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

Issue 875891 link

Starred by 8 users

Issue metadata

Status: Verified
Owner:
Closed: Oct 23
Cc:
EstimatedDays: ----
NextAction: ----
OS: Mac
Pri: 1
Type: Bug-Regression



Sign in to add a comment

M69 regression: Mangled Window UI on MacPro3,1

Project Member Reported by somo...@chromium.org, Aug 20

Issue description

Chrome Version: 69.0.3497.42
OS: macOS

What steps will reproduce the problem?

On a MacPro3,1 (2008), the new material design is not rendering omnibox or any other window-attached UI correctly.

This reproduces identically and 100% reproducibly on the same machine under 10.11.6 and 10.13.6; the former is the last Apple-supported OS version for this hardware, the latter works just fine but is unsupported. 

M68 is fine on this machine.

What is the expected result?

Non-mangled UI. Ability to see omnibox contents. Etc.

What happens instead?

Mangled UI in the window. See attached screenshots of the variations of mangledness.

This machine has the default video card that came with it:

  Chipset Model:	NVIDIA GeForce 8800 GT
  Type:	GPU
  Bus:	PCIe
  Slot:	Slot-1
  PCIe Lane Width:	x16
  VRAM (Dynamic, Max):	512 MB
  Vendor:	NVIDIA (0x10de)
  Device ID:	0x0602
  Revision ID:	0x00a2
  ROM Revision:	3233
 
 
auto.png
46.5 KB View Download
default.png
50.7 KB View Download
hybrid.png
46.7 KB View Download
normal.png
56.0 KB View Download
refresh.png
46.8 KB View Download
touchable.png
47.0 KB View Download
touchablerefresh.png
47.4 KB View Download
Labels: Needs-Triage-M69 Needs-Bisect
Labels: Needs-Feedback
Can you repro with Chrome Canary ?

Thanks in advance.
Labels: -Needs-Feedback
Yes, reproduces on Canary 70.0.3528.0
Cc: swarnasree.mukkala@chromium.org
Labels: Triaged-ET Needs-Feedback
Unable to reproduce the issue on reported chrome version #69.0.3497.42 using MacBook Air OS 10.13.6, by following the steps in comment#0.
Note : MacPro 3,1 is not available at ET team, hence could not check the behaviour on it to reproduce the issue.

Attached screencast for reference.

@Reporter: Could you please review the attached screen-cast and confirm if anything being missed here and request you to retry this issue with fresh profile without any extensions/apps or reset all the flags and let us know if issue still persists.

Thanks.!
875891.mp4
3.3 MB View Download
I can't repro this on a MacBookPro14,2 running 10.13.6 either. It has been consistently reproducible on the MacPro3,1.

I reset all the flags multiple times during my repro that generated the various screen shots. That's definitely not the root cause.

I'm physically not near the machine for the next few days, so I won't be able to try with a fresh profile until I'm back.
Project Member

Comment 6 by sheriffbot@chromium.org, Aug 21

Labels: -Needs-Feedback
Thank you for providing more feedback. Adding the requester to the cc list.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
Labels: TE-NeedsTriageFromMTV
As per comment#5 the issue seems to be related to MacPro3,1 ,as it is not available at TE end.Hence adding TE-NeedsTriageFromMTV label and requesting MTV team to have a look into this if the same can be arranged and tested.

Thanks.!
Cc: ligim...@chromium.org
Labels: Needs-Feedback
MTV also do not have MacPro3,1. 

Stephan, could you please try in Chrome stable and update whether you can still repro the same behavior?

https://www.google.com/chrome/?platform=mac
Per previous requests for more info:

Issue does not reproduce on stable (68.0.3440.106)

Issue consistently reproduces on beta (69.0.3497.57), including with a fresh profile

Issue also consistently reproduces on Canary (70.0.3533.0)

