Norm, thank you for reporting this. It indeed turned out to be a recently
introduced regression related to some new pattern analysis code. I checked
in a fix tonight.
Your stylesheet ran without reporting an error, which may not be the same
thing as running correctly. I would greatly appreciate it if you could do
an update from the CVS and give it a try, and report if everything seems OK
to you. We're trying to stabalize the current trunk, and would like to
catch any further problems as soon as possible.
I run the files you sent through the JProbe profiler (obviously,
performance is not great). It basically just about croaked my machine, but
I did get some interesting information:
319561 Xerces AttrImpl, and
95608 Xerces ElementImpl, which look like they are all created from the
Xerces validater. Indeed, per Thomas' argument that dynamic validation
should be turned off, this is a good argument.
54815 Qualified name objects. Not sure how many of these are unique. I
suspect most of them, but I could be mistaken.
28056 TemplateSubPatternAssociations, which are normally associated with
matched templates. There are many more of these than matched templates,
which may or may not be legitimate.
13613 Hashtables. It's hard to tell where these are all coming from... 70%
are from Object.clone. I don't think Xalan uses Hashtables explicitly all
that much, but the SAX2 NamespaceSupport object does, which Xalan does use.
5051 XPath expressions, or close enough.
6228 ArrayIndexOutOfBoundsExceptions. Definately all Xerces (but part of
normal operation, as far as I know).
2815 location path expressions of the type that need to hold state (i.e.
iterators). The number is the number of iterator pools that are created.
1723 NullPointerExceptions. Again, all Xerces (but part of normal
operation, as far as I know).
1447 literal result elements in your stylesheets.
1354 xsl:apply-templates statements.
1325 xsl:templates. !!!! Give a poor processor a break!
372 variables.
259 params.
349737 Strings. (While surely some of these are from Xalan, I don't think
the majority are... but it's a little hard to tell)
288179 char arrays.
207840 int arrays. (Some of these are the DTM).
188408 Object arrays (Xerces, I think)
105447 Vectors.
-scott
Norman Walsh
<[EMAIL PROTECTED] To: [EMAIL PROTECTED]
om> cc: (bcc: Scott Boag/CAM/Lotus)
Subject: Re: ArrayIndexOutOfBounds in
OpMap
06/19/2001
11:55 AM
Please
respond to
xalan-dev
/ Norman Walsh <[EMAIL PROTECTED]> was heard to say:
| / [EMAIL PROTECTED] was heard to say:
| | Norm, thanks for the report. This is definately a regression, correct?
|
| I'm not entirely sure. I think so, but this particular case hasn't
| been a standard part of my build process until today.
Sadly, xalan 1.2.2 won't process this set of documents either.
In this template:
<xsl:template match="olink[@type='title']">
<xsl:message>Load: <xsl:value-of select="@targetdocent"/></xsl:message>
<xsl:variable name="xml"
select="document(unparsed-entity-uri(@targetdocent),.)"/>
...
It reports:
Load: lib.xml
Error! java.net.URL
Xalan: was not successful.
XSLProcessor: done
This is processing ../docsrc/reference.xml with reference.xsl. And note
that I've added the '.' to the document() function since it otherwise
uses the wrong base URI.
Be seeing you,
norm
--
[EMAIL PROTECTED] | To what excesses will men not go for the sake
XML Standards Engineer | of a religion in which they believe so little
Technology Dev. Group | and which they practice so imperfectly!--La
Sun Microsystems, Inc. | Bruy\`ere
