Sami, any thoughts on the best way to do this?

Modal dialogs are implemented with nested run loops. One fix would be to reset our estimates during and/or after nested run loops.

That sounds reasonable to me, but I don't have a good sense of how often nested run loops are used.

They're pretty rare, and tend to run fairly long, correct? If there were lots of short nested run loops, I'd say it wouldn't make sense to completely reset - we'd just want to exclude the nested run loops from the reported estimate.

Yeah, they're (thankfully) extremely rare. If there's a neat way to exclude them that would work too.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/c4067e72471b45ccb8d775189e53c6bcd3cb2c91

commit c4067e72471b45ccb8d775189e53c6bcd3cb2c91
Author: tdresser <tdresser@chromium.org>
Date: Fri Mar 31 20:34:59 2017

Expected Queueing Time Metric ignores tasks with nested message loops.

We're seeing cases where the Expected Queueing Time metric is reporting
values which are far too high. We hypothesize that this is due to
nested message loops, which currently cause us to report a task with
length equal to the duration of the nested message loop.

Exclude tasks containing nested message loops from this metric.

BUG= 675060 

Review-Url: https://codereview.chromium.org/2788013003
Cr-Commit-Position: refs/heads/master@{#461210}

[modify] https://crrev.com/c4067e72471b45ccb8d775189e53c6bcd3cb2c91/third_party/WebKit/Source/core/frame/PerformanceMonitor.h
[modify] https://crrev.com/c4067e72471b45ccb8d775189e53c6bcd3cb2c91/third_party/WebKit/Source/platform/scheduler/base/queueing_time_estimator.cc
[modify] https://crrev.com/c4067e72471b45ccb8d775189e53c6bcd3cb2c91/third_party/WebKit/Source/platform/scheduler/base/queueing_time_estimator.h
[modify] https://crrev.com/c4067e72471b45ccb8d775189e53c6bcd3cb2c91/third_party/WebKit/Source/platform/scheduler/base/queueing_time_estimator_unittest.cc
[modify] https://crrev.com/c4067e72471b45ccb8d775189e53c6bcd3cb2c91/third_party/WebKit/Source/platform/scheduler/base/task_queue_manager.cc
[modify] https://crrev.com/c4067e72471b45ccb8d775189e53c6bcd3cb2c91/third_party/WebKit/Source/platform/scheduler/base/test_task_time_observer.h
[modify] https://crrev.com/c4067e72471b45ccb8d775189e53c6bcd3cb2c91/third_party/WebKit/Source/platform/scheduler/renderer/renderer_scheduler_impl.cc
[modify] https://crrev.com/c4067e72471b45ccb8d775189e53c6bcd3cb2c91/third_party/WebKit/Source/platform/scheduler/renderer/renderer_scheduler_impl.h
[modify] https://crrev.com/c4067e72471b45ccb8d775189e53c6bcd3cb2c91/third_party/WebKit/public/platform/scheduler/base/task_time_observer.h

With these changes, the data we're getting now makes sense.