I am growing increasingly concerned given how near M69 is.
Project Member

Comment 10 by sheriffbot@chromium.org, Aug 26

Labels: -Needs-Feedback
Thank you for providing more feedback. Adding the requester to the cc list.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
Cc: sdy@chromium.org
Labels: -Pri-2 Pri-1
Owner: ccameron@chromium.org
Status: Assigned (was: Unconfirmed)
This is oddly similar to what I've been seeing in VMWare Fusion.
Screen Shot 2018-08-27 at 4.16.38 PM.png
2.0 MB View Download
Does this reproduce with any of the following flags (try them independently, many are subsets of the others)
  --disable-mac-overlays
  --disable-remote-core-animation
  --disable-gpu-rasterization
  --disable-gpu

Also, please attach the output of about:gpu (as a PDF) for all affected machines.
Oh, and also run with
  --disable-gpu-memory-buffer-compositor-resources
Results, all using 69.0.3497.57:

--disable-mac-overlays

blank window (see attached)

--disable-remote-core-animation

blank window

--disable-gpu-rasterization

somewhat better, but weird autosuggest list (see attached)

--disable-gpu

no badness visible

--disable-gpu-memory-buffer-compositor-resources

blank window again

about:cpu attached

Also attached is what was written out after quitting the various invocations in case it helps.

disable-mac-overlays.png
8.0 KB View Download
disable-gpu-rasterization.png
58.5 KB View Download
about_gpu_macpro31.txt
8.1 KB View Download
error_spew.txt
4.1 KB View Download
The following flags combination fixes things
 --disable-gpu-rasterization --disable-features=UiGpuRasterization

The bug here is that we've blacklisted GPU raster (in about:gpu)
    Some GPUs on Mac can perform poorly with GPU rasterization. Disable
    all known NVidia GPUs other than the Geforce 6xx and 7xx series, which
    have been tested.:  issue 613272 ,  issue 614468 

But ... somehow UI isn't picking up that blacklist. Umm, that's serious and needs to be fixed. I'm sure it's just 1 line of plumbing that we didn't pick up (but which line?).

sdy@, can you give those flags a try on VMWare?
Confirmed, those two flags also help VMWare.
Project Member

Comment 19 by bugdroid1@chromium.org, Aug 28

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/53d324c3aa89d291dc4f3f095b403dde94883188

commit 53d324c3aa89d291dc4f3f095b403dde94883188
Author: Christopher Cameron <ccameron@chromium.org>
Date: Tue Aug 28 03:33:43 2018

Trigger GPU raster (as opposed to forcing it)

Enabling GPU raster via "trigger" will respect GPU blacklists, while
enabling GPU raster via "force" will not respect GPU blacklists.

We had previously been "forcing" GPU raster, which caused it to be
enabled for on some platforms where it would then spectacularly explode.

Bug:  875891 
Change-Id: I435aa86ee70018aca4eafc7fc057380c5238d430
Reviewed-on: https://chromium-review.googlesource.com/1192545
Reviewed-by: enne <enne@chromium.org>
Commit-Queue: ccameron <ccameron@chromium.org>
Cr-Commit-Position: refs/heads/master@{#586570}
[modify] https://crrev.com/53d324c3aa89d291dc4f3f095b403dde94883188/ui/compositor/compositor.cc

Cc: abdulsyed@chromium.org gov...@chromium.org
This appears not to be in a Canary yet. This will be needing a merge to M69 once it is verified.
NextAction: 2018-08-30
Pls update bug with canary result tomorrow. 
Labels: M-69 ReleaseBlock-Stable Target-70 M-70 Target-69
Adding "RBS" label to track this merge.

Note: We're cutting M69 stable RC tomorrow @ 4:00 PM PT, so if canary result looks good and safe to merge, pls request a merge tomorrow morning.
Project Member

Comment 23 by bugdroid1@chromium.org, Aug 29

Labels: merge-merged-3535
The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/d1f93b058cdfd5ccd3d322d98205336b40690592

commit d1f93b058cdfd5ccd3d322d98205336b40690592
Author: Christopher Cameron <ccameron@chromium.org>
Date: Wed Aug 29 20:19:14 2018

Trigger GPU raster (as opposed to forcing it)

Enabling GPU raster via "trigger" will respect GPU blacklists, while
enabling GPU raster via "force" will not respect GPU blacklists.

We had previously been "forcing" GPU raster, which caused it to be
enabled for on some platforms where it would then spectacularly explode.

