Please change the title to "workerStart (Navigation Timing 2 API) is zero even when the HTML document is served by service worker"

Able to reproduce the issue using latest stable #59.0.3071.86 on Mac 10.12.4, Win 10 and Linux Ubuntu 14.04.

Below is the bisect info
========================

Good Build: 59.0.3044.0  (Revision: 457648)
Bad Build:  59.0.3045.0  (Revision: 457915)


Using per-revision bisect providing the bisect results:

You are probably looking for a change made after 457789 (known good), but no later than 457790 (first known bad).

CHANGELOG URL:
 https://chromium.googlesource.com/chromium/src/+log/a9691a68bb9cb7e7d008bfdd881fa27023c8181f..ea6b98729307be8655a597ff4ac559710b7d68c6

Review-Url: https://codereview.chromium.org/2757893002

vasilii@: Could you please take a look into this issue and reassign if this issue is not related to your change.

Note: Issue is seen in M61 as well

Thanks..!!

Kenji, please triage

Shubhie, can you take a look?
Having a functional workerStart is critical for key PWA partners working latency mitigations.

Thanks!

The per-revision bisect result doesn't make sense: that change just affects tests and not the Chrome binary.

I don't understand the PerformanceNavigationTiming results. workerStart is 0 but so are many other things.

Going to a random non-service worker site I see:
requestStart: 0
responseEnd: 145.85
responseStart: 0
secureConnectionStart: 0
startTime: 0
workerStart: 0

If startTime, requestStart, responseStart are 0, for a service-worker enabled site I don't know what workerStart would be.  I wonder if this is just WAI.

I should have recorded how I tested that.

Repro steps:
1) Go to http://www.example.com/
2) In console run
performance.getEntriesByType("navigation")[0];

Looks like something is broken in our NT2 implementation.. Using the example.com test page:

> nt = performance.getEntriesByType('navigation')[0]
> (performance.timing.responseStart - performance.timing.requestStart) == (nt.responseStart - nt.requestStart)
false
> (performance.timing.responseStart - performance.timing.requestStart) 
5
> (nt.responseStart - nt.requestStart)
0

^ orthogonal to workerStart, but a good sign that something's not right in our plumbing. NT1 and NT2 should agree.

Ditto for workerStart, it seems to always report 0. Hmm...

Did a bit of digging, it looks time-origin_ is 0 in navigation timing case, ...because it's not initialized in navigation timing case.

This is the change that broke it, I think: https://codereview.chromium.org/2647643004

I have a wip patch that fixes it (by the way it doesn't make (performance.timing.responseStart - performance.timing.requestStart) == (nt.responseStart - nt.requestStart) true due to precision issues)

And the wpt test was broken too... will fix that as well.

Awesome, thanks! was just wondering why the tests didn't catch this :)

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/7c6d16f59e7c39e4bd6f2d32645956273e963e5f

commit 7c6d16f59e7c39e4bd6f2d32645956273e963e5f
Author: Kinuko Yasuda <kinuko@chromium.org>
Date: Tue Jul 11 08:35:38 2017

Fix NavigationTiming2

time_origin_ is not initialized in PerformanceResourceTiming ctor
for PerformanceNavigationTiming.  This got broken in crrev.com/2647643004

Bug:  732622 
Change-Id: I735ea6777d88b0e89031654216d74f860e05bcb8
Reviewed-on: https://chromium-review.googlesource.com/566261
Reviewed-by: Kunihiko Sakamoto <ksakamoto@chromium.org>
Commit-Queue: Kinuko Yasuda <kinuko@chromium.org>
Cr-Commit-Position: refs/heads/master@{#485559}
[modify] https://crrev.com/7c6d16f59e7c39e4bd6f2d32645956273e963e5f/third_party/WebKit/LayoutTests/external/wpt/navigation-timing/nav2_test_attributes_values.html
[modify] https://crrev.com/7c6d16f59e7c39e4bd6f2d32645956273e963e5f/third_party/WebKit/Source/core/timing/PerformanceNavigationTiming.cpp
[modify] https://crrev.com/7c6d16f59e7c39e4bd6f2d32645956273e963e5f/third_party/WebKit/Source/core/timing/PerformanceNavigationTiming.h
[modify] https://crrev.com/7c6d16f59e7c39e4bd6f2d32645956273e963e5f/third_party/WebKit/Source/core/timing/PerformanceResourceTiming.cpp
[modify] https://crrev.com/7c6d16f59e7c39e4bd6f2d32645956273e963e5f/third_party/WebKit/Source/core/timing/PerformanceResourceTiming.h

Matt: can you verify workerStart on ToT when you have time?

Yep, now there are sane numbers on current Dev for both the non-sw case <http://www.example.com/> and the sw case <https://jakearchibald.github.io/trained-to-thrill/>. Nice!