[qmtest] working with the xml result stream

[Mark Mitchell]

On Tue, 2003-09-30 at 09:18, Stefan Seefeld wrote:
> I'v run into a little issue with the generated xml output using
> xml-result-stream.
> 
> I run qmtest on a variety of platforms, and I generate one xml
> result file for each. Then I process these files with xslt to
> generate a 'result matrix'. At some point I'm determining the
> number of failures/passes/etc., and the xslt way to do this is
> with an xpath expression such as
> 
> "count($target/results/result[outcome='FAIL'])"
> 
> The problem is that the content of 'outcome' nodes varies
> between platforms, i.e. on linux it may be
> 
> <outcome>PASS</outcome>
> 
> and on windows
> 
>    <outcome>
>     PASS
>    </outcome>
> 
> This means that I can't just compare the string values that
> easily, since the content differs in whitespace.
> 
> Thus my question: wouldn't it be more natural to put the outcome
> value into an attribute ?
> A temporary workaround for my xpath expression may be to use
> "contains(outcome, 'FAIL')", but that seems quite hackish.
> 
> Checking with the document type definition for result files
> I think it would be perfectly fine to make such a change,
> i.e. the outcome is actually an enum, so the dtd / schema could
> well be tightened a bit to make processing easier.
> 
> Any thoughts ?

Yes, that would definitely be much better!

Are you willing to work on a patch?  If so, you should make
XMLResultReader._GetResultFromDomNode accept both the old and new format
-- we try to maintain backward compatibility wherever possible.

Thanks!

-- 
Mark Mitchell
CodeSourcery, LLC
mark at codesourcery.com