diff --git a/tests/resources/results.html b/tests/resources/results.html
index db1f73207ff..6797a8185a2 100644
--- a/tests/resources/results.html
+++ b/tests/resources/results.html
@@ -331,14 +331,14 @@ function expandExpectations(expandLink)
     var row = parentOfType(expandLink, 'tr');
     var parentTbody = row.parentNode;
     var existingResultsRow = parentTbody.querySelector('.results-row');
-    
+
     var enDash = '\u2013';
     expandLink.textContent = enDash;
     if (existingResultsRow) {
         updateExpandedState(existingResultsRow, true);
         return;
     }
-    
+
     var newRow = document.createElement('tr');
     newRow.className = 'results-row';
     var newCell = document.createElement('td');
@@ -355,7 +355,7 @@ function expandExpectations(expandLink)
         } else
             result = resultIframe(link.href);
 
-        appendHTML(newCell, result);    
+        appendHTML(newCell, result);
     }
 
     newRow.appendChild(newCell);
@@ -466,7 +466,7 @@ function processGlobalStateFor(testObject)
 
     globalState().hasHttpTests = globalState().hasHttpTests || test.indexOf('http/') == 0;
 
-    var actual = testObject.actual;    
+    var actual = testObject.actual;
     var expected = testObject.expected || 'PASS';
 
     if (actual == 'MISSING') {
@@ -504,7 +504,7 @@ function processGlobalStateFor(testObject)
         globalState().timeoutTests.push(testObject);
         return;
     }
-    
+
     globalState().failingTests.push(testObject);
 }
 
@@ -547,11 +547,11 @@ function imageResultsCell(testObject, testPrefix, actual) {
         globalState().hasImageFailures = true;
 
         if (testObject.reftest_type && testObject.reftest_type.indexOf('!=') != -1) {
-            row += resultLink(testPrefix, '-expected-mismatch.html', 'ref mismatch html');
+            row += resultLink(testPrefix, '-expected-mismatch.sky', 'ref mismatch html');
             row += resultLink(testPrefix, '-actual.png', 'actual');
         } else {
             if (testObject.reftest_type && testObject.reftest_type.indexOf('==') != -1) {
-                row += resultLink(testPrefix, '-expected.html', 'ref html');
+                row += resultLink(testPrefix, '-expected.sky', 'ref html');
             }
             if (globalState().shouldToggleImages) {
                 row += resultLink(testPrefix, '-diffs.html', 'images');
@@ -571,7 +571,7 @@ function imageResultsCell(testObject, testPrefix, actual) {
 }
 
 function tableRow(testObject)
-{    
+{
     var row = '<tbody class="' + (testObject.is_unexpected ? '' : 'expected') + '"';
     if (testObject.reftest_type && testObject.reftest_type.indexOf('!=') != -1)
         row += ' mismatchreftest=true';
@@ -581,7 +581,7 @@ function tableRow(testObject)
 
     var testPrefix = stripExtension(testObject.name);
     row += '<td>';
-    
+
     var actual = testObject.actual;
     if (actual.indexOf('TEXT') != -1) {
         globalState().hasTextFailures = true;
@@ -591,7 +591,7 @@ function tableRow(testObject)
             row += textResultLinks(testObject.name, testPrefix);
         }
     }
-    
+
     if (actual.indexOf('AUDIO') != -1) {
         row += resultLink(testPrefix, '-expected.wav', 'expected audio');
         row += resultLink(testPrefix, '-actual.wav', 'actual audio');
@@ -708,7 +708,7 @@ function testList(tests, header, tableId)
             // FIXME: only include timeout actual/diff results here if we actually spit out results for timeout tests.
             html += textResultLinks(test, stripExtension(test));
         }
-        
+
         if (testObject.has_repaint_overlay)
             html += resultLink(stripExtension(test), '-overlay.html?' + encodeURIComponent(testLinkTarget(test)), 'overlay');
 
