[Mono-devel-list] Re: Fixes to XSLT

[Andrew Skiba]

Hello, Atsushi.

Atsushi Eno wrote:
> Yes, but my point is, there seems a bunch of differences between
> your results and mine. I haven't counted as yet, but apparently
> those "non conformant" cases should be less than 500 in my box.
> And actually on running your code without your patch, I didn't see
> any cases that Mono throws an exception. Can you please provide
> the precise example case (test file name) that Mono throws an
> exception while MS does not?

Sure. Just a random example. Write xsl like this:

<xsl:stylesheet version="1.0" 
xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
<xsl:attribute-set name="simple">
    <xsl:attribute name="attr"/>
</xsl:attribute-set>

<xsl:template match="/">
	<e xsl:use-attribute-sets=" simple "/>
</xsl:template>
</xsl:stylesheet>

It throws a NullReferenceException in Mono, but works good in MS.NET. 
This is a snippet from 
MSFT_Conformance_tests/Attributes/Attribute_WhitespaceInUseAttributeSets.xsl

> I would be also highly appreciated if you post the patch and example
> code to public mailing list (like mono-devel-list) and then other
> people can check which is the correct failure numbers (that's why
> I keep CC-ing to ximian-mono-list).

Yes, I saw that you CC to ximian mono, that's why I did not send it once 
more.

> Hmm... I don't think it is the way. In Class1.cs there is:
> 
>     XmlDocument catalog = new XmlDocument();
>     catalog.Load(@"..\catalog-out.xml");
> 
> So at least catalog-out.xml should be placed on the parent directory.
Ooops. Sure. Class1.exe is in MSFT_Conformance_Tests, and the 
catalog-out.xml is near catalog.xml. My fault.

> I doubt... because the cause of the error is apparently trivial
> directory name difference which is however hardcoded in Class1.cs.
> Also, that catalog-out.xml shows filenames in MSFT_Conformance_Tests,
> but catalog.xml is only in Xalan_Conformance_Tests.
No, we did not use Xalan conformance tests. We focused on MSFT 
conformance tests, may be this is the problem here. Yesterday I spent 
all day investigating the enum problem, which we discussed in the list, 
so I did not get my hands on your test suite. I got interesting results, 
and I'm going to send them to the list. Then I'll deal with your test 
suite, to find why it does not catch things I fixed.

Andrew Skiba.

> 
> Atsushi Eno