Re: [Gestalt-users] Gexslt crash with xsl:import: URI format?

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 422-6466

>>>>> "Florent" == Florent Georges <dar...@ya...> writes:

    >> so that it looks like:

    >> -- attributes_prepared: attributes_prepared

    Florent>   Done.

    >> Then compile afresh (geant compile_ise_debug - that should take
    >> only a few seconds, as it will be an incremental compile), and
    >> run the test again, posting the new stack trace here.

    Florent>   Running the new build on the reduced sample I sent
    Florent> today (with the three little files), I get the following
    Florent> error:

    Florent> gexslt: system execution failed.  Following is the set of
    Florent> recorded exceptions:

    Florent> -------------------------------------------------------------------------------
    Florent> Class / Object Routine Nature of exception Effect
    Florent> -------------------------------------------------------------------------------
    Florent> XM_XSLT_FUNCTION prepare_attributes @6 not_excluded:
    Florent> <000000000349B80C> Precondition violated.  Fail

    Florent>   So in the same file I commented out the following line
    Florent> in function_fingerprint:

    Florent>     if not attributes_prepared then prepare_attributes
    Florent> end

Which is the wrong try.

In general in Eiffel, if a routine has a precondition, you must first
check that you can meet the pre-condition before you may call the
routine.

So the obvious try would be:

if not attributes_prepared and not is_excluded then
  prepare_attributes
end

It looks like that would have been sufficient to fix your problem.

But I didn't do this.
The is_excluded was my way of implementing [xsl:]use-when conditional
compilation - if the use-when clause returns false, then I marked the
element and is_excluded. So whole parts of the tree are logically
excluded.
Hence the pre-conditions.
But it's slightly error prone. Rather than wait for another such error
to be uncovered, or to write a proof checker, I now (as of today)
physically remove the excluded elements prior to begining the compile
process. (MHK does it with a filter, but that is conceptually harder -
I prefer the straight-forward approach)

This seems to fix your problem.

    Florent>   Then the transformation (on the simplified sample) run
    Florent> at the end, and produced no output 

I get that too.

    Florent> nodes, that is correct).  Then I came back to the
    Florent> original files, and there I got the following stacktrace
    Florent> (and biginning of output):

    Florent>   I let you deduce the conclusion ;-).  I put the two
    Florent> files (input + stylesheet) there:

    Florent>     http://www.fgeorges.org/tmp/hello-world.xsl
    Florent> http://www.fgeorges.org/tmp/test-hello-world.xts.xsl

I trust you meant stylesheet + input?

If not, then I gets dozens of compile errors. If you did, then I get
no errors, and a load of output looking like this:

<?xml version="1.0" encoding="UTF-8"?>

      <elem1 xmlns:xs="http://www.w3.org/2001/XMLSchema" xmlns:hw="http://www.fgeorges.org/hello-world"/><elem2 xmlns:xs="http://www.w3.org/2001/XMLSchema" xmlns:hw="http://www.fgeorges.org/hello-world" id="id"/>

      type="text/xsl" href="http://www.fgeorges.org/xslt-unit/report-to-html.xsl"

            <elem1 xmlns:xs="http://www.w3.org/2001/XMLSchema" xmlns:hw="http://www.fgeorges.org/hello-world"/><elem2 xmlns:xs="http://www.w3.org/2001/XMLSchema" xmlns:hw="http://www.fgeorges.org/hello-world" id="id"/>

         Tests: 
         hello-world(), arity 0

         hello-world(), arity 0

         Tests: 
         hello-world(), arity 1

         hello-world(), arity 1

         Tests: 
         hello-world()

         hello-world()

            Tests: 
            'elem' template rule

            'elem' template rule

            Tests: 
            False negative

            False negative

           test: 

                [success]

                [FAILED !]

           test: 

                [success]

                [FAILED !]

           test: 
         hello-world(), arity 0

                [success]

                [FAILED !]

         hello-world(), arity 0

           test: 
         hello-world(), arity 1

                [success]

                [FAILED !]

         hello-world(), arity 1

           test: 
         'elem' #1

                [success]

                [FAILED !]

         'elem' #1

           test: 
         'elem' #2

                [success]

                [FAILED !]

         'elem' #2

           test: 

                [success]

                [FAILED !]

Since this is not well-formd XML, I was dubious, but as Saxon 8.7.1
produces identical output, I assume the problem is fixed.

I should have this checked in to CVS before Tuesday.

remember, the procedure to update is:

cd %GOBO%
cvs update (however you do this in windows)
geant install

Then re-compile gexslt with:

cd %GOBO\src\gexslt
geant compile

then copy gexslt.exe to %GOBO\bin

If by any chance I haven't checked in by then, you can try the fix I
showed you above (and repeat if any further instances arise).
-- 
Colin Adams
Preston Lancashire