@@ -777,7 +777,7 @@ TableSorter._sortRows = function(newHeader, reversed)
         // Only need to support lexicographic sort for now.
         var aText = TableSorter._textContent(a, sortColumn);
         var bText = TableSorter._textContent(b, sortColumn);
-        
+
         // Forward sort equal values by test name.
         if (sortColumn && aText == bText) {
             var aTestName = TableSorter._textContent(a, 0);
@@ -814,11 +814,11 @@ TableSorter.handleClick = function(e)
 TableSorter._sort = function(newHeader)
 {
     TableSorter._updateHeaderClassNames(newHeader);
-    
+
     var reversed = newHeader.classList.contains('reversed');
     var sortArrow = reversed ? TableSorter._backwardArrow : TableSorter._forwardArrow;
     newHeader.innerHTML = TableSorter._sortedContents(newHeader, sortArrow);
-    
+
     TableSorter._sortRows(newHeader, reversed);
 }
 
@@ -839,15 +839,15 @@ PixelZoomer._zoomImageContainer = function(url)
     container.className = 'zoom-image-container';
 
     var title = url.match(/\-([^\-]*)\.png/)[1];
-    
+
     var label = document.createElement('div');
     label.className = 'label';
     label.appendChild(document.createTextNode(title));
     container.appendChild(label);
-    
+
     var imageContainer = document.createElement('div');
     imageContainer.className = 'scaled-image-container';
-    
+
     var image = new Image();
     image.src = url;
     image.style.display = 'none';
@@ -857,7 +857,7 @@ PixelZoomer._zoomImageContainer = function(url)
     imageContainer.appendChild(image);
     imageContainer.appendChild(canvas);
     container.appendChild(imageContainer);
-    
+
     return container;
 }
 
@@ -866,19 +866,19 @@ PixelZoomer._createContainer = function(e)
     var tbody = parentOfType(e.target, 'tbody');
     var row = tbody.querySelector('tr');
     var imageDiffLinks = row.querySelectorAll('a[href$=".png"]');
-    
+
     var container = document.createElement('div');
     container.className = 'pixel-zoom-container';
-    
+
     var html = '';
-    
+
     var togglingImageLink = row.querySelector('a[href$="-diffs.html"]');
     if (togglingImageLink) {
         var prefix = togglingImageLink.getAttribute('data-prefix');
         container.appendChild(PixelZoomer._zoomImageContainer(prefix + '-expected.png'));
         container.appendChild(PixelZoomer._zoomImageContainer(prefix + '-actual.png'));
     }
-    
+
     for (var i = 0; i < imageDiffLinks.length; i++)
         container.appendChild(PixelZoomer._zoomImageContainer(imageDiffLinks[i].href));
 
@@ -935,7 +935,7 @@ PixelZoomer.handleMouseMove = function(e) {
         return;
 
     var container = document.querySelector('.pixel-zoom-container');
-    
+
     var resultContainer = (e.target.className == 'result-container') ?
         e.target : parentOfType(e.target, '.result-container');
     if (!resultContainer || !resultContainer.querySelector('img')) {
@@ -959,7 +959,7 @@ PixelZoomer.handleMouseMove = function(e) {
         PixelZoomer._createContainer(e);
         return;
     }
-    
+
     PixelZoomer._drawAll();
 }
 
@@ -1072,7 +1072,7 @@ TestNavigator._toggleCurrentTestFlagged = function()
 TestNavigator.flagTest = function(testTbody, shouldFlag)
 {
     var testName = testTbody.querySelector('.test-link').innerText;
-    
+
     if (shouldFlag) {
         testTbody.classList.add('flagged');
         TestNavigator.flaggedTests[testName] = 1;
@@ -1250,7 +1250,7 @@ function updateTogglingImages()
 {
     var shouldToggle = document.getElementById('toggle-images').checked;
     globalState().shouldToggleImages = shouldToggle;
-    
+
     if (shouldToggle) {
         forEach(document.querySelectorAll('table:not(#missing-table) tbody:not([mismatchreftest]) a[href$=".png"]'), convertToTogglingHandler(function(prefix) {
             return resultLink(prefix, '-diffs.html', 'images');
diff --git a/tools/webkitpy/layout_tests/controllers/single_test_runner.py b/tools/webkitpy/layout_tests/controllers/single_test_runner.py
index a8ca9739aed..31d4d8cda39 100644
--- a/tools/webkitpy/layout_tests/controllers/single_test_runner.py
+++ b/tools/webkitpy/layout_tests/controllers/single_test_runner.py
@@ -71,7 +71,7 @@ class SingleTestRunner(object):
 
