The issue is that the test failed before the testharness conversion, but with testharness we do not have -expected results, so we can't say that we expect it to fail.

The correct thing to do here is to make the test pass, and leave this bug open with info on what the underlying problem is.

For the record, this is what the test comment says:

    // This test does not currently succeed because skia does not handle
    // canvases more than 32k pixels wide. For now, this test serves the
    // purpose of verifying that this use case does not crash the browser.
    // Crasher bug:  crbug.com/281504 .
    // The result, if Skia handled large canvases, would be [0, 255, 0, 255]

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/142d306b53b0ad88777c23dc9ecebd79487c063f

commit 142d306b53b0ad88777c23dc9ecebd79487c063f
Author: schenney <schenney@chromium.org>
Date: Mon Mar 13 20:57:26 2017

Remove the expectation for fast/canvas/canvas-large-pattern.html

When the test was converted to testharness.js, we lost the expectations
file that recorded the failing result, so now the test fails on the
bots and needs a TestExpectations line. But this is not a transient
expectation, so update the test to record the expected failure and
make sure there's a bug for the underlying problem (which we might
choose to WontFix).

TBR=junov@chromium.org
BUG=696724

Review-Url: https://codereview.chromium.org/2746973004
Cr-Commit-Position: refs/heads/master@{#456477}

[modify] https://crrev.com/142d306b53b0ad88777c23dc9ecebd79487c063f/third_party/WebKit/LayoutTests/TestExpectations
[modify] https://crrev.com/142d306b53b0ad88777c23dc9ecebd79487c063f/third_party/WebKit/LayoutTests/fast/canvas/canvas-large-pattern.html