New issue
Advanced search Search tips

Issue 911334 link

Starred by 5 users

Issue metadata

Status: Verified
Owner:
Closed: Dec 12
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 1
Type: Bug-Regression



Sign in to add a comment

eve: Two-finger scrolling stops working periodically

Project Member Reported by derat@chromium.org, Dec 4

Issue description

On a ToT eve (a.k.a. Pixelbook) build, two-finger scrolling frequently stops working entirely in web content. After a variable number of seconds, it starts working again. This is trivial to reproduce across multiple devices.

Chrome 73.0.3629.0
Chrome OS 11338.0.0

I wasn't able to file a feedback report since I did my own build, but I suspect it's easy to see this on other devices. I first noticed it on M72 a few weeks ago. adlr@ says that he sees it too, I think. I don't know if other hardware is similarly affected.

(Note that this is distinct from issue 889707. That tracks stutter in two-finger flings that has existed at least since M72. This tracks a new, worse regression where two-finger scrolling stops working entirely.)
 
I've uploaded a video demonstrating the problem at https://drive.google.com/open?id=1VdnoecAwbL0GVjqyZa8KJMEpwq4LWmbs.
Status: Started (was: Assigned)
I confirm that the bug is device dependent:
I can repo the is on eve but not on samus both devices with exact same Chromium and Platform versions:
Chromium: 73.0.3631.0 (Developer Build) (64-bit)
Platform: 11244.0.0 (official Build) dev-channel samus/eve test.

Still working on the traces to see what are the differences between the two devices. But the bug is very unlikely to be Chromium related since Chromium code does not have any device specific logic and all ChromeOS devices are treated the same.
I just did a very clear-cut repro, where I did a scroll, grabbed my touchpad logs, and I saw that the touchpad code generated scroll events. However, the web page didn't scroll.

B/c of this it looks like it's not a firmware issue. May still be Eve (which does hover events), but doesn't look like firmware issue.
Why is this RBB for M72?
Having scrolling stop working every few seconds is an extremely poor user experience.
Project Member

Comment 6 by bugdroid1@chromium.org, Dec 7

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/0996ec8f771b7c0cc28aa56b201abf3ace2d4176

commit 0996ec8f771b7c0cc28aa56b201abf3ace2d4176
Author: Sahel Sharify <sahel@chromium.org>
Date: Fri Dec 07 17:19:41 2018

Revert "Register fling scheduler on GSB instead of GFS."

Revert "Register fling scheduler on GSB instead of GFS." and
Revert "Fixed fling regression from registering observer on GSB instead of GFS."

This reverts the following two commits:
0b48b62c4392e7cd594d9fbe8684e93abc72aa1b
0e58f2caab6a230fdb0f2b1993a44c2f4546c2f5

The reason for reverting is that registering fling scheduler observer on GSB
instead of GFS does not fix the one frame jank at the beginning of the
fling. This change also caused some performance regressions and touchpad
scrolling regression on eve.

