loop building with ARP/wARP - REFMAC/gfortran problem? and work-around

From: Huw Jenkins
Date: 6 September 2011 15:48

On 25 Aug 2011, at 14:12, Gregory Bowman wrote:

> When I try to run ARP/wARP classic for loop building, I get the following message in the logfile:
>
> QUITTING ... ARP/wARP module stopped with an error message:
> REFMAC

I get the same error running auto_tracing.sh from ARP/wARP 7.2.  (CCP4 6.2.0, OS X 10.6.8)

As a work-around getting ARP/wARP to use the refmac binary from http://www.ysbl.york.ac.uk/~garib/refmac/data/refmac_experimental/refmac5.6_macintel.tar.gz (version 5.6.0119) allows ARP/wARP to run.

Some information that may help work out what the problem is below:

The only helpful error message I get in the logs is:

At line 723 of file /sw64/src/fink.build/ccp4-6.2.0-102/ccp4-6.2.0/src/refmac5_/read_extra_restraints.f (unit = 9, file = '/tmp/huwtj/refmac5_temp1.53659_BOND_R')
Fortran runtime error: Sequential READ or WRITE not allowed after EOF marker, possibly use REWIND or BACKSPACE

 ### CCP4 6.2: Refmac_5.6.0117     version 5.6.0117 : 13/06/11##

QUITTING ... ARP/wARP module stopped with an error message:
REFMAC5

At first I thought this was a 32/64bit issue but after compiling CCP4 as 32bit I get exactly the same error:

At line 723 of file /sw/src/fink.build/ccp4-6.2.0-102/ccp4-6.2.0/src/refmac5_/read_extra_restraints.f (unit = 9, file = '/tmp/huwtj/refmac5_temp1.81086_BOND_R')
Fortran runtime error: Sequential READ or WRITE not allowed after EOF marker, possibly use REWIND or BACKSPACE

 ### CCP4 6.2: Refmac_5.6.0117     version 5.6.0117 : 13/06/11##

QUITTING ... ARP/wARP module stopped with an error message:
REFMAC5

In both cases this is CCP4 built/installed by fink on OS X 10.6.8.

When I try to compile refmac from the latest source code with gfortran 4.6.1 I get a lot of warnings (which go away if I add -fno-whole-file to XFFLAGS) and the executable produced fails in the same way when called by auto_tracing.sh. This suggests to me it's a compiler issue not any differences between refmac 5.6.0117 and 5.6.0119, unfortunately that's about the limit of my understanding.

Hopefully some of that is useful to someone!


Huw
----------
From: Miguel Ortiz Lombardía

Hi Huw,

I recently found the same problem and solved it by compiling refmac
5.6.0119 using gnu compilers from the 4.4 series (4.4.6 I think). I'm
attaching the makefiles I used. However, this will only work if you
compile the whole ccp4 suite with the same compilers...

To do it I ran:

source <...wherever_CCP4_is...>/include/ccp4.setup-sh
FC=gfortran F77=gfortran CC=gcc-4 CXX=g++-4 ./configure
--with-netlib-lapack 2>&1 | tee configure.log
make 2>&1 | tee make.log
make install 2>&1 | tee install.log

Using Apple lapack libraries didn't work.

HTH.


--
Miguel

Architecture et Fonction des Macromolécules Biologiques (UMR6098)
CNRS, Universités d'Aix-Marseille I & II
Case 932, 163 Avenue de Luminy, 13288 Marseille cedex 9, France
http://www.afmb.univ-mrs.fr/Miguel-Ortiz-Lombardia

----------
From: Saul Hazledine
Hello,
  Further investigation (and lots of help from Huw Jenkins) has confirmed that the following error:
occurs to most ARP/wARP classic model building jobs when used with the Fink CCP4 version 6.2.0-102. Also, as suggested yesterday on this board, the error occurs when CCP4 has been compiled with gcc/gfortran 4.6.1.

There are two main workarounds at the moment. The easiest is to turn off conditional restraints as shown in the image linked below:

http://min.us/moKdRlW4l

The other is to compile Refmac with gfortran version 4.4.

This error only appears in the lastest Fink compiled version of CCP4 - the DMG installed binaries of CCP4 do not exhibit this problem.

We are trying to find the underlying cause of the problem and I hope that this error isn't causing too much inconvenience.

Saul Hazledine