We are getting a little closer with gfortran and Amber 9:
nmode compiles and passes all tests
sander compiles and passes many tests, but (on Pentium IV):
the "findmask" routine doesn't get the right atoms in the targetted
MD
case
the code segfaults on tests that involve minimization
These errors *may* be indicating some problem with our code, but they may
also
be gfortran bugs (of which there have been many).
It's not a high priority to track these down, but if somebody does, that
would
be great. The "configure" script can now take gfortran as one of the
allowed
compilers -- this might be turned off for release if we can't fix this.
[I used a self-built gcc-4.0.2. If anyone has a really modern Linux
distro,
like FC4(?), you might see what that compiler gives.]
...dac
Received on Wed Apr 05 2006 - 23:49:50 PDT