fast/xpath/namespace-vs-predicate.xhtml fails unexpectedly in no-SVG build

• Component: DOM
• Status: RESOLVED FIXED

[Mark Rowe (bdash)]

http://build.webkit.org/results/trunk-mac-intel-nosvg/2750/results.html shows that fast/xpath/namespace-vs-predicate.xhtml is failing on the no-SVG buildbot.  The actual output contains the following error:
CONSOLE MESSAGE: line 38: TypeError: Null value

This is emitted by the following line of code:
        // Remove this node from the document so it doesn't make the DumpRenderTree output
        // different when SVG is enabled and disabled.
        document.getElementById("r_00").parentNode.removeChild(document.getElementById("r_00"));

As the comment describes, this is intended to ensure that the <svg:g> node is removed to maintain identical rendering in SVG and no-SVG builds.  It is not clear why this is throwing an exception.

[Alexey Proskuryakov]

Do you know when this regressed? I'm not quite sure what kind of a bug this is: Firefox seems to have a similar problem with elements in unknown namespaces.

[Mark Rowe (bdash)]

Sorry I don't.  It's probably been like that for quite a while now.

[Alexey Proskuryakov]

Created attachment 16916 [details]
unknown namespace test case

Firefox and TOT WebKit behave identically on this test (actually, Tiger WebKit does the same). I still don't see why.

[Alexey Proskuryakov]

Created attachment 18150 [details]
fix the test

It's actually clear: we do not support xml:id (see bug 16505), and plain "id" is only treated as ID in HTML and SVG content (unless a doctype says that some attribute is an ID, but it doesn't in this case, and we don't even support that, bug 12971).

A no-SVG build isn't supposed to know about SVG id attribute.

Comment on attachment 18150 [details]
fix the test

r=me

[Alexey Proskuryakov]

Committed revision 29024.

[Lucas Forschler]

Mass moving XML DOM bugs to the "DOM" Component.