TBR=ccameron@chromium.org

(cherry picked from commit 53d324c3aa89d291dc4f3f095b403dde94883188)

Bug:  875891 
Change-Id: I435aa86ee70018aca4eafc7fc057380c5238d430
Reviewed-on: https://chromium-review.googlesource.com/1192545
Reviewed-by: enne <enne@chromium.org>
Commit-Queue: ccameron <ccameron@chromium.org>
Cr-Original-Commit-Position: refs/heads/master@{#586570}
Reviewed-on: https://chromium-review.googlesource.com/1195850
Reviewed-by: ccameron <ccameron@chromium.org>
Cr-Commit-Position: refs/branch-heads/3535@{#13}
Cr-Branched-From: 4d18295073a6bf8f2f4d26e4e059767e091eaaf3-refs/heads/master@{#586474}
[modify] https://crrev.com/d1f93b058cdfd5ccd3d322d98205336b40690592/ui/compositor/compositor.cc

Update: Canary 70.0.3536.0 yields visible improvement re rendering of the address bar, but note the invisible content of the autocomplete drop-down in the attached screen shot.
70.0.3536.0_improvement.png
32.0 KB View Download
Thanks for testing -- does that still happen with the flags:
 --disable-gpu-rasterization --disable-features=UiGpuRasterization

Also, what about just the flag
  --disable-features=UiGpuRasterization
in isolation
... and also
  --disable-features=VizDisplayCompositor
The NextAction date has arrived: 2018-08-30
Now at 70.0.3537.0.

Tested all three separately:

 --disable-gpu-rasterization --disable-features=UiGpuRasterization

 --disable-features=UiGpuRasterization

 --disable-features=VizDisplayCompositor

All three still exhibit the invisible drop-down content from #c24

Labels: Merge-Request-69
Adding a merge request for Chrome 69, as this does indeed improve the situation
Project Member

Comment 31 by sheriffbot@chromium.org, Aug 30

Labels: -Merge-Request-69 Merge-Review-69 Hotlist-Merge-Review
This bug requires manual review: We are only 4 days from stable.
Please contact the milestone owner if you have questions.
Owners: amineer@(Android), kariahda@(iOS), cindyb@(ChromeOS), govind@(Desktop)

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
Labels: -ReleaseBlock-Stable -Merge-Review-69 Merge-Approved-69
Approving merge to M69 branch 3497 based on comment #30 and per offline chat with  ccameron@, this will be a safe merge. Pls merge now. Thank you.

Also removing "RBS" from bug, pls add back if needed. 
Project Member

Comment 33 by bugdroid1@chromium.org, Aug 30

Labels: -merge-approved-69 merge-merged-3497
The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/7a16d1b0751b22d9f4ed298210eecca0a630e222

commit 7a16d1b0751b22d9f4ed298210eecca0a630e222
Author: Christopher Cameron <ccameron@chromium.org>
Date: Thu Aug 30 20:21:15 2018

Trigger GPU raster (as opposed to forcing it)

Enabling GPU raster via "trigger" will respect GPU blacklists, while
enabling GPU raster via "force" will not respect GPU blacklists.

We had previously been "forcing" GPU raster, which caused it to be
enabled for on some platforms where it would then spectacularly explode.

TBR=ccameron@chromium.org

(cherry picked from commit 53d324c3aa89d291dc4f3f095b403dde94883188)

Bug:  875891 
Change-Id: I435aa86ee70018aca4eafc7fc057380c5238d430
Reviewed-on: https://chromium-review.googlesource.com/1192545
Reviewed-by: enne <enne@chromium.org>
Commit-Queue: ccameron <ccameron@chromium.org>
Cr-Original-Commit-Position: refs/heads/master@{#586570}
Reviewed-on: https://chromium-review.googlesource.com/1197224
Reviewed-by: ccameron <ccameron@chromium.org>
Cr-Commit-Position: refs/branch-heads/3497@{#851}
Cr-Branched-From: 271eaf50594eb818c9295dc78d364aea18c82ea8-refs/heads/master@{#576753}
[modify] https://crrev.com/7a16d1b0751b22d9f4ed298210eecca0a630e222/ui/compositor/compositor.cc

