On Thu, Nov 03, 2005, John Mongan wrote:
> The following gives instructions for fixing your build, which will
> probably break the next time you update:
Well, yes it does (break)....
On windows/cygwin, I get
gcc -c -m32 -O2 -I. -DNDEBUG -DpgiFortran ncx.c
ncx.c:1119:2: #error "ix_float implementation"
ncx.c:1542:2: #error "ix_double implementation"
when compiling netcdf.
Of course, I will debug this sometime, but I think we need to allow (at
least
for a while) builds without netcdf support, for machines where this is not
immediately portable. Maybe no-netcdf should even be the default, so that
one
needs a "-bintraj" flag to configure in order to turn it on.
Then the bintraj.f file could just have stubs for things that are in the
netcdf library, for use when -bintraj was not specified. The calls to
netcdf configuration could also be bypassed unless -bintraj was set.
This doesn't look very hard -- let me know if you want me to make a stab
at
it, or whether you can do it.
[Minor point: I don't think you can assume that module files are always
called
"foo.mod". So, the dependence on netcdf.mod in src/sander/Makefile
probably
needs some tweaking to make it really portable. This is another reason
that
we should anticipate that some platforms won't immediately be able to
support
netcdf, and so there should be an option to skip it...]
But, generally, it looks very promising!
...dave
Received on Wed Apr 05 2006 - 23:49:51 PDT
