This is an archived snapshot of W3C's public bugzilla bug tracker, decommissioned in April 2019. Please see the home page for more details.
Test Parenexpr-19 should simply return a copy of the input document. In fact, however, it returns a copy of the input document with whitespace text nodes removed. I think there are many other tests suffering from the same problem. At present I handle this using a comparator that ignores whitespace text nodes in the result (with a comment that this has been done). At the very least, I think results should be published with and without whitespace stripping, or the catalog should say whether whitespace stripping on the source document is required. In fact, however, I think that according to the data model spec, whitespace stripping in this case is incorrect if the standard "construction from an InfoSet" is used. In the mapping from the InfoSet, whitespace text nodes are discarded only if the [element content whitespace] property is true, and this will only be the case if there is a DTD. Other tests showing this problem (in most cases I only did careful checking of the first one in each group; the others are listed because they displayed the same symptoms): Axes085 NodeTest003/4/5 unabbreviatedSyntax-4/5/8/10/11/12/13/20/21/22/24/25/26/28/29/30/31 abbreviatedSyntax-2/3/5/6/14/16/17/18/19/20/21/24 filterexpressionhc1/4/5 fn-union-node-args-002/3/5/7/9/10/11/15/16 fn-intersect-node-args-011/15/16 fn-except-node-args-003/017 ForExpr005 WhereExpr004 ReturnExpr004 orderDecl-4/5/6/7/21/22 extexpr-6 position-1/3/4/5/6/7/8/9/10/13-18/20-22 last-1/3/4/5/8/9/10/13-18/20-22 ancestor-11/12/13 ancestorself-11/12/13 Michael Kay
Also affects ForExprType038/039
Michael et all: There may be other tests that are affected by this. I will try and find a way to automate findig those tests. For the time I corrected all the identified ones from the NIST side. Assigning the bug to ravindranath as this affects some of his tests as well. Thanks, Carmelo
There was actually a certain amount of automation in the way I produced the list: if my test driver finds a discrepancy, it tests to see if the discrepancy is in whitespace text nodes only, and if so, it tries running the test again with whitespace stripping applied to the source document. However, as I mentioned, it also involved some manual checking and I did this only for a couple of tests in each group.
Fixed following test output: fn-union-node-args-002/3 fn-intersect-node-args-011 fn-except-node-args-003/017 Did not find any white space problems with the following: fn-union-node-args-5/7/9/10/11/15/16 fn-intersect-node-args-015/16 fn-except-node-args-017
Closed by mistake. Reopening and assigning to Andrew.
All: All issues pertaining this particular bug seems fixed. I will go ahead and close it fixed for the sake of time. Thanks, Carmelo
Some test cases are still not fixed. I will take a look a this once more. Carmelo
We believe that all are done but the following: ForExpr005 WhereExpr004 ReturnExpr004 orderDecl-4/5/6/7/21/22
Michael: Thanks. Corrected all the missing whitespaces. Please close bug when able to verify. Thanks, Carmelo
All; For ForExpr005, we should NOT have get whitespace text node in front of <Deny>. The expect result should be: =============== <Deny> <security> <right>Read</right> </security> <user>system\Toshiko</user> </Deny> instead of =============== <Deny> <security> <right>Read</right> </security> <user>system\Toshiko</user> </Deny> Thanks Jinghao > Michael: > Thanks. Corrected all the missing whitespaces. Please close bug when able to > verify. > Thanks, > Carmelo
Mike et all: Looks like this was all taken care of. I am marking the bug as fix. Thanks, Carmelo