Martin Costabel <costabel <at> wanadoo.fr> writes:

> Mike Zulauf wrote:
> > 
> > Poking around the various related libraries, it looks to me like kio  is 
> > defined in liblam, which isn't linked.  Is this the problem?  Is  it 
> > time to inform Jeffrey Whitaker, or is there more to do first?
> 
> I don't have libmpi (from the lampi package, I suppose?). Could you try 
> to uninstall this temporarily and see if octave then builds correctly?

That seems to have done the trick.  It got through the "config" stage, and is
now building (I'm doing this 
on the slower of my two machines).

I've now got both config logs placed online at:
<http://www.met.utah.edu/skrueger/homepages/mazulauf/stuff/octave.config.1.log>
<http://www.met.utah.edu/skrueger/homepages/mazulauf/stuff/octave.config.2.log>

It looks like the configure script looks for MPI capability, and attempts to
link/test against it if it is 
available - but it doesn't do it properly (the first log).  In the second log
(when MPI has been removed), 
I'm guessing that it determines that MPI is unavailable, and proceeds.  (The
test fails, but in a way it 
understands?)

I'd guess that it wouldn't fail in the first case if it had also attempted to
link with liblam.  Maybe it 
expects libraries as for MPICH or something. . .


> The error seems to come from the "new behavior" (AKA "not a bug") of 
> Tiger's dyld that is annoying us regularly these days. It crashes binary 
> executables that were compiled and linked without problem, because it 
> tries (without being asked) to resolve undefined symbols in dependent 
> libraries but is too stupid to find them.

This could be the case - this is all above my head a bit.


> OTOH, if octave does not have a dependency on lampi, Fink policy says it 
> should not try to link to its libraries. So there is something to do for 
> the octave maintainer, it seems.

Octave doesn't depend on lammpi, and as far as I can tell neither do any of it's
dependencies (I got no 
warnings when I removed lammpi packages).  But it does look as if the configure
script does test for 
MPI capability, and fails (in this case) when it exists.  Whatever the cause,
seems like something the 
maintainer should know about, correct?

If so, I'll email him offline. . .

Thanks,
Mike




-------------------------------------------------------
This SF.Net email is sponsored by Oracle Space Sweepstakes
Want to be the first software developer in space?
Enter now for the Oracle Space Sweepstakes!
http://ads.osdn.com/?ad_id=7412&alloc_id=16344&op=click
_______________________________________________
Fink-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/fink-users

Reply via email to