LayoutTests/css2.1/t040103-ident-03-c.html was wrong, the test and its baselines need to be fixed.
Created attachment 42892 [details] Proposed Patch
> LayoutTests/css2.1/t040103-ident-03-c.html was wrong Why? Please document the reasons here for later reference. This is an official test from W3C, has a fix been made upstream? Or is there a bug filed in W3C bugzilla?
Could you point me where this test is copied from in W3C? I don't know this test is from W3C so not sure whether it is fixed upstream. I think this test is wrong so it should be fixed.
Googling for "t040103-ident-03-c" gives a number of hits, including a discussion of its correctness in <http://lists.w3.org/Archives/Public/www-style/2007Aug/0128.html>. A version of this test in <http://www.w3.org/Style/CSS/Test/Print/1.0/20070802/t040103-ident-03-c.htm> seems to have the change incorporated, but I don't know if it's current. <http://www.w3.org/Style/CSS/Test/CSS2.1/current/> links to a seriously refactored version of the test suite, where this test evolved into <http://www.w3.org/Style/CSS/Test/CSS2.1/current/html4/ident-003.htm>. You need to check with someone involved with the CSS 2.1 working group for the exact status of their test suite. Maybe we should just replace the whole css2.1 directory with a more current version.
Comment on attachment 42892 [details] Proposed Patch We've historically not modified these files directly. Given ap's comments above, marking this as r-. I agree, we should probably just update our copy of css2.1
Good to know, Thanks ap and eric for your info!
Are someone working to update our copy of CSS2.1 test suite? If not, I'm happy to look into this.
Feel free to take this if you have a chance.
It seems W3C released a new alpha version of test suite. http://www.w3.org/Style/CSS/Test/CSS2.1/20100127/html4/index.htm I'm not sure if we want to just update css2.1 directory because the number of test cases significantly increased (from ~400 to ~8200). If we want to avoid increasing the test running time, one idea I came up with is - update existing tests with the new release, - add new tests and their expected results into somewhere different directory, and - modify run-webkit-tests not to run the new tests by default Any suggestions will be appreciated, thanks!
*** Bug 23269 has been marked as a duplicate of this bug. ***
Hmm, for a few of the LayoutTests/css2.1 tests where I've seen issues I've opened up the source and sees a comment that says why the test is wrong [1]. Could we such a comment to the test? We could open a separate bug for updating the entire suite. There is also test.html.disabled if we wanted to skip it, but I'm all for leaving failing "apparent" failing results checked in. [1]: http://trac.webkit.org/browser/trunk/LayoutTests/css2.1/t1504-c543-txt-decor-00-d-g.html#L28
This test is still wrong and expected png is committed for all ports. It is annoying a little bit that I see "This should be green" in red as expected. :) Is there any will to fix it in the future?