70865 - LocationStep isn't attributefriendly

Status: VERIFIED FIXED

(Core :: XSLT, defect)

Description

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

While walking the tree LocationStep isn't attribute frienly. In many places it 
does |node = node->getParentNode()| where node could be an attribute. It should 
rather do |node = cs->getParentNode(node)| .

Also the code for following:: (and preceding::) has bugs. There is an |if| that 
checks if the current node is the last sibling and if so walks to the parent 
following sibling. However the parent may also be the last sibling so we need a 
|while| rather then an |if|

attaching a patch to fix

Comment 1

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

Attachment: patch to fix

Comment 2

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

the attached patch dosn't work... will attach new version once I have some more 
fixes in

Comment 3

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

Attachment: This is tested agains all axes tests in buster

Comment 4

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

the attached patch passes almost all axes tests in buster. The remaing ones I 
afaict are problems with either whitespace or attribute order. Note that this 
depends on bug 70979 for module

Comment 5

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

Attachment: Latest version, fixes some reverse axes walking

Comment 6

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

The patch fixes the following things:

1. The old code didn't work when the contextnode was an attribute. According to
   the XPath datamodel the parent of an attribute is it's element so we need to
   use the ContextState::getParentNode(Node*) function rather then
   Node::getParentNode() when trying to find the parent

2. The FOLLOWING and PRECEDING axes were very broken

3. LocationStep::matches didn't calculate the correct context nodeset when
   evaluating the predicates

Comment 7

[Axel Hecht]

r=[e-mail]

Comment 8

[Mike Shaver (:shaver emeritus)]

+        NodeSet* nodes = (NodeSet*)evaluate(cs->getParentNode(node),cs);
+        result = nodes->contains(node);

Can evaluate return null?  If so, you might want to defend against it.

Other than that, sr=shaver.

Comment 9

[Axel Hecht]

fix checked in

Axel

Comment 10

[Axel Hecht]

bitching buttons, verfication spam