Labels: -M-69 -Target-69 ReleaseBlock-Stable
Adding "RBS" back for M70 per chat with ccameron@. 
Hopefully this just affects the omnibox. If this also affects the 3-dot menu, then we should panic more.

sdy@, how is this in your VMWare machine? Is the omnibox also busted, or is everything more-or-less working.
The omnibox is pretty busted, the menu is okay (but pretty slow).
canary_current.mp4
2.0 MB View Download
sdy@, what happens if you also use
  --disable-zero-copy --disable-gpu-memory-buffer-compositor-resources

(and also, is there any error spew in about:gpu)
Way worse with those flags, and spew in about:gpu. No spew that I noticed without the flags. All attached.
about_gpu_vmware_70.0.3537.0.txt
7.9 KB View Download
about_gpu_disable-zero-copy-disable-resources_70.0.3537.0.txt
7.8 KB View Download
disable-zero-copy-disable-resources_70.0.3537.0.mp4
3.1 MB View Download
I think that the spew may have been clipped from the above comment.
I think I misread one of the outputs, so disregard that, sorry. The text attachments in the above comments are faithful to what I see in about:gpu.
s/above comments/above comment/
I also see no debug spew being generated in chrome://gpu when the invisible menu contents phenomenon is happening.
Labels: -TE-NeedsTriageFromMTV -Needs-Bisect
Updating labels.

Comment 45 Deleted

sdy@, one more request from ccameron@  -- can you verify with this RC build #69.0.3497.81?
With that build, the omnibox is still blank (same on Canary), but everything else looks OK.
NextAction: ----
Confirming same on the MacPro3,1.

With the exception of the transparent autocomplete menu, everything looks fine with 69.0.3497.81.
Project Member

Comment 49 by bugdroid1@chromium.org, Sep 11

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/a361a69018d5b46ba3a29f7818e717451a92a09f

commit a361a69018d5b46ba3a29f7818e717451a92a09f
Author: Christopher Cameron <ccameron@chromium.org>
Date: Tue Sep 11 22:58:54 2018

Mac: Blacklist all GPU features for GL <4.1

This is a very small set of machines, which have been exhibiting
- rendering artifacts when GPU compositing is enabled
- severe rendering artifacts when GPU raster is enabled

Bug:  875891 
Cq-Include-Trybots: luci.chromium.try:android_optional_gpu_tests_rel;luci.chromium.try:linux_optional_gpu_tests_rel;luci.chromium.try:mac_optional_gpu_tests_rel;luci.chromium.try:win_optional_gpu_tests_rel
Change-Id: Ie49a9111b9723977a2b8758f824bba7843238d32
Reviewed-on: https://chromium-review.googlesource.com/1220180
Reviewed-by: ccameron <ccameron@chromium.org>
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Reviewed-by: Zhenyao Mo <zmo@chromium.org>
Commit-Queue: ccameron <ccameron@chromium.org>
Cr-Commit-Position: refs/heads/master@{#590516}
[modify] https://crrev.com/a361a69018d5b46ba3a29f7818e717451a92a09f/gpu/config/software_rendering_list.json

Cc: ccameron@chromium.org
 Issue 884340  has been merged into this issue.
Would this bug also affect CSS rendering of rules like backface-visibility and transform? I'm seeing this same Omnibox issue on an older iMac and it's also rendering some elements invisible (they can still be found & outlined in the Dev Tools though – and disabling choice CSS rules makes them visible again). I won't open a new bug if the cause is the same.
[bulk edit] - This issue is marked as a stable blocker for M70. We are two weeks away from M70 Stable. Please take a look urgently!
ccameron@ Do we need to merge this to M70?
Friendly ping to look into this issue and to provide further update on this issue as it has been marked as a stable blocker.

Thanks!
Yes, let's merge r590516 to M70.
ccameron@, can you please merge the above change to M70 branch: 3538?

