[AMBER-Developers] NetCDF Update

From: Daniel Roe <daniel.r.roe.gmail.com>
Date: Thu, 31 Oct 2013 11:14:23 -0600

Hi All,

The NetCDF in GIT master has been updated from version 3 (several
years out of date) to version 4. The build is completely backwards
compatible, and no additional features/requirements (such as using the
HDF5 libraries) have been added.

After version 4.1 NetCDF has been split into separate interfaces for C
and Fortran. Because of this, the NetCDF build process is now slightly
different. Basically the Fortran interface requires that the C
interface be compiled beforehand, so now both interfaces are
configured and compiled one after the other during Amber's main
configure. This increases initial configure time, but subsequent
builds are faster since the NetCDF configure/build step can be
skipped. The configure script is smart enough to recognize when the
interface needs to be rebuilt (such as when switching compiler types
etc). The '--with-netcdf' keyword has been re-enabled as well so you
can specify a separate NetCDF build.

NetCDF now uses an automake build which has some wrinkles when
combined with versioning control software such as GIT. Basically when
you switch branches, file timestamps are reset. This leads to the
automake framework thinking certain files need to be rebuilt when they
really dont. I've added a script that updates certain critical file
timestamps if necessary to prevent this, so you may see this during

Warning: NetCDF build timestamps need to be updated. This should
         ONLY happen during development after switching branches
         in GIT. Fixing timestamps (this will take about 7 seconds).

This will ONLY be necessary after you switch branches. It works fine
for me on my system, but I have occasionally seen on other systems
minor changes to files like config.h.in in the netcdf fortran build.
The changes are innocuous but please avoid committing them (just reset
the file with 'git checkout <file>') and let me know when it happens.
Users won't encounter this issue unless they somehow screw with their
file timestamps.

These changes have been in the master-update-netcdf branch for a while
now, which has been tested by myself, Jason, and Dave among others,
and no issues have been seen. If you do encounter problems let me know
right away. Thanks!

If there are any other questions let me know.


Daniel R. Roe, PhD
Department of Medicinal Chemistry
University of Utah
30 South 2000 East, Room 201
Salt Lake City, UT 84112-5820
(801) 587-9652
(801) 585-6208 (Fax)
AMBER-Developers mailing list
Received on Thu Oct 31 2013 - 10:30:03 PDT
Custom Search