         if self._reference_files:
             # Detect and report a test which has a wrong combination of expectation files.
-            # For example, if 'foo.html' has two expectation files, 'foo-expected.html' and
+            # For example, if 'foo.html' has two expectation files, 'foo-expected.sky' and
             # 'foo-expected.txt', we should warn users. One test file must be used exclusively
             # in either layout tests or reftests, but not in both.
             for suffix in ('.txt', '.png', '.wav'):
@@ -402,22 +402,17 @@ class SingleTestRunner(object):
         if not reference_driver_output.image_hash and not actual_driver_output.image_hash:
             failures.append(test_failures.FailureReftestNoImagesGenerated(reference_filename))
         elif mismatch:
-            if reference_driver_output.image_hash == actual_driver_output.image_hash:
-                diff, err_str = self._port.diff_image(reference_driver_output.image, actual_driver_output.image)
-                if not diff:
-                    failures.append(test_failures.FailureReftestMismatchDidNotOccur(reference_filename))
-                elif err_str:
-                    _log.error(err_str)
-                else:
-                    _log.warning("  %s -> ref test hashes matched but diff failed" % self._test_name)
+            diff, err_str = self._port.diff_image(reference_driver_output.image, actual_driver_output.image)
+            if not diff:
+                failures.append(test_failures.FailureReftestMismatchDidNotOccur(reference_filename))
+            elif err_str:
+                _log.error(err_str)
 
-        elif reference_driver_output.image_hash != actual_driver_output.image_hash:
+        else:
             diff, err_str = self._port.diff_image(reference_driver_output.image, actual_driver_output.image)
             if diff:
                 failures.append(test_failures.FailureReftestMismatch(reference_filename))
             elif err_str:
                 _log.error(err_str)
-            else:
-                _log.warning("  %s -> ref test hashes didn't match but diff passed" % self._test_name)
 
         return TestResult(self._test_name, failures, total_test_time, has_stderr, pid=actual_driver_output.pid)
diff --git a/tools/webkitpy/layout_tests/controllers/test_result_writer_unittest.py b/tools/webkitpy/layout_tests/controllers/test_result_writer_unittest.py
index 213b6779fde..5b9d85ed156 100644
--- a/tools/webkitpy/layout_tests/controllers/test_result_writer_unittest.py
+++ b/tools/webkitpy/layout_tests/controllers/test_result_writer_unittest.py
@@ -54,16 +54,16 @@ class TestResultWriterTests(unittest.TestCase):
 
     def test_reference_exists(self):
         failure = test_failures.FailureReftestMismatch()
-        failure.reference_filename = '/src/exists-expected.html'
-        files = {'/src/exists-expected.html': 'yup'}
+        failure.reference_filename = '/src/exists-expected.sky'
+        files = {'/src/exists-expected.sky': 'yup'}
         written_files = self.run_test(failures=[failure], files=files)
-        self.assertEqual(written_files, {'/tmp/exists-expected.html': 'yup'})
+        self.assertEqual(written_files, {'/tmp/exists-expected.sky': 'yup'})
 
         failure = test_failures.FailureReftestMismatchDidNotOccur()
-        failure.reference_filename = '/src/exists-expected-mismatch.html'
-        files = {'/src/exists-expected-mismatch.html': 'yup'}
+        failure.reference_filename = '/src/exists-expected-mismatch.sky'
+        files = {'/src/exists-expected-mismatch.sky': 'yup'}
         written_files = self.run_test(failures=[failure], files=files)
-        self.assertEqual(written_files, {'/tmp/exists-expected-mismatch.html': 'yup'})
+        self.assertEqual(written_files, {'/tmp/exists-expected-mismatch.sky': 'yup'})
 
