Looks like  bug 355305  is actually covering this and the mentioned skia roll is associated with it, so marking this as a duplicate.

I don't see how 355305 has anything to do with this? The roll in question should fix  bug 355305 .

Original author forgot to cc reed@ as apparently intended.
+junov@ because of display list 2d canvas and general Blinkiness; it's been so long I don't even remember how to get the text diff to see how things are broken, aside from trying to reproduce locally.

Note also virtual/gpu/fast/canvas/DrawImageSinglePixelStretch is now *passing* unexpectedly.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/59e1a445434dc2bd2c9b83ff6c135e9307a3e897

commit 59e1a445434dc2bd2c9b83ff6c135e9307a3e897
Author: vabr <vabr@chromium.org>
Date: Tue Jul 26 14:35:25 2016

Disable virtual/display_list_2d_canvas/fast/canvas/DrawImageSinglePixelStretch.html

The test started failing after the skia roll https://crrev.com/407777.

NOTRY=true
BUG= 631472 , 355305 
TBR=reed@google.com,xidachen@chromium.org

Review-Url: https://codereview.chromium.org/2178383002
Cr-Commit-Position: refs/heads/master@{#407790}

[modify] https://crrev.com/59e1a445434dc2bd2c9b83ff6c135e9307a3e897/third_party/WebKit/LayoutTests/TestExpectations

"DrawImage with a source of a single pixel should draw one uniform color throughout. Neighboring pixels in the source image shouldn't affect the destination rect's output pixels"

Sounds like this test is contradicting  issue 355305  and the recent kFast_SrcRectConstraint changes.

Do we need to update or delete it?

Note: the test was already failing for virtual/gpu/, but the failure is included in the baseline/expected result: https://cs.chromium.org/chromium/src/third_party/WebKit/LayoutTests/virtual/gpu/fast/canvas/DrawImageSinglePixelStretch-expected.txt?rcl=0&l=7

The recent Skia change extends this behavior to the CPU rasterizer.

This test has been labeled failure here: https://codereview.chromium.org/2178383002

Yes, it is due to the recent Skia change. 

It will probably take a bit of time for the bots to cycle in the change in #3.

The failure is  gone now.