Bug: 882907, 911975 , 911334 ,903161,903156, 903155 
Change-Id: I6aba2c0314eeab616fcb1c18c448c0b3c075d4e6
Reviewed-on: https://chromium-review.googlesource.com/c/1363893
Reviewed-by: Dave Tapuska <dtapuska@chromium.org>
Commit-Queue: Sahel Sharify <sahel@chromium.org>
Cr-Commit-Position: refs/heads/master@{#614730}
[modify] https://crrev.com/0996ec8f771b7c0cc28aa56b201abf3ace2d4176/content/browser/renderer_host/input/fling_controller.cc
[modify] https://crrev.com/0996ec8f771b7c0cc28aa56b201abf3ace2d4176/content/browser/renderer_host/input/fling_controller.h
[modify] https://crrev.com/0996ec8f771b7c0cc28aa56b201abf3ace2d4176/content/browser/renderer_host/input/fling_controller_unittest.cc
[modify] https://crrev.com/0996ec8f771b7c0cc28aa56b201abf3ace2d4176/content/browser/renderer_host/input/fling_scheduler.cc
[modify] https://crrev.com/0996ec8f771b7c0cc28aa56b201abf3ace2d4176/content/browser/renderer_host/input/fling_scheduler.h
[modify] https://crrev.com/0996ec8f771b7c0cc28aa56b201abf3ace2d4176/content/browser/renderer_host/input/fling_scheduler_android.cc
[modify] https://crrev.com/0996ec8f771b7c0cc28aa56b201abf3ace2d4176/content/browser/renderer_host/input/fling_scheduler_android.h
[modify] https://crrev.com/0996ec8f771b7c0cc28aa56b201abf3ace2d4176/content/browser/renderer_host/input/fling_scheduler_unittest.cc
[modify] https://crrev.com/0996ec8f771b7c0cc28aa56b201abf3ace2d4176/content/browser/renderer_host/input/gesture_event_queue.cc
[modify] https://crrev.com/0996ec8f771b7c0cc28aa56b201abf3ace2d4176/content/browser/renderer_host/input/gesture_event_queue_unittest.cc
[modify] https://crrev.com/0996ec8f771b7c0cc28aa56b201abf3ace2d4176/content/browser/renderer_host/input/mock_input_router_client.h

Does this need to be merged to the 3626 branch?  If so, it needs to happen before 12:00 PST tomorrow or eve will be skipped for the first M72 beta release.
Project Member

Comment 8 by sheriffbot@chromium.org, Dec 11

Labels: Pri-0
Priority is being increased due to the next release approaching.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
Cc: sahel@chromium.org
Owner: bokan@chromium.org
sahel is OOO, re-assigning to bokan
The revert in #6 should be in any build after and including 73.0.3634.0. Could someone verify that the issue is resolved by that CL?
Labels: Merge-Request-72
Pinged a few people but no one's able to verify the fix. Given how close we are to cutting beta we need to merge now if we want it in the next 72 beta. Given it's been in Canary without causing issues for a few releases I think it should be safe though I haven't verified the issue itself. TPM, WDYT?
Project Member

Comment 12 by sheriffbot@chromium.org, Dec 11

Labels: -Merge-Request-72 Merge-Review-72 Hotlist-Merge-Review
This bug requires manual review: Reverts referenced in bugdroid comments after merge request.
Please contact the milestone owner if you have questions.
Owners: govind@(Android), kariahda@(iOS), djmm@(ChromeOS), abdulsyed@(Desktop)

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
Labels: -Merge-Review-72 Merge-Approved-72
Project Member

Comment 14 by bugdroid1@chromium.org, Dec 11

Labels: -merge-approved-72 merge-merged-3626
The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20

commit 3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20
Author: Sahel Sharify <sahel@chromium.org>
Date: Tue Dec 11 21:28:04 2018

Revert "Register fling scheduler on GSB instead of GFS."

Revert "Register fling scheduler on GSB instead of GFS." and
Revert "Fixed fling regression from registering observer on GSB instead of GFS."

This reverts the following two commits:
0b48b62c4392e7cd594d9fbe8684e93abc72aa1b
0e58f2caab6a230fdb0f2b1993a44c2f4546c2f5

The reason for reverting is that registering fling scheduler observer on GSB
instead of GFS does not fix the one frame jank at the beginning of the
fling. This change also caused some performance regressions and touchpad
scrolling regression on eve.

