Norman,

I am having difficulty starting any of the STARJAVA applications
 [ussc@saturn bin]$ ./treeview
Exception in thread "main" java.lang.NoClassDefFoundError:
uk/ac/starlink/util/Loader
        at uk.ac.starlink.treeview.Driver.main(Driver.java:32)

is this related to util?

Note, I think there is some confusion (at least on my part) on some of the
properties in the build.xml files. In .stardev.properties I set,

star.dir=/apps/cvsbuild/star
star.install=/apps/cvsbuild/starjava
star.jar.dir=/apps/cvsbuild/starjava/lib
star.archive.dir=/apps/cvsbuild/export
stardev=/apps/cvsbuild/star
perl=/usr/bin/perl

What I am particularly confused about is star.jar.dir, in all of the
build.xml files it is set to ${star.dir}/lib, but isn't ${star.dir} supposed
to point to the Classic Starlink? If you chose to install in a different
location set by star.install then star.jar.dir will point to the wrong
location. This affects the HDX test, which fails in a different way if I
don't set star.jar.dir in .stardev.properties.

You just have to look at the previous test.log to see what I mean. I think
the problem is in Loader.java in the util package, as the build says that it
cannot find the native jnihds library. Changing star.jar.dir to point to
${star.install}/lib in the HDX build.xml file doesn't help.

Steve.

-----Original Message-----
From: Norman Gray [mailto:[log in to unmask]]
Sent: 28 November 2003 16:43
To: [log in to unmask]
Subject: Re: BUILD FAILED Executing HDX target: test

Steve,

On Friday, November 28, 2003, at 03:27 PM, Rankin, SE (Stephen) wrote:

> STARJAVA TEST failed again on HDX.
>
> See:
> http://cvsweb.starlink.ac.uk/cvsweb.cgi/buildlog/
> test.log?rev=1.66&content-t
> ype=text/x-cvsweb-markup

I think that quite a lot of XML stuff -- or at least some important
things -- must have changed between 1.4.1 and 1.4.2!

There are now two failures in the HDX code, and I think they're related
(one of them appears linked to a HDX.isValid() problem that Mark mailed
me about a little while ago, so they _may_ in fact be distinct, though
that seems a bit of a coincidence). It's clearly something subtle,
since it's happening in a sanity check that a previous sanity check
was supposed to make redundant. It's sufficiently obscure that I'm not
yet convinced it's a code failure rather than a wrong test, like
yesterday's util failure.

That's code for `it's not going to be fixed this evening'.

So: Do you want me to check in testcases with the appropriate tests
disabled, so that the build can run OK tonight; or just leave it until
Monday. The latter seems preferable to me.

Aha -- I see the problem! I've got 1200 lines of Hdx regression tests
-- something's bound to go wrong if you look closely enough....

See you,

Norman


--
----------------------------------------------------------------------
Norman Gray http://www.astro.gla.ac.uk/users/norman/
Physics and Astronomy, University of Glasgow [log in to unmask]