the above test refers to an <a href="/sometestfile.html"> and checks that the resulting toString() produces valid output. If the test is being run over a filesystem, this will translate to "/sometestfile.html" on most unix systems, but on Windows this is actually implementation-dependent, and could legitimately return either "file:///sometestfile.html" or "file://C:/sometestfile.html" (or another drive letter). Since this test doesn't really seem to be testing the semantics of relative-URL parsing on file: URLs, I think it makes sense to change it to "http://localhost/sometestfile.html" to get something platform-independent.
Created attachment 32541 [details] patch to make test case less platform-specific
Comment on attachment 32541 [details] patch to make test case less platform-specific How exactly is file:/// platform-specific? Are there platforms that don't use file URLs?
Comment on attachment 32541 [details] patch to make test case less platform-specific > + * fast/dom/anchor-toString-expected.txt: > + * fast/dom/anchor-toString.html: Strange indentation here. r=me Sorry, I missed the comment explaining what "platform-dependent" meant. I think the comment would have been a lot clearer if it just specifically called out Windows drive letters in URLs as the issue.
(In reply to comment #0) > Since this test doesn't really seem to be testing the semantics of relative-URL > parsing on file: URLs What do you mean about relative-URL parsing? Aren't these absolute URLs?
The test case has <a href="/sometestfile.html">, which is a relative URL, not an absolute URL. You are right that the comment probably would've been clearer if I'd just mentioned windows drive letters. Also, the formatting in the changelog was off due to tabs not expanding properly, I think.
Landed as http://trac.webkit.org/changeset/45722.