New issue
Advanced search Search tips

Issue 638699 link

Starred by 3 users
Status: Duplicate
Merged: issue 637640
Owner: ----
Closed: Aug 2016
Cc:
csharrison@chromium.org
mmenke@chromium.org
altimin@chromium.org
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 3
Type: Bug



Sign in to add a comment

Win canary gets slower and slower over time

Project Member Reported by mmenke@chromium.org, Aug 17 2016 
Version: Win 54.0.2831.0 (Official Build) canary (32-bit)
OS: Win 7

What steps will reproduce the problem?
(1)  Open up codereview.
(2)  Review code for a while.
(3)  Pages eventually start taking much longer to load at some point (~22 seconds instead of ~7)

Killing the renderer (Or restarting) fixes the issue for a while, but it starts up again.  I've repeated this process at least 3 times today.  It goes from reasonably performant to slow, with nothing in the middle.  Maybe the issue is a memory leak, maybe not - need to investigate further.

Attaching a "bad" trace with the issue, and a "good" trace recorded immediately after - all I did was kill all renderer processes in windows task manager, and then reload the page.
trace_bad.json.gz
7.7 MB Download

Comment 1 by mmenke@chromium.org, Aug 17 2016 

Oops, looks like the good trace wasn't attached, let's try again.
trace_good.json.gz
3.4 MB Download

Comment 2 by csharrison@chromium.org, Aug 17 2016

Cc: haraken@chromium.org
Components: Blink>Scheduling Blink>HTML>Parser
So strange. It looks like the bad trace has some bogus scheduling going on wrt html parsing. The traces definitely don't look like we're hammering CPU.

What categories did you use? It might make sense if this continues to repro to just log with all categories enabled.

haraken@, could https://codereview.chromium.org/2252693003/ have changed behavior here?

Comment 3 by mmenke@chromium.org, Aug 17 2016 

Just the default set.  I'll record a log with more stuff once it starts happening again.

Comment 4 by csharrison@chromium.org, Aug 17 2016

Cc: -haraken@chromium.org
RE: #2 nvm looks like that patch landed in 2832 but the traces were logged on 2831. Removing haraken from cc.

Comment 5 by hichris...@gmail.com, Aug 17 2016 

Possibly related:  bug 637640 .

Comment 6 by mmenke@chromium.org, Aug 17 2016 

Just to update, it doesn't look like the issue is running out of memory.  I'll try and gather more data next time it occurs.

Comment 7 by mmenke@chromium.org, Aug 17 2016

Mergedinto: 637640
Status: Duplicate (was: Untriaged)
►

Sign in to add a comment