Add new filter repaint test for turbulence, rebaseline existing test

third_party/WebKit/LayoutTests/TestExpectations

 # ====== Oilpan-only failures from here ======
 # Most of these actually cause the tests to report success, rather than
 # failure. Expected outputs will be adjusted for the better once Oilpan
 # has been well and truly enabled always.
 
 crbug.com/420008 inspector/tracing/worker-js-frames.html [ Pass Timeout Failure ]
 crbug.com/420008 virtual/threaded/inspector/tracing/worker-js-frames.html [ Pass Timeout Failure ]
 crbug.com/596940 inspector-protocol/runtime/runtime-execution-contexts-events.html [ Pass Timeout ]
 crbug.com/597572 [ Debug ] http/tests/inspector/resource-tree/resource-tree-reload.html [ Pass Failure ]
 
@@ skipping 552 matching lines @@
 
 # Test sometimes flakes on most platforms
 crbug.com/571531 imported/csswg-test/css-flexbox-1/css-flexbox-height-animation-stretch.html [ Pass Failure ]
 
 crbug.com/637055 fast/css/outline-offset-large.html [ Skip ]
 
 crbug.com/529938 virtual/spv2/fast/block/positioning/absolute-length-of-neg-666666.html [ Skip ]
 
 crbug.com/617785 fast/borders/block-mask-overlay-image-outset.html [ Skip ]
 
-# Some SVG tests fail due to incorrect visual rects.
-crbug.com/598051 css3/filters/effect-reference-hidpi.html [ Skip ]
+crbug.com/640264 css3/filters/effect-reference-hidpi.html [ NeedsRebaseline ]
+crbug.com/640264 fast/repaint/turbulence-filter-repaint.html [ NeedsRebaseline ]

[Stephen White, 2016/08/30 13:50:15]

Could you move this to css3/filters? There are a bunch of repaint tests 
there as effect-reference-repaint-foo.html. You could also rewrite it in 
the form of those tests if you like (or not, not a big deal). They do 
have reftest results, which is kind of nice.

[pdr., 2016/08/30 16:45:39]

Done, I've rewritten the test as css3/filters/filter-repaint-turbulence.html. I
think the text+pixel results (testRunner.dumpAsTextWithPixelResults()) will be
useful for this test because it shows where the underinvalidation occurs. I can
switch it to just be text output if you'd like though.

[Stephen White, 2016/08/30 17:05:51]

Sounds good. We can always flip it over to text-only when it's fixed.

 
 # Either "combo" or split should run: http://testthewebforward.org/docs/css-naming.html
 crbug.com/410320 imported/csswg-test/css-writing-modes-3/orthogonal-parent-shrink-to-fit-001.html [ Skip ]
 crbug.com/492664 imported/csswg-test/css-writing-modes-3/text-orientation-script-001.html [ Skip ]
 
 # These tests pass but images do not match because of wrong half-leading in vertical-lr
 crbug.com/537080 imported/csswg-test/css-writing-modes-3/vertical-alignment-003.xht [ Failure ]
 crbug.com/537080 imported/csswg-test/css-writing-modes-3/vertical-alignment-009.xht [ Failure ]
 
 # These tests fail / timeout until https://www.chromestatus.com/features/5643236399906816 ships
@@ skipping 698 matching lines @@
 crbug.com/638618 inspector/editor/text-editor-search-switch-editor.html [ Skip ]
 
 # This test breaks when the feature introduced in the bug is enabled. We keep
 # this tested without the feature by using the virtual test. See
 # virtual/stable/http/tests/navigation/beacon-cross-origin-redirect-blob-expected.txt
 crbug.com/490015 http/tests/navigation/beacon-cross-origin-redirect-blob.html [ Skip ]
 # This test doesn't need to be run with the stable release mode.
 crbug.com/490015 virtual/stable/http/tests/navigation/beacon-blob-with-non-simple-type.html [ Skip ]
 # This test fails with the stable release mode.
 crbug.com/490015 virtual/stable/http/tests/navigation/same-and-different-back.html [ Skip ]