Bug: 882907, 911975 , 911334 ,903161,903156, 903155 
Change-Id: I6aba2c0314eeab616fcb1c18c448c0b3c075d4e6
Reviewed-on: https://chromium-review.googlesource.com/c/1363893
Reviewed-by: Dave Tapuska <dtapuska@chromium.org>
Commit-Queue: Sahel Sharify <sahel@chromium.org>
Cr-Original-Commit-Position: refs/heads/master@{#614730}(cherry picked from commit 0996ec8f771b7c0cc28aa56b201abf3ace2d4176)
Reviewed-on: https://chromium-review.googlesource.com/c/1372714
Reviewed-by: David Bokan <bokan@chromium.org>
Cr-Commit-Position: refs/branch-heads/3626@{#262}
Cr-Branched-From: d897fb137fbaaa9355c0c93124cc048824eb1e65-refs/heads/master@{#612437}
[modify] https://crrev.com/3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20/content/browser/renderer_host/input/fling_controller.cc
[modify] https://crrev.com/3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20/content/browser/renderer_host/input/fling_controller.h
[modify] https://crrev.com/3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20/content/browser/renderer_host/input/fling_controller_unittest.cc
[modify] https://crrev.com/3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20/content/browser/renderer_host/input/fling_scheduler.cc
[modify] https://crrev.com/3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20/content/browser/renderer_host/input/fling_scheduler.h
[modify] https://crrev.com/3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20/content/browser/renderer_host/input/fling_scheduler_android.cc
[modify] https://crrev.com/3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20/content/browser/renderer_host/input/fling_scheduler_android.h
[modify] https://crrev.com/3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20/content/browser/renderer_host/input/fling_scheduler_unittest.cc
[modify] https://crrev.com/3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20/content/browser/renderer_host/input/gesture_event_queue.cc
[modify] https://crrev.com/3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20/content/browser/renderer_host/input/gesture_event_queue_unittest.cc
[modify] https://crrev.com/3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20/content/browser/renderer_host/input/mock_input_router_client.h

Labels: -Pri-0 Pri-1
Lowering priority since the revert got merged to 72 branch.
Labels: -ReleaseBlock-Beta
And -RBB. Will check with sahel@ when she returns if there's anything more to do here.
Status: Fixed (was: Started)
I am sorry I did not respond in time, I lost my access to my corp account on my phone. Thank you bokan@ for merging the revert.
Status: Verified (was: Fixed)
Thanks for reverting and merging. I can confirm that I no longer see these complete pauses on an eve device with Chrome 73.0.3638.0 and Chrome OS 11395.0.0.
I can also confirm that the issues seem to have vanished in Version 72.0.3626.15 (Official Build) dev (64-bit). Touchpad operation is buttery smooth now.
Also seems to have addressed issue 889707, where I first posted about this situation?
Labels: Merge-Merged-72-3626
The following revision refers to this bug: 
https://chromium.googlesource.com/chromium/src.git/+/3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20

Commit: 3cc06e54913c18ff3a45e9dda912bc7b5e5d5e20
Author: sahel@chromium.org
Commiter: bokan@chromium.org
Date: 2018-12-11 21:28:04 +0000 UTC

Revert "Register fling scheduler on GSB instead of GFS."

Revert "Register fling scheduler on GSB instead of GFS." and
Revert "Fixed fling regression from registering observer on GSB instead of GFS."

This reverts the following two commits:
0b48b62c4392e7cd594d9fbe8684e93abc72aa1b
0e58f2caab6a230fdb0f2b1993a44c2f4546c2f5

The reason for reverting is that registering fling scheduler observer on GSB
instead of GFS does not fix the one frame jank at the beginning of the
fling. This change also caused some performance regressions and touchpad
scrolling regression on eve.

Bug: 882907, 911975 , 911334 ,903161,903156, 903155 
Change-Id: I6aba2c0314eeab616fcb1c18c448c0b3c075d4e6
Reviewed-on: https://chromium-review.googlesource.com/c/1363893
Reviewed-by: Dave Tapuska <dtapuska@chromium.org>
Commit-Queue: Sahel Sharify <sahel@chromium.org>
Cr-Original-Commit-Position: refs/heads/master@{#614730}(cherry picked from commit 0996ec8f771b7c0cc28aa56b201abf3ace2d4176)
Reviewed-on: https://chromium-review.googlesource.com/c/1372714
Reviewed-by: David Bokan <bokan@chromium.org>
Cr-Commit-Position: refs/branch-heads/3626@{#262}
Cr-Branched-From: d897fb137fbaaa9355c0c93124cc048824eb1e65-refs/heads/master@{#612437}

Sign in to add a comment