Thank you!
Project Member

Comment 57 by bugdroid1@chromium.org, Oct 12

Labels: merge-merged-3538
The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/2f179eaa22bc4c40bb438eba31ffe2be210206d8

commit 2f179eaa22bc4c40bb438eba31ffe2be210206d8
Author: Christopher Cameron <ccameron@chromium.org>
Date: Fri Oct 12 20:20:31 2018

Mac: Blacklist all GPU features for GL <4.1

This is a very small set of machines, which have been exhibiting
- rendering artifacts when GPU compositing is enabled
- severe rendering artifacts when GPU raster is enabled

TBR=ccameron@chromium.org

(cherry picked from commit a361a69018d5b46ba3a29f7818e717451a92a09f)

Bug:  875891 
Cq-Include-Trybots: luci.chromium.try:android_optional_gpu_tests_rel;luci.chromium.try:linux_optional_gpu_tests_rel;luci.chromium.try:mac_optional_gpu_tests_rel;luci.chromium.try:win_optional_gpu_tests_rel
Change-Id: Ie49a9111b9723977a2b8758f824bba7843238d32
Reviewed-on: https://chromium-review.googlesource.com/1220180
Reviewed-by: ccameron <ccameron@chromium.org>
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Reviewed-by: Zhenyao Mo <zmo@chromium.org>
Commit-Queue: ccameron <ccameron@chromium.org>
Cr-Original-Commit-Position: refs/heads/master@{#590516}
Reviewed-on: https://chromium-review.googlesource.com/c/1277584
Cr-Commit-Position: refs/branch-heads/3538@{#988}
Cr-Branched-From: 79f7c91a2b2a2932cd447fa6f865cb6662fa8fa6-refs/heads/master@{#587811}
[modify] https://crrev.com/2f179eaa22bc4c40bb438eba31ffe2be210206d8/gpu/config/software_rendering_list.json

Labels: CommitLog-Audit-Violation Merge-Without-Approval
Here's a summary of the rules that were executed: 
 - OnlyMergeApprovedChange: Rule Failed -- Revision 2f179eaa22bc4c40bb438eba31ffe2be210206d8 was merged to refs/branch-heads/3538 branch with no merge approval from a TPM! 
Please explain why this change was merged to the branch!
 - AcknowledgeMerge: Notification Required -- 
Labels: Merge-Merged-70-3538
The following revision refers to this bug: 
https://chromium.googlesource.com/chromium/src.git/+/2f179eaa22bc4c40bb438eba31ffe2be210206d8

Commit: 2f179eaa22bc4c40bb438eba31ffe2be210206d8
Author: ccameron@chromium.org
Commiter: ccameron@chromium.org
Date: 2018-10-12 20:20:31 +0000 UTC

Mac: Blacklist all GPU features for GL <4.1

This is a very small set of machines, which have been exhibiting
- rendering artifacts when GPU compositing is enabled
- severe rendering artifacts when GPU raster is enabled

TBR=ccameron@chromium.org

(cherry picked from commit a361a69018d5b46ba3a29f7818e717451a92a09f)

Bug:  875891 
Cq-Include-Trybots: luci.chromium.try:android_optional_gpu_tests_rel;luci.chromium.try:linux_optional_gpu_tests_rel;luci.chromium.try:mac_optional_gpu_tests_rel;luci.chromium.try:win_optional_gpu_tests_rel
Change-Id: Ie49a9111b9723977a2b8758f824bba7843238d32
Reviewed-on: https://chromium-review.googlesource.com/1220180
Reviewed-by: ccameron <ccameron@chromium.org>
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Reviewed-by: Zhenyao Mo <zmo@chromium.org>
Commit-Queue: ccameron <ccameron@chromium.org>
Cr-Original-Commit-Position: refs/heads/master@{#590516}
Reviewed-on: https://chromium-review.googlesource.com/c/1277584
Cr-Commit-Position: refs/branch-heads/3538@{#988}
Cr-Branched-From: 79f7c91a2b2a2932cd447fa6f865cb6662fa8fa6-refs/heads/master@{#587811}
Labels: -Merge-Without-Approval
Just tested 70.0.3538.67 in VMWare. The omnibox is transparent (I only see the shadow), but other parts of the window, including menus, work fine.
Tested 70.0.3538.67 on my MacPro3,1. Screenshot enclosed. Looks good here.
875891_verify_macpro31.png
18.3 KB View Download
Status: Verified (was: Assigned)

Sign in to add a comment