http://hg.csswg.org/test/file/5f94e4b03ed9/contributors/opera/submitted/css3-conditional has @supports tests from Opera. Firefox passes all reftests but doesn't fully pass the js test (filed https://bugzilla.mozilla.org/show_bug.cgi?id=816045 ). Opera passes all tests. I think WebKit with @supports enabled doesn't pass all tests.
I also noticed that http://trac.webkit.org/export/131783/trunk/LayoutTests/css3/supports.html has a few fails in Opera and Firefox (apart from the test expecting the quotes to be absent, I think CSSOM requires the quotes). I haven't investigated which tests are correct per spec.
For test 014 i see this in the spec: supports_negation : NOT S* supports_condition_in_parens ; (in fact it's S* all around, i think Tab changed a couple of those that weren't at some point.) The other failing tests are 004, 021, 024, 031, 033. These are legitimate bugs, 004 is a tricky one with nesting, the rest are due to lack of error handling. I'll fix what i can. The fails from the css3/supports.html in opera and firefox are related to whitespace between not/and/or and parens, or because the condition uses a -webkit prefixed property, so there's nothing serious there.
If some tests are wrong per the current spec, I'm happy to fix them, but it would be helpful if you could point out which are wrong so I'm confident about which to fix. :-)
Oh, I glossed over 014 in your comment. Is that the only one?
Fixed 014. http://hg.csswg.org/test/rev/a9c10115a98c
Yeah, i think that's the only one. Thanks!
FTR, my 014 change was wrong per the current spec, but it's possible that the spec will change so that the test becomes valid. https://bugzilla.mozilla.org/show_bug.cgi?id=816045#c6 http://lists.w3.org/Archives/Public/www-style/2012Dec/0084.html
We are tracking separate efforts to improve WPT pass rates, so I don't think this bug is useful now.