Also task switching away then loading more apps then switching back to Chrome, causes total unresponsiveness.

Users reporting this on play store reviews.

Deleted: Screenshot_20171124-153652.png 272 KB

	Screenshot_20171124-153652.png 272 KB View Download

Tested the issue on Pixel XL (Android 8.0.0 / OPR1.170623.027) using Chrome Canary# 64.0.3277.0 and surfed the browser for upto 45 minutes and did not observe any sluggish / unresponsive behavior.

Note: Played videos, browsed to various websites (search, images, videos) and could not observe any abnormal behavior.
Uninstalled and did a fresh install and could not reproduce the issue.

Thank You.

I got a Chirp about renderer hang count blowing up like crazy as well: https://uma.googleplex.com/timeline_v2?sid=110c92c98023ee1c7090761400c8420f

There have been doubts to the accuracy of this metric in the past, so I'm not sure how much I'd trust it alone, but between that and this bug (plus user feedback) we should probably investigate more.  Tentatively tagging RB-Beta for M64, but will defer to cmasso@'s assessment as well.

Meant to CC yfriedman@ here as well given this affects a Stability.Counts metric...  As context, renderer hang count is blowing up in canary.

Final update from me, I have seen canary crash repeatedly over the last few days - I haven't noticed hangs personally.  Perhaps the two are related, though...

I spent the whole long weekend suffering with this bug. I am surprised it did not reproduce. I can certainly try and bisect myself. I'm still pretty confident it started in 3276 as it was so noticeable (but I waited one Canary release before filing it just in case it was going to go away in a revert).

Testing is trying to repro.

perhaps try with accessibility enabled (installed tasker and make sure it's listed as Enabled in settings->Accessibility) - because last time that was the root cause of not being able to repro a bug by TE.

I think i saw this anecdotally too. Occasionally trying to launch a webapk, and the splashscreen just hangs there for a long while without doing anything

Let's block Dev on this since this seems pretty bad. We might need more people to work on bisecting this as well.

graphs from c5 split by version do confirm this is probably a regression in 3276 - https://uma.googleplex.com/timeline_v2?sid=ec25364903c40aef423872630ff95abd

chromium revs: https://chromium.googlesource.com/chromium/src/+log/64.0.3275.0..64.0.3276.0?pretty=fuller&n=10000

v8 revs: https://chromium.googlesource.com/v8/v8/+log/fdf73522..b759ad89?pretty=fuller&n=10000

I can see if I can get a local repro on a Chromium build in that range.

+stability sheriff since we need more hands on this

I don't have exact repro steps, but I've hit it often enough today that I have a suspected CL.

Tested on Pixel / OPR1.170623.034

Good build: 64.0.3275.3
Bad build: 64.0.3276.0
Regression range: https://chromium.googlesource.com/chromium/src/+log/64.0.3275.3..64.0.3276.0?pretty=fuller&n=10000

Good commit: 518573
Bad commit: 518574
Suspected CL: https://chromium.googlesource.com/chromium/src/+/35431df0957d5dd0fddfe210843128b1e3def0aa#

https://chromium.googlesource.com/v8/v8/+log/df985bbc..0001378a

 Issue 788962  has been merged into this issue.

hablich@, can you help with triaging this blocker?

I'm unable to repro this with a Pixel 2 so far.

candrada@, is it possible for you to bisect into the V8 roll? The most relevant CL in that roll is probably https://chromium.googlesource.com/v8/v8/+/0481b23e2562fcb001494de3765a3ef61427fd00.

I don't know how to bisect into the V8 roll.  Is there someone who can help?

I've seen the page freeze after about 5-10 minutes of browsing web pages or playing Google Solitaire.  Happens in custom tabs, chrome tabs, and incognito tabs.  The page freezes up, and I cannot tap on anything on the page or scroll.  I am able to open other tabs and switch tabs though.

adamk@ can you help bisecting into the V8 roll?

Adding jkummerow@ since adamk@ is not available today.

Every V8 roll's CL description contains information about its contents, in this case:

Summary of changes available at:
https://chromium.googlesource.com/v8/v8/+log/df985bbc..0001378a

You can use that to bisect into the roll as follows:

// Check out the roll commit, and get other DEPS into the right state.
chromium/src> git checkout 35431df0957d5dd0fddfe210843128b1e3def0aa
chromium/src> gclient sync
// Bisect into V8.
chromium/src> cd v8
chromium/src/v8> git bisect start
// The currently checked-out revision should be 0001378a.
// You can run "git log -1" to confirm, and/or compile and test
// to verify that this is indeed a broken revision.
chromium/src/v8> git bisect bad
// Then check out the previously rolled V8 revision.
// Optionally compile and test to verify that it is good.
chromium/src/v8> git checkout df985bbc
chromium/src/v8> git bisect good
// Then follow regular bisect steps (compile, test, mark as good or bad).

Thanks for the steps.

aluo@ helped me get a build up (thanks aluo@!), and we bisected and confirmed that https://chromium.googlesource.com/v8/v8/+/0481b23e2562fcb001494de3765a3ef61427fd00 is indeed the culprit.

neis@, can we get a revert/fix?

Assigning to the author of that change, bbudge, who's probably in a better position to figure out whether a straight revert or fix is the right next step.

This CL is intended to solve the memory problem:

https://chromium-review.googlesource.com/c/v8/v8/+/790932

The fix or a revert needs to be on the to-be-created branch V8 6.4. I guess this still needs to be merged to 64.0.3278.0 to unblock Android dev.

Merged: https://chromium-review.googlesource.com/c/v8/v8/+/795721

This is active on V8's branch chromium/3278: https://chromium.googlesource.com/v8/v8/+log/chromium/3278

This change should be automatically be picked up when you rebuild (at least according to internal Chrome's buildspec DEPS file.

Forgot to add: This should no longer be blocking Android Dev.

Still, fix needs to be on 64 beta.

@bbudge would you please take a look?

I can repro on a Pixel XL - Chrome Canary is 64.0.3279.0.

It looks like the fix got merged to 3278 but not 3279 or 3280. I'm not sure how Canary is released, so without more merges it might be another day before the fix is in.

In case the fix CL isn't effective, I've prepared the full revert. It's a fairly large change at this point though.

https://chromium-review.googlesource.com/c/v8/v8/+/797924

Maybe also  https://crbug.com/v8/7146  is related?

which version(s) of Canary is this fix in? I'd like to try and verify if I am still seeing this behavior or not.

Today's Canary (64.0.3281.0) doesn't become unresponsive in my limited testing this morning.

I haven't been able to repro the issue either on latest canary so far.

I think this is fixed.

Fix was merged before M64 branch cut.