Bug 16743 – Acid3 expects NodeIterator.nextNode to forward exceptions from the filter

• Component: DOM
• Status: RESOLVED DUPLICATE

[Eric Seidel (no email)]

Test 5: FAIL (method [object NodeIterator].nextNode() didn't forward exception)

    // DOM Traversal
    function () {
      // test 5: NodeFilters and Exceptions
      var doc = getTestDocument();
      var iteration = 0;
      var exception = "Roses";
      var test = function(node) {
        iteration += 1;
        switch (iteration) {
          case 1, 3, 4, 6, 7, 8, 9, 12: throw exception;
          case 2, 5, 10, 11: return true;
          default: throw 0;
        };
      };
      var check = function(o, method) {
        var ok = false;
        try {
          o[method]();
        } catch (e) {
          if (e === exception)
            ok = true;
        }
        assert(ok, "method " + o + "." + method + "() didn't forward exception");
      };
      var i = doc.createNodeIterator(doc.documentElement, 0xFFFFFFFF, test, true);
      check(i, "nextNode"); // 1
      assert(i.nextNode() == doc.documentElement, "i.nextNode() didn't return the right node"); // 2
      check(i, "previousNode"); // 3
      var w = document.createTreeWalker(doc.documentElement, 0xFFFFFFFF, test, true);
      check(w, "nextNode"); // 4
      assert(w.nextNode() == doc.documentElement.firstChild, "w.nextNode() didn't return the right node"); // 5
      check(w, "previousNode"); // 6
      check(w, "firstChild"); // 7
      check(w, "lastChild"); // 8
      check(w, "nextSibling"); // 9
      assert(w.previousSibling() == null, "w.previousSibling() didn't return the right node"); // 10
      assert(w.nextSibling() == null, "w.nextSibling() didn't return the right node"); // 11
      check(w, "previousSibling"); // 12
      return 1;
    },

[Darin Adler]

Our traversal functions are so broken, this is only the tip of the iceberg. But it's straightforward to fix.

[Eric Seidel (no email)]

If you'd like to check in a test case which demonstrates the total brokeness (with lots of FAILs), that's fine by me. :)

[Darin Adler]

(In reply to comment #2)
> If you'd like to check in a test case which demonstrates the total brokeness
> (with lots of FAILs), that's fine by me. :)

Sure, I'll try to find the time to do that.

In the past, I had decided it was pointless to worry about these rarely-used classes, which is why I didn't bother.

The design is fairly tricky -- there are all sorts of edge cases because of the involvement of the filter function and the fact that it can manipulate the very tree its filtering, so writing a test is a bit challenging.

[Darin Adler]

This is an unusual requirement. There's nowhere else in the DOM where we expect exceptions to propagate through the DOM. This causes problems for the language-independent DOM -- it's not necessarily simple to propagate exceptions from any binding language through the DOM.

Are we sure this is a requirement? It's not hard for me to implement, but I'm worried it's not the right thing to do.

[Darin Adler]

This code is incorrect JavaScript:

        switch (iteration) {
          case 1, 3, 4, 6, 7, 8, 9, 12: throw exception;
          case 2, 5, 10, 11: return true;
          default: throw 0;
        };

Those cases are only for the values 12 and 11. It needs to be changed to this:

    switch (iteration) {
        case 1: case 3: case 4: case 6: case 7: case 8: case 9: case 12: throw "Roses";
        case 2: case 5: case 10: case 11: return true;
        default: throw 0;
    }

CC'ing Hixie so he sees this.

[Darin Adler]

I found another error in the test:

      assert(w.previousSibling() == null, "w.previousSibling() didn't return the right node"); // 10

The above line will not call the filter, because there's no node to filter (it's null). So this does not increment the iteration count. I don't know what the contents of the test document are, so I'm not sure about the validity of the next two test steps. I had to change them for my test case.

[Darin Adler]

*** This bug has been marked as a duplicate of 4714 ***

[Ian 'Hixie' Hickson]

I tried to fix the test, let me know if i missed something.
Thanks for the help.

[Lucas Forschler]

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