     def test_reference_is_missing(self):
         failure = test_failures.FailureReftestMismatch()
diff --git a/tools/webkitpy/layout_tests/models/test_expectations_unittest.py b/tools/webkitpy/layout_tests/models/test_expectations_unittest.py
index 5971a284c5e..2847de0ce57 100644
--- a/tools/webkitpy/layout_tests/models/test_expectations_unittest.py
+++ b/tools/webkitpy/layout_tests/models/test_expectations_unittest.py
@@ -185,9 +185,9 @@ class MiscTests(Base):
         try:
             filesystem = self._port.host.filesystem
             filesystem.write_text_file(filesystem.join(self._port.layout_tests_dir(), 'failures/expected/needsrebaseline.html'), 'content')
-            filesystem.write_text_file(filesystem.join(self._port.layout_tests_dir(), 'failures/expected/needsrebaseline-expected.html'), 'content')
+            filesystem.write_text_file(filesystem.join(self._port.layout_tests_dir(), 'failures/expected/needsrebaseline-expected.sky'), 'content')
             filesystem.write_text_file(filesystem.join(self._port.layout_tests_dir(), 'failures/expected/needsmanualrebaseline.html'), 'content')
-            filesystem.write_text_file(filesystem.join(self._port.layout_tests_dir(), 'failures/expected/needsmanualrebaseline-expected.html'), 'content')
+            filesystem.write_text_file(filesystem.join(self._port.layout_tests_dir(), 'failures/expected/needsmanualrebaseline-expected.sky'), 'content')
             self.parse_exp("""Bug(user) failures/expected/needsrebaseline.html [ NeedsRebaseline ]
 Bug(user) failures/expected/needsmanualrebaseline.html [ NeedsManualRebaseline ]""", is_lint_mode=True)
             self.assertFalse(True, "ParseError wasn't raised")
diff --git a/tools/webkitpy/layout_tests/models/test_run_results_unittest.py b/tools/webkitpy/layout_tests/models/test_run_results_unittest.py
index 99d22462ab7..300bdbfa7e3 100644
--- a/tools/webkitpy/layout_tests/models/test_run_results_unittest.py
+++ b/tools/webkitpy/layout_tests/models/test_run_results_unittest.py
@@ -105,7 +105,7 @@ class InterpretTestFailuresTest(unittest.TestCase):
         self.port = host.port_factory.get(port_name='test')
 
     def test_interpret_test_failures(self):
-        test_dict = test_run_results._interpret_test_failures([test_failures.FailureReftestMismatchDidNotOccur(self.port.abspath_for_test('foo/reftest-expected-mismatch.html'))])
+        test_dict = test_run_results._interpret_test_failures([test_failures.FailureReftestMismatchDidNotOccur(self.port.abspath_for_test('foo/reftest-expected-mismatch.sky'))])
         self.assertEqual(len(test_dict), 0)
 
         test_dict = test_run_results._interpret_test_failures([test_failures.FailureMissingAudio()])
diff --git a/tools/webkitpy/layout_tests/port/base_unittest.py b/tools/webkitpy/layout_tests/port/base_unittest.py
index 494830d3d11..c1f77ebec51 100644
--- a/tools/webkitpy/layout_tests/port/base_unittest.py
+++ b/tools/webkitpy/layout_tests/port/base_unittest.py
@@ -287,10 +287,10 @@ class PortTest(unittest.TestCase):
         self.assertTrue(Port.is_test_file(filesystem, '', 'foo.svg'))
         self.assertTrue(Port.is_test_file(filesystem, '', 'test-ref-test.html'))
         self.assertFalse(Port.is_test_file(filesystem, '', 'foo.png'))
-        self.assertFalse(Port.is_test_file(filesystem, '', 'foo-expected.html'))
+        self.assertFalse(Port.is_test_file(filesystem, '', 'foo-expected.sky'))
         self.assertFalse(Port.is_test_file(filesystem, '', 'foo-expected.svg'))
         self.assertFalse(Port.is_test_file(filesystem, '', 'foo-expected.xht'))
-        self.assertFalse(Port.is_test_file(filesystem, '', 'foo-expected-mismatch.html'))
+        self.assertFalse(Port.is_test_file(filesystem, '', 'foo-expected-mismatch.sky'))
         self.assertFalse(Port.is_test_file(filesystem, '', 'foo-expected-mismatch.svg'))
         self.assertFalse(Port.is_test_file(filesystem, '', 'foo-expected-mismatch.xhtml'))
         self.assertFalse(Port.is_test_file(filesystem, '', 'foo-ref.html'))
@@ -321,7 +321,7 @@ class PortTest(unittest.TestCase):
     def test_reference_files(self):
         port = self.make_port(with_tests=True)
         self.assertEqual(port.reference_files('passes/svgreftest.svg'), [('==', port.layout_tests_dir() + '/passes/svgreftest-expected.svg')])
-        self.assertEqual(port.reference_files('passes/xhtreftest.svg'), [('==', port.layout_tests_dir() + '/passes/xhtreftest-expected.html')])
+        self.assertEqual(port.reference_files('passes/xhtreftest.svg'), [('==', port.layout_tests_dir() + '/passes/xhtreftest-expected.sky')])
         self.assertEqual(port.reference_files('passes/phpreftest.php'), [('!=', port.layout_tests_dir() + '/passes/phpreftest-expected-mismatch.svg')])
 
