On Wed, 29 Nov 2006, Brad Cavanagh wrote:
> On Wed, 29 Nov 2006, Peter W. Draper wrote:
>
>> I've also changed the build procedure so that it will use $FC, if defined
>> in preference to the built-in definition, and fixed up that problem with
>> the definition of JAVA_INCLUDE, so all those hacks you have in place
>> should be removed.
>
> I did a fresh checkout, and it picked up g95 just fine. Ta.
>
> Unfortunately g95 doesn't know about -lcc_dynamic, so I removed it, and
> essentially made the g95 line look more like what's described in
> http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0603&L=STARDEV&P=R1167&I=-3
>
> I didn't add the -install_dir flag.
OK, seems we'll need to hand-craft this until the old problem (restFP)
goes away (or I stop building ppc on 10.3).
>> Looking forward to that exception.
>
> Hopefully this is useful for you. I can't seem to get anything useful out of
> it. :-)
>
> [junit] java.lang.UnsatisfiedLinkError: couldn't load library splat:
> /Users/bradc/development/starlink/java/source/splat/lib/i386/libsplat.jnilib:
> [junit] at uk.ac.starlink.util.Loader.loadLibrary(Loader.java:146)
> [junit] at
> ......
> /star/starjava/lib/i386/libsplat.jnilib:
> [junit] at java.lang.ClassLoader$NativeLibrary.load(Native Method)
> [junit] at
Isn't very useful after all. Just amounts to what we knew already, it
exists and doesn't load (fully, clearly it does load a little, we can see
that from when HDS was causing a core dump, that traceback is from
libsplat, not libjnihds as I expected). At least the architecture is right
and the places it is looking.
Could you try removing the "-undefined suppress" and see what symbols are
missing? Maybe there's an extra library we should be linking against. Also
what g95 are you using? Maybe that's doing something dodgy, or is dodgy.
Try FC="g95 -v" so we can see what the linker is really being passed.
Failing all that, post me the notebook!
Cheers,
Peter.
|