Re: [AMBER] Any idea about this Amber compile bug?

From: Scott Brozell <sbrozell.rci.rutgers.edu>
Date: Thu, 1 Sep 2016 23:13:28 -0400

Hi,

1.
Check your environment for an errant FC variable.
If FC contains slashes then that might be causing the error from sed.

Configure now lets FC override fc, eg:
1145: if [ -z "$FC" ]; then fc=gfortran; else fc="$FC"; fi
1335: if [ -z "$FC" ]; then fc=ifort; else fc="$FC"; fi
1689: if [ -z "$FC" ]; then fc=pgf90; else fc="$FC"; fi
which was not envisioned when i wrote extract_and_emit_compiler_versions.
We now need some more bullet proofing in that routine.

2.
If 1 doesn't fix it then send verbose and verbatim details including
the result of requesting the version from your fortran compiler.
Send it to me or to am dev.

scott


On Thu, Sep 01, 2016 at 08:53:07PM -0400, Jason Swails wrote:
> It's in AmberTools/src. The configure script calls it directly so you should be able to follow the logic from there.
>
> What it sounds like is that the Intel compilers changed their version output format *again*, so you'll need to fix that parsing.
>
> HTH,
> Jason
>
> --
> Jason M. Swails
>
> > On Sep 1, 2016, at 4:45 PM, David Cerutti <dscerutti.gmail.com> wrote:
> >
> > Hi Dan (and Devs),
> >
> > I'm running into a problem compiling amber on MSU HPCC. I googled and
> > found that you had helped someone with nearly the exact same error two
> > years ago, so I tried your solution with soft-links:
> >
> > http://archive.ambermd.org/201405/0396.html
> >
> > However, this still did not work. I notice that, prior to the time at
> > which I get the error the other user reported, there was another failure
> > relating to Fortran compiler version identification. (The issue in 2014 did
> > not appear to be prefaced by this.)
> >
> > Checking for updates...
> > Checking for available patches online. This may take a few seconds...
> >
> > Available AmberTools 17 patches:
> >
> > No patches available
> >
> > Available Amber 17 patches:
> >
> > No patches available
> >
> > Using the AmberTools miniconda installation in
> > /home/cerutti/amber//miniconda
> > version 2.7.12
> >
> > Obtaining the intel compiler suite versions, e.g.:
> > /opt/software/ClusterStudio/2016.3/bin/icc -v
> > The C version is 16.0.3
> > sed: -e expression #1, char 8: unknown option to `s'
> > The Fortran version is
> > ./configure2: line 1384: [: : integer expression expected
> > ./configure2: line 1385: [: : integer expression expected
> > ./configure2: line 1462: [: : integer expression expected
> > ./configure2: line 1543: [: : integer expression expected
> >
> > Testing the /opt/software/ClusterStudio/2016.3/bin/icc compiler:
> > ...
> > <<<
> >
> > Notice the references to ./configure2; I'm looking around for where that
> > script might have been placed so that I can see what's happening around
> > those lines, but I don't see it. Can you point out where this might be?
> > There are a lot of issues with the modules system and platform-specific
> > settings on our HPCC machines, but this may lead us to some more general
> > ways to shore up the compilation.
> >
> > Thanks,
> >
> > Dave
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Sep 01 2016 - 20:30:02 PDT
Custom Search