     def test_operating_system(self):
diff --git a/tools/webkitpy/layout_tests/port/test.py b/tools/webkitpy/layout_tests/port/test.py
index 99f6874e33b..2afbe10e634 100644
--- a/tools/webkitpy/layout_tests/port/test.py
+++ b/tools/webkitpy/layout_tests/port/test.py
@@ -206,22 +206,22 @@ layer at (0,0) size 800x34
               expected_text='\nfoo\n\n', actual_text='\nfoo\r\n\r\r\n')
 
     # For reftests.
-    tests.add_reftest('passes/reftest.html', 'passes/reftest-expected.html', same_image=True)
+    tests.add_reftest('passes/reftest.html', 'passes/reftest-expected.sky', same_image=True)
 
     # This adds a different virtual reference to ensure that that also works.
-    tests.add('virtual/passes/reftest-expected.html', actual_checksum='xxx', actual_image='XXX', is_reftest=True)
+    tests.add('virtual/passes/reftest-expected.sky', actual_checksum='xxx', actual_image='XXX', is_reftest=True)
 
-    tests.add_reftest('passes/mismatch.html', 'passes/mismatch-expected-mismatch.html', same_image=False)
+    tests.add_reftest('passes/mismatch.html', 'passes/mismatch-expected-mismatch.sky', same_image=False)
     tests.add_reftest('passes/svgreftest.svg', 'passes/svgreftest-expected.svg', same_image=True)
-    tests.add_reftest('passes/xhtreftest.xht', 'passes/xhtreftest-expected.html', same_image=True)
+    tests.add_reftest('passes/xhtreftest.xht', 'passes/xhtreftest-expected.sky', same_image=True)
     tests.add_reftest('passes/phpreftest.php', 'passes/phpreftest-expected-mismatch.svg', same_image=False)
-    tests.add_reftest('failures/expected/reftest.html', 'failures/expected/reftest-expected.html', same_image=False)
-    tests.add_reftest('failures/expected/mismatch.html', 'failures/expected/mismatch-expected-mismatch.html', same_image=True)
-    tests.add_reftest('failures/unexpected/crash-reftest.html', 'failures/unexpected/crash-reftest-expected.html', same_image=True, crash=True)
-    tests.add_reftest('failures/unexpected/reftest.html', 'failures/unexpected/reftest-expected.html', same_image=False)
-    tests.add_reftest('failures/unexpected/mismatch.html', 'failures/unexpected/mismatch-expected-mismatch.html', same_image=True)
+    tests.add_reftest('failures/expected/reftest.html', 'failures/expected/reftest-expected.sky', same_image=False)
+    tests.add_reftest('failures/expected/mismatch.html', 'failures/expected/mismatch-expected-mismatch.sky', same_image=True)
+    tests.add_reftest('failures/unexpected/crash-reftest.html', 'failures/unexpected/crash-reftest-expected.sky', same_image=True, crash=True)
+    tests.add_reftest('failures/unexpected/reftest.html', 'failures/unexpected/reftest-expected.sky', same_image=False)
+    tests.add_reftest('failures/unexpected/mismatch.html', 'failures/unexpected/mismatch-expected-mismatch.sky', same_image=True)
     tests.add('failures/unexpected/reftest-nopixel.html', actual_checksum=None, actual_image=None, is_reftest=True)
-    tests.add('failures/unexpected/reftest-nopixel-expected.html', actual_checksum=None, actual_image=None, is_reftest=True)
+    tests.add('failures/unexpected/reftest-nopixel-expected.sky', actual_checksum=None, actual_image=None, is_reftest=True)
     tests.add('reftests/foo/test.html')
     tests.add('reftests/foo/test-ref.html')
 
