New issue
Advanced search Search tips

Issue 824136 link

Starred by 1 user

Issue metadata

Status: WontFix
Owner:
Closed: Apr 2018
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

8% regression in octane at 544316:544336

Project Member Reported by verwa...@chromium.org, Mar 21 2018

Issue description

See the link to graphs below.
 
Project Member

Comment 1 by 42576172...@developer.gserviceaccount.com, Mar 21 2018

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=824136

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=0398faf1413a0642e30b3d63f3f1daa35706dee0e1b4d343370b7ccfc4a00086


Bot(s) for this bug's original alert(s):

android-nexus6
Cc: benjhayden@chromium.org hablich@chromium.org
Ben, any clue why pinpoint already running for nearly a week?
Cc: dtu@chromium.org
Owner: dtu@chromium.org
Dave:
1. It looks like pinpoint raised AccessTokenRefreshError?
2. Is there a dashboard view of long-running (possibly stalled) pinpoint jobs?
3. It might be a nice feature if we could link to specific build/test/value quests in the UI using hash params.

Comment 5 by dtu@chromium.org, Mar 28 2018

Owner: marja@chromium.org
Status: Assigned (was: Untriaged)
The Nexus 5X and Nexus 6 bots (both WebView and non-WebView) are way overloaded, so jobs take a long time. It's not stalled, though! AccessTokenRefreshError looks like a transient error that will not affect the job results.

Looks like the result here is gonna be this CL, though:

[in-place weak refs prework] No dependent code handling when maps die.
https://chromium.googlesource.com/v8/v8/+/58b0dea85cee7ebba093dc8f6cbb2d2267824b79
Project Member

Comment 6 by 42576172...@developer.gserviceaccount.com, Mar 31 2018

Cc: u...@chromium.org marja@chromium.org
📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/1108b079440000

[in-place weak refs prework] No dependent code handling when maps die. by marja@chromium.org
https://chromium.googlesource.com/v8/v8/+/58b0dea85cee7ebba093dc8f6cbb2d2267824b79

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

Comment 7 by marja@chromium.org, Apr 5 2018

I can repro this regression (around 5%) locally on Linux, running only the TypeScript test (commenting out other tests in Octane).

Comment 8 by marja@chromium.org, Apr 9 2018

This regression is not as reproducible as it first seemed; I can't repro it any more locally :P (I.e., first it looked like I can repro it but that was just... untrue. No idea what was going on.)

This also doesn't occur consistently across bots:

https://chromeperf.appspot.com/report?sid=dd93e8573143343dfa346a9cb5d96bf5e458b3b77b43d95deb6c39b983bf88e7&start_rev=542374&end_rev=547380

The regression is only visible in a couple of Android bots.

Comment 9 by marja@chromium.org, Apr 16 2018

Status: WontFix (was: Assigned)
Also checked the roll where this commit is contained; didn't see anything suspicious in the unaddressed regressions.

Sign in to add a comment