Service Worker: (Re-commit) Make ServiceWorkerRegistration.update() return a promise. (Blink 1/3)

Service Worker: (Re-commit) Make ServiceWorkerRegistration.update() return a promise. (Blink 1/3)

* The original CL was committed but reverted due to a leak report: https://codereview.chromium.org/1267703003/#msg29.
This CL re-commits the original CL with a patch for the leak. This change will
also be existed temporarily until the chromium-side patch will land.

--
As per the resolution of f2f, ServiceWorkerRegistration.update() should return a
promise that transforms the promise returned by Update algorithm.

In this CL, ServiceWorkerContextCore::UpdateServiceWorker method has been
overloaded to cover two invocation paths: a scheduled update without a
provider_host and a callback (Soft Update in the spec) and a call initiated from
the script surface using ServiceWorkerRegistration.update().

This is a web-exposed API change. But it has no compatibility risk because
existing user code doesn't expect to receive anything from update().

Spec: https://slightlyoff.github.io/ServiceWorker/spec/service_worker/#service-worker-registration-update
Spec discussion: https://github.com/slightlyoff/ServiceWorker/issues/311

Companion CL (Chromium): https://codereview.chromium.org/1270513002/
Companion CL (Blink layout test): https://codereview.chromium.org/1268663003/

BUG=513655
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=199875

[jungkees, 2015-08-01 04:51:52]

jungkee.song@samsung.com changed reviewers:
+ falken@chromium.org, jinho.bang@samsung.com, kinuko@chromium.org,
l.gombos@samsung.com, michaeln@chromium.org, nhiroki@chromium.org,
tkent@chromium.org

[jungkees, 2015-08-01 05:01:18]

tkent@, nhiroki@, zino@,

The commit from the original CL was reverted with a leak report:
https://codereview.chromium.org/1264143002/.  Here's the analysis I carried out:
https://codereview.chromium.org/1267703003/#msg30. And here's a solution that
zino@ suggested, and I incorporated that in this snapshot as such. PTAL.

[tkent, 2015-08-02 23:32:42]

lgtm

https://codereview.chromium.org/1260833003/diff/1/public/platform/WebServiceW...
File public/platform/WebServiceWorkerRegistration.h (right):

https://codereview.chromium.org/1260833003/diff/1/public/platform/WebServiceW...
public/platform/WebServiceWorkerRegistration.h:33: // internally call
this->update(p) to work with the existing code base.
nit: |this->update(p)| -> |update(provider)|

https://codereview.chromium.org/1260833003/diff/1/public/platform/WebServiceW...
public/platform/WebServiceWorkerRegistration.h:38: this->update(provider);
nit: remove |this->|

[zino, 2015-08-03 00:43:13]

non-onwer lgtm.

[nhiroki, 2015-08-03 03:19:04]

lgtm

[jungkees, 2015-08-03 03:29:28]

The CQ bit was checked by jungkee.song@samsung.com

[jungkees, 2015-08-03 03:29:28]

The patchset sent to the CQ was uploaded after l-g-t-m from tkent@chromium.org,
jinho.bang@samsung.com
Link to the patchset: https://codereview.chromium.org/1260833003/#ps20001
(title: "Remove |this| for calling its own method.")

[commit-bot: I haz the power, 2015-08-03 03:29:38]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1260833003/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1260833003/20001

[commit-bot: I haz the power, 2015-08-03 05:34:39]

Committed patchset #2 (id:20001) as
https://src.chromium.org/viewvc/blink?view=rev&revision=199875

[jungkees, 2015-08-03 08:24:57]

johnme@, I checked it from the flakiness dashboard that this commit still causes
http/tests/serviceworker/update.html to fail on WebKit Linux Leak and WebKit
Linux Oilpan Leak. But I have also checked locally using layout test's
--enable-leak-detection option that this failure goes away when the
chromium-side companion patch is applied. (That must be because the resolver
object is properly consumed and destroyed whereas our temporary patch might miss
something. I'm still looking at it.) 

I'm just waiting for this patch is rolled into the chromium repo and planning to
commit the chromium-side patch as soon as it can go in. So, if possible, I'd
like to try what I just explained here.

/cc tkent@, nhiroki@, zino@.

[johnme, 2015-08-03 08:42:34]

On 2015/08/03 08:24:57, jungkees wrote:
> johnme@, I checked it from the flakiness dashboard that this commit still
causes
> http/tests/serviceworker/update.html to fail on WebKit Linux Leak and WebKit
> Linux Oilpan Leak. But I have also checked locally using layout test's
> --enable-leak-detection option that this failure goes away when the
> chromium-side companion patch is applied. (That must be because the resolver
> object is properly consumed and destroyed whereas our temporary patch might
miss
> something. I'm still looking at it.) 
> 
> I'm just waiting for this patch is rolled into the chromium repo and planning
to
> commit the chromium-side patch as soon as it can go in. So, if possible, I'd
> like to try what I just explained here.
> 
> /cc tkent@, nhiroki@, zino@.

Sounds good, let's hope that does the trick.

[jungkees, 2015-08-03 12:15:54]

The chromium-side patch has just landed:
https://codereview.chromium.org/1270513002/#msg39. I'll keep an eye on the
flakiness dashboard.

[jungkees, 2015-08-03 13:28:33]

On 2015/08/03 12:15:54, jungkees wrote:
> The chromium-side patch has just landed:
> https://codereview.chromium.org/1270513002/#msg39. I'll keep an eye on the
> flakiness dashboard.

Yes, it starts passing again! :)

[johnme, 2015-08-03 13:59:28]

On 2015/08/03 13:28:33, jungkees wrote:
> On 2015/08/03 12:15:54, jungkees wrote:
> > The chromium-side patch has just landed:
> > https://codereview.chromium.org/1270513002/#msg39. I'll keep an eye on the
> > flakiness dashboard.
> 
> Yes, it starts passing again! :)

Thanks for following up :)