On Mon, 19 Sep 2005, Norman Gray wrote:
> On 2005 Sep 17 , at 01.05, Tim Jenness wrote:
>
> > If rpms, starjava and ssn78 are ignored
>
> And the latter only requires a
>
> test -f STAR_SUPPRESS_BUILD || make dist
>
> in the general build script, for it to be properly suppressed.
>
> By the way, this suppress-build mechanism is quite general, though
> it's only been used by ssn78 so far. Is this something the rpms and
> starjava builds could use?
The persistent starjava15 failure on X86_64 is down to the fact that
the JNIAST native libraries have not been recompiled for the latest
JNIAST classes (JNIAST v4.0-1/AST v4.something) for this architecture.
You can tell because the log fails with a JVM crash (not just a java
stacktrace) during the JNIAST run-tests target. They haven't been
recompiled for SPARC either, though that doesn't seem to be leading
to a build/test failure at present. I'm not really sure where the
relevant hardware is for this etc so I wasn't planning to do it myself.
I mentioned that it was work outstanding to Peter a month or so back,
not sure if he decided it wasn't worth doing or it just slipped off the
bottom of his list. However, it probably makes sense for someone
either to build and check in these libraries, or to use Norman's
suppress-build mechanism for starjava on the offending platforms.
Mark
--
Mark Taylor Astronomical Programmer Physics, Bristol University, UK
[log in to unmask] +44-117-928-8776 http://www.star.bris.ac.uk/~mbt/
|