New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 631313 link

Starred by 1 user

Issue metadata

Status: Duplicate
Merged: issue 627183
Owner:
Closed: Jul 2016
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 1
Type: Bug-Regression



Sign in to add a comment

Chrome hanging due to constant ViewHostMsg_PageScaleFactorChanged message

Project Member Reported by rbyers@chromium.org, Jul 26 2016

Issue description

Chrome Version       : 53.0.2785.23
OS Version: 8530.24.0

I've been having a bunch of trouble with pages hanging - tabs becoming entirely unresponsive.  I think this started after a recent dev channel update.  I'm not sure exactly how to reproduce this.  I've definitely seen it from Google docs tabs, possibly only when they're zoomed with browser zoom.

The last time it occurred I took a trace at (attached) and saw a constant stream (~80/ms) of ViewHostMsg_PageScaleFactorChanged messages swamping the renderer and browser IO threads as well as the browser main thread.

It looks like this message originates in VisualViewport::didSetScaleOrLocation.  What page behavior could cause it to be sent repeatedly?  Should we at least be throttling to no more than one per frame?

If this is indeed a regression in M53, then it's probably a beta blocker - it's making my device almost unusable for any real length of time.
 

Comment 1 by rbyers@chromium.org, Jul 26 2016

Labels: M-53

Comment 2 by rbyers@chromium.org, Jul 26 2016

trace_PageScaleFactorChanged_Hang-trace.json.gz
5.0 MB Download

Comment 3 by rbyers@chromium.org, Jul 26 2016

Just hit this again in a renderer that was not on google docs or zoomed (https://www.consumeraffairs.com/travel/platepass-com.html and https://www.platepass.com/).  Still no idea how to trigger it though, but is happening pretty regularly for me.

Project Member

Comment 4 by sheriffbot@chromium.org, Jul 26 2016

Labels: -M-53 M-54 MovedFrom-53
Moving this nonessential bug to the next milestone.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Comment 5 by bokan@chromium.org, Jul 26 2016

Cc: -bokan@chromium.org
Labels: -M-54 -MovedFrom-53 M-53 ReleaseBlock-Stable
Owner: bokan@chromium.org
Status: Assigned (was: Unconfirmed)
Thanks Rick, I'll use a dev chromebook for a while and see if I can repro, about how often do you see this (e.g. 1/day)?

We should only ever send that message out of VisualViewport::didSetScaleOrLocation if you're zoomed in (since the visual viewport location can't change unless you're zoomed) so it's strange it happens when unzoomed.

In any case, we definitely should throttle it since there's no need to send more than once per frame but I want to get at the root cause here since that sounds like a bug in our logic somewhere.

Bringing back M-53 label since we don't want to ship this.

Comment 6 by rbyers@chromium.org, Jul 26 2016

Thanks David!  I'd say I was hitting this (or something with similar impact) about once in every 30 minutes of usage.  If you can't repro and find the trigger, then perhaps you just want to add some additional traces on the blink side to narrow down which code path this may be coming from (possibly in addition to throttling)?

Comment 7 by bokan@chromium.org, Jul 26 2016

I've managed to repro, aggressively pinch zooming in and out on the trackpad seems to do it. There's a parallel email chain discussing this problem and it looks likely like it has to do with some recent changes to how trackpad pinch-zoom is handled for OOPIFs (site isolation is on for google.com accounts). I'll keep poking.

Comment 8 by bokan@chromium.org, Jul 26 2016

Mergedinto: 627183
Status: Duplicate (was: Assigned)

Sign in to add a comment