210314 - [FIX]Need to properly handle attribute selectors with wildcard namespace

Status: RESOLVED FIXED

(Core :: CSS Parsing and Computation, defect, P2)

Description

[Boris Zbarsky [:bzbarsky]]

For some CSS selectors, the matching code needs to get a list of all attributes
with a given name (no matter what namespace they are in) from a node.  We have
something along those lines with the "attributes" property on nsIDOMNode, and I
guess the style system could use that in a pinch.  Doing things that way is
probably a little slow, though....

Do we want some sort of API on nsIContent for this?

Comment 1

[Boris Zbarsky [:bzbarsky]]

Attachment: testcase

To pass this testcase, we have to be able to examine all attributes named
"attribute", no matter what namespace they are in, and compare their values to
a given string.

Comment 2

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

If we absolutly have to have a "fast" way to do this we should IMHO add a new
function for it. I do not want to make it so that you can pass in
kNameSpaceID_Unknown or some such as namespace to the existing functions.

I think it was a bad desition by CSS group to allow this syntax (mostly because
it's very bad style to have attributes with same name but in different namespace
and give them a same/similar meaning) and I don't think it's worth taking a
performance hit to implement it.

IMHO the style-code could just loop through all attributes and call
GetAttrNameAt() until a good attribute is found 

Comment 3

[Boris Zbarsky [:bzbarsky]]

Ah.  I had missed GetAttrNameAt().  Yeah, that should be just fine for what we
need here.  Morphing and taking to do the style system end.  ;)

Comment 4

[Boris Zbarsky [:bzbarsky]]

A thought.  Perhaps the attr getters should assert if kNamespaceID_Unknown is
passed in for the namespace id, since the behavior is undefined?

Comment 5

[Boris Zbarsky [:bzbarsky]]

Attachment: Patch using GetAttrNameAt

Comment 6

[Boris Zbarsky [:bzbarsky]]

Attachment: Patch to assert if GetAttr is called with kNamespaceID_Unknown

GetAttrNameAt patch checked in.

This one will assert in situations like the one we had here, so that should let
us find other broken callers.

Comment 7

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 126344 [details] [diff] [review]
Patch to assert if GetAttr is called with kNamespaceID_Unknown

jst, would you review?

Comment 8

[Johnny Stenback (:jst)]

Comment on attachment 126344 [details] [diff] [review]
Patch to assert if GetAttr is called with kNamespaceID_Unknown

r+sr=jst

Comment 9

[Boris Zbarsky [:bzbarsky]]

Patch checked in, with similar assertions added to HasAttr.

Comment 10

[Boris Zbarsky [:bzbarsky]]

*** Bug 169905 has been marked as a duplicate of this bug. ***