It's not
that big :)

I'm still trying to see if there's a way to write a test for the migration path,
will update.

akalin - FYI / TBR

[nilesh, 2012-05-29 23:01:36]

LGTM
Disclaimer:
I dont understand a lot of this :)
nzea would be able to provide a more meaningful review.

http://codereview.chromium.org/10451060/diff/5025/sync/notifier/invalidation_...
File sync/notifier/invalidation_notifier.h (right):

http://codereview.chromium.org/10451060/diff/5025/sync/notifier/invalidation_...
sync/notifier/invalidation_notifier.h:27: #include
"sync/notifier/state_writer.h"
I think you can remove the include now.

[Nicolas Zea, 2012-05-29 23:32:53]

Mostly LG

https://chromiumcodereview.appspot.com/10451060/diff/5025/sync/notifier/inval...
File sync/notifier/invalidation_notifier_unittest.cc (right):

https://chromiumcodereview.appspot.com/10451060/diff/5025/sync/notifier/inval...
sync/notifier/invalidation_notifier_unittest.cc:40:
base::WeakPtr<InvalidationStateTracker>()),
Seems like this would be a good place to test the logic. Passing in a mock/fake
here, and ensuring it only gets called once if the initial state was empty,
never otherwise.

[tim (not reviewing), 2012-05-30 02:02:47]

https://chromiumcodereview.appspot.com/10451060/diff/5025/sync/notifier/inval...
File sync/notifier/invalidation_notifier_unittest.cc (right):

https://chromiumcodereview.appspot.com/10451060/diff/5025/sync/notifier/inval...
sync/notifier/invalidation_notifier_unittest.cc:40:
base::WeakPtr<InvalidationStateTracker>()),
On 2012/05/29 23:32:53, nzea wrote:
> Seems like this would be a good place to test the logic. Passing in a
mock/fake
> here, and ensuring it only gets called once if the initial state was empty,
> never otherwise.

Done, thanks!

[Nicolas Zea, 2012-05-30 17:14:57]

LGTM

[tim (not reviewing), 2012-05-30 21:46:14]

Rebased with Fred's PushClient change, and also added code + test to
InvalidatorStorage to encode the state in base 64.

Please have another look (zea), thanks!

[Nicolas Zea, 2012-05-30 22:39:58]

LGTM

http://codereview.chromium.org/10451060/diff/14001/chrome/browser/sync/invali...
File chrome/browser/sync/invalidations/invalidator_storage_unittest.cc (right):

http://codereview.chromium.org/10451060/diff/14001/chrome/browser/sync/invali...
chrome/browser/sync/invalidations/invalidator_storage_unittest.cc:8: #include
"base/string_util.h"
abc order

[commit-bot: I haz the power, 2012-05-30 23:17:13]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/tim@chromium.org/10451060/14003

[commit-bot: I haz the power, 2012-05-31 00:34:43]

Try job failure for 10451060-14003 (retry) on win_rel for step "compile"
(clobber build).
It's a second try, previously, step "compile" failed.
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=win_rel&nu...