@@ -237,7 +237,7 @@ layer at (0,0) size 800x34
     tests.add('reftests/foo/second-mismatching-ref.html', actual_checksum='ghi', actual_image='ghi')
 
     # The following files shouldn't be treated as reftests
-    tests.add_reftest('reftests/foo/unlistedtest.html', 'reftests/foo/unlistedtest-expected.html', same_image=True)
+    tests.add_reftest('reftests/foo/unlistedtest.html', 'reftests/foo/unlistedtest-expected.sky', same_image=True)
     tests.add('reftests/foo/reference/bar/common.html')
     tests.add('reftests/foo/reftest/bar/shared.html')
 
diff --git a/tools/webkitpy/layout_tests/run_webkit_tests_unittest.py b/tools/webkitpy/layout_tests/run_webkit_tests_unittest.py
index eb6bcf5a876..32cc45ecf42 100644
--- a/tools/webkitpy/layout_tests/run_webkit_tests_unittest.py
+++ b/tools/webkitpy/layout_tests/run_webkit_tests_unittest.py
@@ -750,16 +750,16 @@ class RunTest(unittest.TestCase, StreamTestingMixin):
         self.assertEqual(['passes/reftest.html'], tests_run)
 
     def test_reftest_expected_html_should_be_ignored(self):
-        tests_run = get_tests_run(['passes/reftest-expected.html'])
+        tests_run = get_tests_run(['passes/reftest-expected.sky'])
         self.assertEqual([], tests_run)
 
     def test_reftest_driver_should_run_expected_html(self):
         tests_run = get_test_results(['passes/reftest.html'])
-        self.assertEqual(tests_run[0].references, ['passes/reftest-expected.html'])
+        self.assertEqual(tests_run[0].references, ['passes/reftest-expected.sky'])
 
     def test_reftest_driver_should_run_expected_mismatch_html(self):
         tests_run = get_test_results(['passes/mismatch.html'])
-        self.assertEqual(tests_run[0].references, ['passes/mismatch-expected-mismatch.html'])
+        self.assertEqual(tests_run[0].references, ['passes/mismatch-expected-mismatch.sky'])
 
     def test_reftest_should_not_use_naming_convention_if_not_listed_in_reftestlist(self):
         host = MockHost()
@@ -777,7 +777,7 @@ class RunTest(unittest.TestCase, StreamTestingMixin):
 
     def test_reftest_with_virtual_reference(self):
         _, err, _ = logging_run(['--details', 'virtual/passes/reftest.html'], tests_included=True)
-        self.assertTrue('ref: virtual/passes/reftest-expected.html' in err.getvalue())
+        self.assertTrue('ref: virtual/passes/reftest-expected.sky' in err.getvalue())
 
     def test_additional_platform_directory(self):
         self.assertTrue(passing_run(['--additional-platform-directory', '/tmp/foo']))
diff --git a/tools/webkitpy/tool/commands/rebaseline_unittest.py b/tools/webkitpy/tool/commands/rebaseline_unittest.py
index ab1ea087c0c..992d170fc75 100644
--- a/tools/webkitpy/tool/commands/rebaseline_unittest.py
+++ b/tools/webkitpy/tool/commands/rebaseline_unittest.py
@@ -263,7 +263,7 @@ Bug(A) [ Debug ] : fast/css/large-list-of-rules-crash.html [ Failure ]
 
     def test_rebaseline_reftest(self):
         self._write("userscripts/another-test.html", "test data")
-        self._write("userscripts/another-test-expected.html", "generic result")
+        self._write("userscripts/another-test-expected.sky", "generic result")
         OutputCapture().assert_outputs(self, self.command._rebaseline_test_and_update_expectations, args=[self.options],
             expected_logs="Cannot rebaseline reftest: userscripts/another-test.html\n")
         self.assertDictEqual(self.command._scm_changes, {'add': [], 'remove-lines': [], "delete": []})