[Fwd: Re: [AMBER-Developers] 28 Mar release candidates for Amber11/AmberTools-1.4]

From: Mark Williamson <mjw.sdsc.edu>
Date: Tue, 30 Mar 2010 12:56:02 -0700

Yes, I came to the same conclusion as Dan on this one, but accidentally
emailed my findings to myself, and not the list last night.

---------------------------- Original Message ----------------------------
Subject: Re: [AMBER-Developers] 28 Mar release candidates for
From: "Mark Williamson" <mjw.sdsc.edu>
Date: Tue, March 30, 2010 2:15 am
To: "Mark Williamson" <mjw.sdsc.edu>

Daniel Roe wrote:

>> All of the errors have to do with evb test cases, specifically:
>> cd evb/malon_dbonds_umb_dg_UFF_3DG_qi_full_2D-PMF && ./Run.evb

I'm seeing this seqfault as well. Fearing that it was related to my
ene/ener changes, I rolled back the CVS tree to before my changes ( cvs
co -r before_mjw_ene_fix amber11 ) and started to test the first fail.
The problem existed before my changes.

 From my testing, the segfault is happening here in pimd_force.f, during
the evaluation of dlnCdd_dl :

         do m = 1, natomCL
             write(6,*) "m is",m
             mdx = ( m - 1 ) * 3
             write(6,*) "mdx is",mdx
             mm = bead_dcrypt( m, div_ndx(n) )
             write(6,*) "m is",mm
             dlnCdd_dl = dlnCdd_dl + kT * dmdlm(mm) * 0.50d0 * (
gradRC(mdx+1,n)**2 + gradRC(mdx+2,n)**2 &
                                   + gradRC(mdx+3,n)**2 ) / mass(mm) /
gnorm !<----------------------------- segfault here....

I think it is occurring because the allocatable array, dmdlm(), is being
referenced, but has not been allocated. This is because itimass is not >
0, (see rdparm.f ~ 558). Generally, all accesses to dmdlm() in the code
are preambled with a (itimass > 0) safeguard, however, here there is not.

I think this code that is triggering this went in at this point:

commit 2f971243d1240774a71a4de69e103b1c6865988e
Author: kimwong <kimwong>
Date: Mon Feb 15 20:29:45 2010 +0000

     KFW: Put in Amber dihedrals and angles for DG-EVB.



AMBER-Developers mailing list
Received on Tue Mar 30 2010 - 13:00:04 PDT
Custom Search