Greetings,
On 2004 May 20, , at 01.35, Alasdair Allan wrote:
> After compleing the build I did an an 'ant test', which bails during
> the
> HDX tests (see attached crash.log file).
>
> The JRE version is 1.4.2_03 and the JVM is
> 1.4.2-34.
This has come back to bite us. Last November/December we tried 1.4.2,
and this same HTX test failed. See
<http://www.jiscmail.ac.uk/cgi-bin/webadmin?
A2=ind03&L=stardev&D=0&T=0&P=82480>. I've changed the subject of this
message to match that thread.
The summary in that message was that this works with JDK 1.4.1, but:
* JDK 1.4.2 comes with version.xalan2_2=Xalan Java 2.4.1 and fails
with the two-endDocument error.
* JDK 1.4.2 with xalan 2.5.2 passes that test, but fails a DOMResult
test for which there is an open bug report.
That is, I believe that the HDX code is not wrong, but it's triggering
a bug in the underlying XML support. I haven't _proved_ this, but
there seems good circumstantial evidence for this.
The current version of Xalan[1] is 2.6.0, and bug 15140 ([2], the open
bug in question) is reported as fixed in that version.
We have no leeway to mandate a version of Java, since OSX users at
least will be reluctant to start mucking around with versions, so we're
stuck with 1.4.2.
So:
1. I could check that everything works with Xalan 2.6.0, and if so
start distributing that along with our stuff. Unappealing.
2. I could devise workarounds which are active for those particular
Xalan versions which are known to fail the tests. Potentially very
messy, and hard, since I haven't been able to localise the bug very
well -- the failing tests are pretty fundamental sanity checks.
3. I could simply skip the test in those cases (based on Xalan
versions) where it's known to fail. The tests in question are rather
proactive -- they test more extreme situations than are terribly likely
to come up in practice, as opposed to being responses to bugs that
appeared in use -- so it's not obvious that there would be any
functionality problems which would emerge. I think the corners of HDX
are not at present exercised very heavily in applications.
I propose (3), and that we switch to using JDK 1.4.2, but if any usage
problems come up which are related to this, then I'll produce
workarounds. How does that sound?
I have no plans to do this between now and Boston, however.
See you,
Norman
[1] http://xml.apache.org/xalan-j/
[2] http://nagoya.apache.org/bugzilla/show_bug.cgi?id=15140
--
------------------------------------------------------------------------
---
Norman Gray
http://www.astro.gla.ac.uk/users/norman/
Physics and Astronomy, University of Glasgow, UK
[log in to unmask]
|