Issue metadata
Sign in to add a comment
|
Regression: Screen appears black after launching chrome with version #73.0.3639.1 |
||||||||||||||||||||||||
Issue description
Chrome Version: 73.0.3639.1 (cohort: 71_Win_98)Revision5c6469ade1252bc3636320e8865710cb1bafa4ce-refs/branch-heads/3639@{#6}(32/64 bit)
OS: Windows-7
What steps will reproduce the problem?
(1) Launch chrome version 73.0.3639.1 and observe.
Actual: Screen appears black after launching chrome.
Expected: Screen should not appear black.
This is a windows-7 specific regression issue,broken in 'M73', will soon update other info:
Good Build: 73.0.3638.0
Bad Build: 73.0.3639.1
Note:
1. Issue is not seen on Windows(8,8.1,10),Mac(10.13.1,10.13.6,10.14.2) and Linux(14.04 LTS) OS.
2. Issue is specific on Windows 7 with (1600*900) screen resolution.
,
Dec 14
Just to update: 1. Not able to reproduce this issue on latest canary #73.0.3640.0 2. Not able to reproduce the issue while disabling the below flag "enable-blink-gen-property-trees" from chrome://flags Please refer attached screen-cast for the same. Thank you.
,
Dec 14
Adding release blocker label for this issue.Please reduce priority or remove if not the case. Note: 1. Issue able to reproduce in Dev RC #73.0.3639.1 2. Issue is seen with default setting hence adding dev blocker Thank You!
,
Dec 14
Suspect in C#1 could be unrelated. +pdr@ as the issue seems to be resolved by disabling the 'blink-gen-property-trees' flags.
,
Dec 14
I agree that this is more likely to be BlinkGenPropertyTrees so I'll steal this from Bruce. I need a little more information to make progress on this. dipteeb or ajha, on the Win7 machine, can you go to chrome://gpu, copy the results to the clipboard, and paste the results here? Can you also confirm: Go to chrome://flags and enable "BlinkGenPropertyTrees", restart chrome, and the initial screen is black. Go to chrome://flags and disable "BlinkGenPropertyTrees", restart chrome, and the initial screen is white.
,
Dec 14
Stephen, from the videos this seems like it could be the default layer background color being different with BlinkGenPropertyTrees. Do you think that is possible? If so, can you point me to where the code is?
,
Dec 14
Since the issue occurs when the flag is enabled, pdr@/schenney@ do you think its a blocker for dev release? Please confirm ASAP, we are planning to deploy 73.0.3639.1 to dev soon.
,
Dec 14
The flag has been disabled in finch and for users with experimental web platform features enabled (see: https://crbug.com/913464), so I do not think this should block dev.
,
Dec 14
,
Dec 14
Assigning to our Win7 expert! dipteeb or ajha, can you respond to comment #5?
,
Dec 14
The view background is painted by ViewPainter. https://codesearch.chromium.org/chromium/src/third_party/blink/renderer/core/paint/view_painter.cc?rcl=fd8ad21305da505a82ed7ab6dc3c49738eaa802e&l=94 Interestingly, the first line is an early-out TODO that you have flagged, so maybe the code does in fact get hit and it's wrong (i.e. we paint nothing and get the black default graphics backing, which may also explain why it's Windows only). Otherwise there are a lot of callers to set the background color.
,
Dec 14
I have not been able to reproduce, as I have only extremely limited access to a Win7 machine (via browserstack). Is the behavior that the browser is black and never recovers, or that it is briefly black during load? It's not clear from the video.
,
Dec 17
Update : Retested above issue on Windows-7 OS using latest Canary #73.0.3643.0 issue is not reproducible. Screen does not appear black on Launching chrome. Kindly review the attached screen-cast. Thank you!
,
Dec 17
My change was reverted shortly after it landed. My change slightly altered how all of Chrome was built for Windows so it *could* have broken anything, but it's hard to predict. Maybe it could explain why the problem went away? But it's hard to imagine how my change could cause a regression that would only happen on a particular OS and particular resolution. I will be relanding my change so I'll watch for this, in case my change is at fault.
,
Dec 17
shrutij, can you go to chrome://flags, enable "BlinkGenPropertyTrees", restart chrome, and then test if initial screen is black?
,
Dec 19
@shrutij or ajha, ping for comment #15? This should be fixed on trunk because we disabled BlinkGenPropertyTrees by default, but we would like to know if this is still reproducable with BlinkGenPropertyTrees enabled.
,
Dec 20
With respect to comment 15: Restart chrome by enabling "BlinkGenPropertyTrees" flag from chrome://flags and initially screen is not seen black. Please refer attached screen cast for the same: Thank you.
,
Dec 20
brucedawson, I think this may have been your change after all. Unless a mistake was made in reproducing this, it seems to not be related to BlinkGenPropertyTrees after all.
,
Dec 28
,
Dec 28
I've marked this as blocking the toolchain switch and I will investigate. Windows 7 only, such fun...
,
Jan 2
I built Chrome with my change patched in and ran in on Windows 7 and saw no problems. I was able to browse to multiple sites. I will leave this bug open for now and when I reland will see if we can reproduce the problem.
,
Jan 8
The angle failures are almost certainly from the same cause. Tracking this in bug 919163 because there are more details there. |
|||||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||||
Comment 1 by dipt...@virtusa.com
, Dec 14Labels: hasbisect
Owner: brucedaw...@chromium.org
Status: Assigned (was: Unconfirmed)