drive: Stop making multiple GetAppList() requests from DriveAppRegistry

drive: Stop making multiple GetAppList() requests from DriveAppRegistry

Calling DriveAppRegistry::Update() results in being ignored when there is an ongoing one already.
Add FakeDriveService::app_list_load_count()

BUG=268707
TEST=unit_tests
R=hidehiko@chromium.org, kinaba@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=215806

[kinaba, 2013-08-06 04:17:23]

I'm wondering if we can remove |needs_to_update_again_|
and just discard Update() calls during an inflight request.

We don't have any mechanism to sense updates in the app list,
so we're just putting aside the event that periodically happens
for Drive users. Thus, we don't have any strict requirement to process all
Update() requests.

WDYT? (+hidehiko)

[hidehiko, 2013-08-06 04:31:01]

On 2013/08/06 04:17:23, kinaba wrote:
> I'm wondering if we can remove |needs_to_update_again_|
> and just discard Update() calls during an inflight request.
> 
> We don't have any mechanism to sense updates in the app list,
> so we're just putting aside the event that periodically happens
> for Drive users. Thus, we don't have any strict requirement to process all
> Update() requests.
> 
> WDYT? (+hidehiko)

As we have no way to guarantee keeping the local AppList up to date, and
updating the AppList would be rare,
I think it should be OK to remove the flag.
It may be better to have an option to force to fetch AppList for users as a last
resort?

[kinaba, 2013-08-06 05:02:16]

On 2013/08/06 04:31:01, hidehiko wrote:
> On 2013/08/06 04:17:23, kinaba wrote:
> > I'm wondering if we can remove |needs_to_update_again_|
> > and just discard Update() calls during an inflight request.
> > 
> > We don't have any mechanism to sense updates in the app list,
> > so we're just putting aside the event that periodically happens
> > for Drive users. Thus, we don't have any strict requirement to process all
> > Update() requests.
> > 
> > WDYT? (+hidehiko)
> 
> As we have no way to guarantee keeping the local AppList up to date, and
> updating the AppList would be rare,
> I think it should be OK to remove the flag.

Thanks, then I'd like to suggest having only single-bit state |is_updating_|,
and just ignore duplicated Update() calls for keeping simplicity.
(But not quite strong opinion about it. What do you think? > hashimoto)

> It may be better to have an option to force to fetch AppList for users as a
last
> resort?

I'd rather like to ensure it is updated at least once per Files.app open,
but anyway this is a separate issue.

[hashimoto, 2013-08-06 05:12:02]

Simply ignoring Update() sounds good, done.
(the previous patch set had a bug around error handling, luckily it's no longer
there...)

[kinaba, 2013-08-06 05:13:29]

lgtm

[hidehiko, 2013-08-06 05:18:52]

lgtm

[hashimoto, 2013-08-06 05:54:28]

Committed patchset #2 manually as r215806 (presubmit successful).