Manual regression range:
Good Build: 56.0.2886.0
Bad Build: 56.0.2888.0

Note: Issue not seen on Mac OS.

Deleted: Download.png 29.6 KB

	Download.png 29.6 KB View Download

You are probably looking for a change made after 424608 (known good), but no later than 424609 (first known bad).

CHANGELOG URL:
-----------------
https://chromium.googlesource.com/chromium/src/+log/751e1f52e94d3d611e27571de632b6a8faa7fc34..42c6683729d8e049fdcc89e552bdcff5fb337227

From the CL above, assigning the issue to the concern owner 

@eugenebut - Could you please check whether this is caused with respect to your change, if not please help us in assigning it to the right owner.

Thanks!

This is very likely caused by my CL. Trent, could you please find another owner for this issue (hopefully I did not break too many things). Thanks!

Adding RB label as this is a recent regression

-> bsep

eugene's CL in that link (https://codereview.chromium.org/2404783002 -> r424609) only affects Mac. But this affects other platforms. I think the bisect went awry - can we do it again?

This looks like a blink issue - my first guess from the coarse bisect in https://chromium.googlesource.com/chromium/src/+log/56.0.2886.0..56.0.2888.0?pretty=fuller&n=10000 is r424596 -> https://codereview.chromium.org/2386033002 "Scale focus outline thickness with zoom level."

There's also r424580 -> https://codereview.chromium.org/2411073002 "[css-flexbox] Remove unnecessary call to adjustAlignmentForChild".

Hmm, I can see how my patch would have caused this. Since I increased the focus ring size maybe it's bleeding outside the area that's damaged on a focus change, so it's not being redrawn?

Strangely enough for me it's ONLY happening at 300% zoom exactly, not any larger or smaller zoom level.

I'm willing to investigate but this is somewhat outside my area of knowledge. If anyone has any pointers that would be very helpful.

I can see this on other webpages not just the downloads page, so I'm changing the components.

Above issue is still reproducible on latest canary version i.e 	56.0.2901.0 (Official Build) 

Please find attached screencast

Deleted: Actual_Result.mp4 409 KB

	Actual_Result.mp4 409 KB View Download

With response to comment #6 & #7

Below is narrow bisect info:
https://chromium.googlesource.com/chromium/src/+log/b5c99e8aa282fdcf0c12bae48ab556cfaff0b74c..f29fc457dc4b6663998a40958f782bfbb89d6dc0?pretty=fuller&n=30

Issue is still reproducible on win10 latest canary version 56.0.2905.0

bsep@, Could you please look  into this or route to appropriate dev for further investigation.

I have a CL fixing it but having some trouble landing it, so hold on: https://codereview.chromium.org/2460683002

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/b18b97a7da62e85fc08eddaefbc95b49c4d7b648

commit b18b97a7da62e85fc08eddaefbc95b49c4d7b648
Author: bsep <bsep@chromium.org>
Date: Tue Nov 01 23:33:17 2016

Fix focus ring damage sometimes being 1 pixel off.

crrev.com/2386033002 changed the focus ring stroke from being 1 pixel to
being scaled based on the zoom level, which changed the type to a float.
When calculating the outdent (and thus damage) the width value lost
precsion and so at certain fractional zoom levels would leave a 1 pixel
line behind. This patch ceils the width value instead, to make sure that
pixel is redrawn when necessary.

R=bokan@chromium.org
BUG= 658648 

Review-Url: https://codereview.chromium.org/2460683002
Cr-Commit-Position: refs/heads/master@{#429153}

[modify] https://crrev.com/b18b97a7da62e85fc08eddaefbc95b49c4d7b648/third_party/WebKit/Source/core/style/ComputedStyle.cpp
[modify] https://crrev.com/b18b97a7da62e85fc08eddaefbc95b49c4d7b648/third_party/WebKit/Source/core/style/ComputedStyleTest.cpp
[modify] https://crrev.com/b18b97a7da62e85fc08eddaefbc95b49c4d7b648/third_party/WebKit/Source/core/style/StyleBackgroundData.h