New issue
Advanced search Search tips

Issue 607324 link

Starred by 3 users

Issue metadata

Status: Fixed
Owner:
Closed: Aug 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: All
Pri: 1
Type: Bug



Sign in to add a comment

PerformanceEntries recorded before PerformanceObserver is disconnected are not delivered

Project Member Reported by igrigo...@chromium.org, Apr 27 2016

Issue description

Version: 52.0.2718.0 canary (64-bit)
OS: OSX

What steps will reproduce the problem?
(1) Enable experimental features flag
(2) Load http://w3c-test.org/submissions/2884/performance-timeline/po-disconnect.html

The failing test is:

> observer.observe({entryTypes: ["mark"]});
> performance.mark("mark1");
> observer.disconnect();

The mark ("mark1") is recorded before the observer is disconnected but is never delivered.

WIP W3C tests:
- http://w3c-test.org/submissions/2884/performance-timeline/
- https://github.com/w3c/web-platform-tests/pull/2884
 
Owner: panicker@chromium.org
Status: Fixed (was: Untriaged)
Status: Available (was: Fixed)
Based on discussion in [1], we should revert this behavior.

[1] https://github.com/w3c/performance-timeline/issues/66#issuecomment-282396844
Cc: igrigo...@chromium.org
Labels: Hotlist-PerformanceAPIs
Ping? Is this behavior going to be reverted?

It seems like the resolution was that disconnect() should *not* flush records and this bug should be reverted.

We'd like to ship this API in Firefox but don't want to ship something that is incompatible with Chrome.
Labels: -Pri-3 Pri-1
Owner: npm@chromium.org
Status: Assigned (was: Available)
Thanks for the ping.

Nicolas, can you take a look?
Marking this P1, as it's blocking Firefox from shipping.
Project Member

Comment 8 by bugdroid1@chromium.org, Aug 9 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/7df7884ee789a33cd850c04d6d462aaacb6df9f0

commit 7df7884ee789a33cd850c04d6d462aaacb6df9f0
Author: Nicolas Pena <npm@chromium.org>
Date: Wed Aug 09 04:53:40 2017

Revert flush behavior when disconnecting PerformanceObserver

This CL reverts the flush behavior introduced in
https://crrev.com/32102c0baf205ff54cd16644ddd823a6cb6538f5

This is reverted after upstream agreement that it is not really the
desired behavior, and to align with Mozilla's behavior:
https://github.com/w3c/performance-timeline/issues/66#issuecomment-282396844

Bug:  chromium:607324 
Change-Id: Ia38649f29f99d9c71f06da1a1a3318cc83a82e12
Reviewed-on: https://chromium-review.googlesource.com/606209
Reviewed-by: Timothy Dresser <tdresser@chromium.org>
Reviewed-by: Shubhie Panicker <panicker@chromium.org>
Reviewed-by: Steve Kobes <skobes@chromium.org>
Commit-Queue: Nicolás Peña <npm@chromium.org>
Cr-Commit-Position: refs/heads/master@{#492826}
[modify] https://crrev.com/7df7884ee789a33cd850c04d6d462aaacb6df9f0/third_party/WebKit/LayoutTests/TestExpectations
[modify] https://crrev.com/7df7884ee789a33cd850c04d6d462aaacb6df9f0/third_party/WebKit/Source/core/timing/PerformanceBase.cpp
[modify] https://crrev.com/7df7884ee789a33cd850c04d6d462aaacb6df9f0/third_party/WebKit/Source/core/timing/PerformanceBaseTest.cpp
[modify] https://crrev.com/7df7884ee789a33cd850c04d6d462aaacb6df9f0/third_party/WebKit/Source/core/timing/PerformanceObserver.cpp

Comment 9 by npm@chromium.org, Aug 11 2017

Status: Fixed (was: Assigned)

Sign in to add a comment