Re: Re: Subject: Re: [xsl] Bug in treating an RTF by Saxon 6.5 and MSXML

Subject: Re: Re: Subject: Re: [xsl] Bug in treating an RTF by Saxon 6.5 and MSXML
From: Dimitre Novatchev <dnovatchev@xxxxxxxxx>
Date: Thu, 16 May 2002 15:45:16 -0700 (PDT)
Joerg Heinicke wrote:

> Hey Dimitre,
> are you a bit indignant? As far as I can see it's possible that the 
> template is applied either with a node-set or a RTF as param, isn't 
> it?

Hi Joerg,

No, I'm never indignant when a bug is found -- when it is in my product
I'm glad that somebody has helped me in testing. When I've discovered a
bug, I'm glad that I've helped other developers.

[code fragments skipped]

> So isn't the template applied with a node-set as Michael claimed?
> Even if a few lines later the template is applied with a RTF as


> Maybe I'm only too stupid for your stylesheets but they are so 
> complex and nearly not understandable with always at least 3 
> parameters and all possible types to pass ;-)


Did I ever say that the code demonstrating the bug was nice?

Code designed to demonstrate bugs is by definition nasty and buggy --
this is why it is a must an XSLT processor should not silently tolerate

In theory one may write XSLT code that is easy to port. In practice
not, especially if the code is tested on an XSLT processor, which
doesn't raise the red flag when necessary -- unfortunately other XSLT
processors will raise the red flag when the code happens to be run on
them. And this may happen many months after the code was written.

As for your desire to understand whether an RTF is really passed as
parameter, this is not difficult at all, especially if you read to the
end the text you included in your message.

Dimitre Novatchev.

Dimitre Novatchev schrieb:
> michael.h.kay@xxxxxxxxxxxxxxxxxxxx wrote: 
>>Dimitre Novatchev [dnovatchev@xxxxxxxxxxxxxxxxx], who should know
>>claimed that MSXML4 and Saxon exhibit a bug in their handling of 
> RTFs. 
>>The quoted lines are all valid, as far as I can see, because $arg1 is

>>not an RTF. (I say "as far as I can see", because I haven't really  
>>worked out what arguments are being supplied to this template). 
> Yes, Dimitre Novatchev ***knows better*** -- the value passed in
> is an RTF. 
> Here's the template with the offending lines again: 
>     <xsl:template match="str-split2words-func:*"> 
>       <xsl:param name="arg1" select="/.."/> 
>       <xsl:param name="arg2"/> 
> <!--Right--> <xsl:copy-of select="vendor:node-set($arg1)/*[1]"/> 
> <!--Err--> <xsl:copy-of select="$arg1/word[position() != last()]"/> 
>       <xsl:choose> 
> <!--Err--> <xsl:when test="contains($arg1/*[1], $arg2)"> 
> <!--Err-->  <xsl:if test="string($arg1/word[last()])"> 
> <!--Err-->   <xsl:copy-of select="$arg1/word[last()]"/> 
>            </xsl:if> 
>           <word/> 
>         </xsl:when> 
>         <xsl:otherwise> 
> <!--Err-->   <word><xsl:value-of  
>                 select="concat($arg1/word[last()], $arg2)"/></word> 
>         </xsl:otherwise> 
>       </xsl:choose> 
>     </xsl:template> 
> Look at the line starting with <!--Right-->  
> Why do you think I had to add the vendor:node-set() conversion there?

> The answer is simple -- because SAXON 6.5 (correctly!) issued an
> message for this line. 
> However, after I corrected this line, SAXON6.5 did not issue any
> for the following <!--Err--> lines. 
> Mike, you may try it immediately -- just change this line: 
> <!--Right--> <xsl:copy-of select="vendor:node-set($arg1)/*[1]"/> 
> to this: 
> <!--Err--> <xsl:copy-of select="$arg1/*[1]"/> 
> and you'll immediately get the correct error message from Saxon. 
> This proves that the error message must be given for the following  
> <!--Err--> lines -- and the bug is that no error message is issued on

> any of them. 
> And of course, feeding this example to a more conformant XSLT
> (in this case XalanJ-2.3) also shows that at all these lines there's
> illegal manipulation of an RTF as if it were a node-set. 
> So do I still need to know better? 
> If this bug continues to be there in future versions, I'll be forced
> use XALAN for development and testing, as it correctly catches more 
> errors. This will guarantee for me that something that works on XALAN

> will work on less conformant processors. 
> Cheers, 
> Dimitre Novatchev. 

Do You Yahoo!?
LAUNCH - Your Yahoo! Music Experience

 XSL-List info and archive:

Current Thread