Math.LOG10E is now 0.4342944819032518, which matches Firefox and IE. See http://code.google.com/p/chromium/issues/detail?id=72555 Will update LayoutTests/platform/chromium/test_expectations.txt when V8 3.1.4 has landed in chromium.
The layout test failing due to the change to Math.LOG10E is fast/js/kde/math.html.
You added an override in the Chromium-side src/webkit/tools/layout_tests/test_expectations.txt file. Please move that to the WebKit side or fix the test or its expectations as soon as possible. Thanks!
Created attachment 82439 [details] Patch
This patch is incomplete. In order for this to work, the test will have to be changed in order to expect a different value on V8 versus JSC.
The patch adds the file LayoutTests/platform/chromium/fast/js/kde/math-expected.txt which only differs from LayoutTests/fast/js/kde/math-expected.txt in the result of Math.LOG10E is 0.4342944819032518 in the first and 0.43429448190325176 in the second. Should that not cover the difference in the value of Math.LOG10E between V8 and JSC? There is another bug on this, https://bugs.webkit.org/show_bug.cgi?id=41033. According to ES5 section 15.8.1.5 LOG10E has a value of approximately 0.4342944819032518.
(In reply to comment #5) > The patch adds the file > > LayoutTests/platform/chromium/fast/js/kde/math-expected.txt > > which only differs from > > LayoutTests/fast/js/kde/math-expected.txt > > in the result of Math.LOG10E is > > 0.4342944819032518 > > in the first and > > 0.43429448190325176 > > in the second. > > Should that not cover the difference in the value of Math.LOG10E between V8 and JSC? No, because the test explicitly checks the value against 0.43429448190325176. That's why the current output is "FAIL" + explanation instead of the output you attached. Please see the actual test source code at http://trac.webkit.org/browser/trunk/LayoutTests/fast/js/kde/script-tests/math.js. The best patch might be one that changes the test to expect the more-correct value, changes the global expectation to expect failure, and adds a Chromium expectation identical to the one you supplied.
Created attachment 82782 [details] Patch
Sorry for making such a bogus patch. I have now uploaded a patch where both the JSC and the V8 value of LOF10E are present and with different expectations.
Comment on attachment 82782 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=82782&action=review > LayoutTests/fast/js/kde/script-tests/math.js:17 > +// JSC and V8 has different values for Math.LOG10E Nit: has -> return. You might want to add to the end that this is the more accurate value. > LayoutTests/fast/js/kde/script-tests/math.js:19 > shouldBe("String()+Math.LOG10E", "'0.43429448190325176'"); Please remove this line. We should only expect one result. It is OK for JSC to have an expected-fail result for just the first value.
JSC folks might want to see this go by. Seems JSC might want to change here too.
Created attachment 83214 [details] Patch
http://trac.webkit.org/changeset/79246 JSC's value has been changed. Is this patch needed?
(In reply to comment #12) > http://trac.webkit.org/changeset/79246 > JSC's value has been changed. Is this patch needed? No. Someone should fix the Chromium expectations to note that we expect to pass this test.
http://trac.webkit.org/changeset/79246/ aligned its behavior to V8.