Hi Dave,
case wrote:
> Hi everyone:
>
> My sander compile on MacOSX (10.5) has suddenly stopped working, with this
> error:
>
> Undefined symbols:
> "_nf_put_var1_int_", referenced from:
> ___netcdf_MOD_nf90_put_var_fourbyteint in libnetcdf.a(netcdf.o)
> ___netcdf_MOD_nf90_put_var_eightbyteint in libnetcdf.a(netcdf.o)
>
> ....many more such problems.
>
> If I revert to CVS version 1.29 of the configure script, things work again.
> It looks like the main difference is that the "new" configure updates add
> "-mtune=generic -msse -mfpmath=sse" to the options send to netcdf configure,
> and to load commands.
>
> OR, it may be related to this comment (from revision 1.30):
>
> "Finally, netCDF should, if on a mac, be configured with -DpgiFortran. This
> usually seems to be correctly guessed, but with the combination of the gcc
> compiler and the -m32 flag, it will not be. Thus, I have ensured that if we
> are building on a Mac, this compiler flag will be added explicitly."
>
> But, as far as I can see, the current version of configure does not in fact
> add -DpgiFortran.
>
> Anyway...does anyone recoginize the problem? I guess I can debug it
> myself, but am hoping that it will be recognized. We've had *lots* of
> changes to the configure script over the past few days, which continues
> to scare me, since I am pretty sure these are not being tested on lots of
> platforms. Also, please don't tweak the configure scripts just to get the
> absolute last millisecond of performance out of the codes.
>
I recall very similar problems from Snow Leopard, which is why I added
the -DpgiFortran to the NetCDF configure. However, that seemed to break
the configure for Mengjuei, so you reverted it back. Since then, I
thought -DpgiFortran was no longer being used. In any case, the netCDF
Makefile should tell you one way or the other. It's not a problem with
the default (i.e., without any -m) configuration option on Snow Leopard.
In general, this problem seems to be related to the number of
underscores used. -DpgiFortran instructs the netCDF configure to use a
certain routine name format. There are other options, like -Df2cFortran.
For debugging, I don't have a Leopard machine to test on. It may involve
a bit of playing around with the compiler options.
I hope this helps a little.
Cheers,
Ben
--
Benjamin P. Roberts
Postdoctoral Research Associate
Quantum Theory Project
University of Florida
2301 New Physics Building #92
PO Box 118435
Gainesville FL 32611-8435
USA
Phone: +1 352 392 6712
Cell: +1 352 222 3677
Member of the Royal Australian Chemical Institute
and of the American Chemical Society
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Nov 16 2009 - 08:30:03 PST
