Brad,
On 2004 Dec 23 , at 02.54, Brad Cavanagh wrote:
>> I'll try checking out again to see if it's fixed now.
>
> Hate to be the bearer of bad news, but it's broken differently now,
> and the
> error message doesn't really make much sense to me:
>
> /bin/sh ./libtool --mode=link g77 -g -O2
> -I/local-star-make-world/include
> -fno-second-underscore -L/local-star-make-world/lib -o convert_idl.la
> -rpath
> /local-star-make-world/lib -module read_ndf.lo write_ndf.lo
> probe_ndf.lo
> crehds.lo hds2idl.lo idlstructdef.lo idlstructarrdef.lo
> idlstructfill.lo
> idlprimfill.lo hdsstructwrite.lo hdsprimwrite.lo gethdstype.lo
> getidltype.lo
> getidlstringsize.lo getobjectdetails.lo getstringarray.lo tagstrip.lo
> getcomp.lo checkarr.lo IDL_Load.lo hds.lo dat.lo dummy.lo `ndf_link`
> -lcc_dynamic
Sorry to be so tardy coming in to this thread.
I think the problem might be the -lcc_dynamic -- that's now what is
included to `solve' the restFP problem, rather than `-L/usr/lib -lgcc',
which turns out to be wrongish (I'd misunderstood the relevant
explanation first time). I changed that a week or so ago, and I
thought I'd confirmed, by the lack of nightly build failures, that this
was working. Perhaps it is working, and this isn't the problem, here.
I am pretty sure (or thought I was) that -lcc_dynamic is the correct
thing to include with a g77 link....
No! Hang on: in your first message on the 20th, you quote
> /bin/sh ./libtool --mode=link gcc -g -O2
> -I/local-star-make-world/include
> -L/local-star-make-world/lib -o convert_idl.la -rpath
> /local-star-make-world/lib -module read_
> ndf.lo write_ndf.lo probe_ndf.lo crehds.lo hds2idl.lo idlstructdef.lo
> idlstructarrdef.lo idlstructfill.lo idlprimfill.lo hdsstructwrite.lo
> hdsprimwrite.lo gethdstype.lo ge
> tidltype.lo getidlstringsize.lo getobjectdetails.lo getstringarray.lo
> tagstrip.lo getcomp.lo checkarr.lo IDL_Load.lo hds.lo dat.lo
> `ndf_link`
The difference is the inclusion of dummy.lo, and the idl/dummy.f which
Peter added, after my instruction/pattern somewhere. The presence of
that file is why the link is being done (correctly) with g77 rather
than gcc, which is in turn why -lcc_dynamic is being included. But
that's correct.
...no: it's the ast_link called by ndf_link that's the problem, because
of its inclusion of the AC_FC_LIBRARY_FLAGS. _That's_ what's resulting
in the Fortran libraries being included twice. ARGGGGGGGHHHHHHHH. I'm
going round in circles again.
I'm afraid that right now I do not have a neat solution to this. And
given that I'm just about to stop for the duration, I fear this will be
unresolved until the New Year. That's a pity, as I had rather hoped to
make a (another) final push for this autoconf stuff before finishing
off.
Nevermind....
Norman
--
----------------------------------------------------------------------
Norman Gray : Physics & Astronomy, Glasgow University, UK
http://www.astro.gla.ac.uk/users/norman/ : www.starlink.ac.uk
|