See comment 22 and 25 on this link.
Thomas Ganter seems to be using Leopard whereas I'm using Tiger, so the only common denominator is PPC. I've isolated down to these 3 tests that are reported fixed but still failing on PPC.
To save you some potential headache. I've used Monolingual to remove the extra languages, and stripped the binary of non-G4 architecture binary code.
FWIW, I do not see this on my G4/1.25DP - both a local debug build and r31238 nightly pass these tests, and give a 95/100 overall. Testing on 10.4.11 with Safari 3.0.4.
Test 69 often gives warnings such as "took 44 attempts".
I did see failure of test 77 once, but that is tracked as bug 17672.
For 78, it's probably bug 17063 because the error messages match.
I don't see anything related to 69 or 74 anywhere however.
(In reply to comment #0)
> To save you some potential headache. I've used Monolingual to remove the extra
> languages, and stripped the binary of non-G4 architecture binary code.
This is an unsupported configuration. Results after performing such a change are not valid.
I forgot to mention that running the universal Webkit nightly without stripping also has this problem.
I can't really unstrip my system so I can't test whether it is some other libraries Webkit is referencing is causing the problem or Webkit itself.
Not sure if Thomas Ganter has ran Monolingual. He certainly didn't mention it. If he hasn't, or done anything weird to his binaries, then it's not a problem of Monolingual.
This should be a duplicate of http://bugs.webkit.org/show_bug.cgi?id=17884
*** This bug has been marked as a duplicate of 17884 ***