Thanks for the small testcase.

I think this is a Skia bug. The testcase doesn't look minimized but it actually is: removing or moving any of the paths seems to fix this.

Florin, could you triage this?

@pdr were you able to repro on something other than iMac 2017 27"?

I'm able to reproduce on a highdpi macbook pro (MacOS 10.13.3, macbook pro 15in 2016). If I pass --disable-gpu, it does not reproduce. I tried my usual flags but couldn't easily get it to repro on linux.

Strange, I cannot repro on a similar config: 15" 2016 MBP 10.13.3.

Punting to Brian for further triage.

Attaching about:gpu in case it is needed. I'm able to repro in Chrome Dev and Chrome Canary and content shell, so it's unlikely to be related to finch/etc.

Deleted: aboutgpu.htm 63.7 KB

aboutgpu.htm 63.7 KB View Download

Greg is the lucky GPU wrangler.

A-ha:

My about://gpu

GPU0	VENDOR = 0x1002, DEVICE= 0x67ef
GPU1	VENDOR = 0x8086, DEVICE= 0x191b *ACTIVE*

pdr's

GPU0	VENDOR = 0x1002, DEVICE= 0x67ef *ACTIVE*
GPU1	VENDOR = 0x8086, DEVICE= 0x191b

Looks like I'm using the integrated (IntelHD?) GPU while the issue repros with the discrete card (RadeonPro).

@pdr: did you do anything special to make Chrome use the discrete GPU?

Hmm I have an AMD Radeom Pro 560 which does not repo the issue

Nice! That's probably it.

I have automatic graphics switching disabled (settings > energy saver > uncheck "Automatic graphics switching"). I'm also plugged into a power source.

I can now repro after forcing Chrome to use the dedicated GPU (disabled "Automatic graphics switching" in Settings).

So it seems to be a RadeonPro driver issue.  I'll see if bisecting yields anything interesting.

Radeon Pro 460:

  Chipset Model:	AMD Radeon Pro 460
  Type:	GPU
  Bus:	PCIe
  PCIe Lane Width:	x8
  VRAM (Total):	4 GB
  Vendor:	AMD (0x1002)
  Device ID:	0x67ef
  Revision ID:	0x00c0
  ROM Revision:	113-C980AF-908
  VBIOS Version:	113-C9801AU-029
  EFI Driver Version:	01.00.908
  Automatic Graphics Switching:	Supported
  gMux Version:	4.0.29 [3.2.8]
  Metal:	Supported, feature set macOS GPUFamily1 v3
  Displays:
Color LCD:
  Display Type:	Built-In Retina LCD
  Resolution:	2880 x 1800 Retina
  Framebuffer Depth:	24-Bit Color (ARGB8888)
  Main Display:	Yes
  Mirror:	Off
  Online:	Yes
  Rotation:	Supported
  Automatically Adjust Brightness:	No

You are probably looking for a change made after 505839 (known good), but no later than 505859 (first known bad).
CHANGELOG URL:
  https://chromium.googlesource.com/chromium/src/+log/2214bd4a3a01c5830d7a26afb4fe20c2dd9ca134..382773265bc845437e2b2abb050768de08ebea3b

Suspecting https://chromium.googlesource.com/chromium/src/+/e6db79b65553a1d4907217f8aaa7cb589ac976c6 rather than the Skia roll, because disabling MSAA (--gpu-rasterization-msaa-sample-count=0) fixes the issue.

Eric, is MSAA supposed to be disabled for this config and for some reason Skia's MSAA status doesn't reflect that?

FYI, we narrowed this down to MSAA + tessellating path renderer and can repro in Skia.  I'll open a separate Skia issue.

For info, this is the program which triggered the problem:
https://rosek86.github.io/smith-app/

There is mouseover cursor, once I change cursor position green circle flickers. I was hoping to solve this by changing circle to a 'path' (two arcs) but it seems that half of one arc flickers too (red circle). Is this a separate problem?

Due to issues with dynamic switching, Chrome typically blacklists MSAA for both GPUs if either is unsupported. Skia shouldn't have these same issues, as we re-create the GrContext on GPU switch. Because of this, we were enabling MSAA on multi-GPU devices when the discrete GPU is in used where we may have previously suppressed it.

It seems like we should just fix the Skia issue (or blacklist MSAA for this device in Skia) - Chrome never intended to blacklist MSAA on this device, and we probably had this bug before my change on single-GPU AMD systems.

Assigning to senorblanco@, as I think the only thing to do here is close this out when the blocking Skia bug is fixed.

Should be fixed as of the Skia roll including Skia 531a48ed788c5fabfe21704286b54d7567f35469
aka Chrome aa14a6793133f867d63bb3154866c8ae1dfe8c6a.

 Issue 847080  has been merged into this issue.

Requesting merge of Skia change https://crrev.com/ to M68. This is a 
fix for a rendering bug that has been reported against multiple sites. The fix is is simple, safe, and has been in Canary since June 1.

This bug requires manual review: M68 has already been promoted to the beta branch, so this requires manual review
Please contact the milestone owner if you have questions.
Owners: cmasso@(Android), kariahda@(iOS), bhthompson@(ChromeOS), abdulsyed@(Desktop)

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

Approved for M68. Branch:3440

This issue has been approved for a merge. Please merge the fix to any appropriate branches as soon as possible!

If all merges have been completed, please remove any remaining Merge-Approved labels from this issue.

Thanks for your time! To disable nags, add the Disable-Nags label.

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

Sorry, this was merged to M68 as https://skia.googlesource.com/skia.git/+/a4f4c3f16123741b900bcd5e29305e183b020a56. Since it was a Skia merge, bugdroid didn't notice. Removed the Merge-Approved-68 label.