05 June 2015

611. Building ecce on debian jessie

* I've confirmed that ECCE built like this installs and works perfectly on a Thinkpad SL410 with intel graphics

* It also compiles and runs perfectly on a home built desktop with external nvidia card (GF119) using the binary non-free nvidia drivers

* It also compiles and runs perfectly on a home built desktop with onboard nvidia (GeForce 7025/nForce 630a) using the nouveau drivers. I had issues on this desktop before, but reinstalled debian jessie from scratch. Before that I used nvidia-legacy drivers, which may or may not  (probably not) have had something to do with it not working.

* UPDATE: instead of putting
#include <freetype.h>
the recommended method is to use
#include FT_FREETYPE_H
I've updated the patch_script.sh below accordingly. Same goes for ftoutln.h vs FT_OUTLINE_H, but the latter didn't work (error saying #include must use "" or <>)

* UPDATE: If you're having issues with undetected -lGL and -lGLU in the wxwidgets step, it's because
667       endif
668       cat configure.orig | sed -e 's%^SEARCH_INCLUDE="\\%SEARCH_INCLUDE="$ECCE_HOME/${ECCE_SYSDIR}3rdparty/mesa/include \\%' >! configure
669       chmod a+x configure

needs to be changed to
667       endif
668       cat configure.orig | sed -e 's%^SEARCH_INCLUDE="\\%SEARCH_INCLUDE="$ECCE_HOME/${ECCE_SYSDIR}3rdparty/mesa/include $ECCE_HOME/${ECCE_SYSDIR}3rdparty/mesa/lib \\%' >! configure
669       chmod a+x configure
I had this issue on a debian wheezy system with the vendor nvidia libraries.
I wouldn't have spotted this bug otherwise.

* Another error from my debian wheezy nvidia system: if you get
checking how to run the C preprocessor... x86_64-linux-gnu-gcc -E ./configure: line 2880: syntax error near unexpected token `Using' ./configure: line 2880: `  AC_MSG_NOTICE(Using external PCRE library from $PCRE_CONFIG)'
then make sure you're not using autoconf2.13, which is an obsolete version. I think I have it due to my system originally being installed back in 2010.

* Finally, I'm currently  working on fixing minor things that have been nagging me in ecce. One is the basis set quicklist (in src/dsm/edsiimpl/ICalcUtils.C), but obviously that's a personal preference. A more serious one is the 256 character limit for csh commands:
Exceeds maximum C shell command length of 256 characters
Note that this isn't the C shell complaining -- this is a built-in limit in ecce (in src/comm/rcommand/RCommand.C). I've changed that limit to 16384 characters (the real limit is much, much higher)
I've also added two basis sets to ECCE, and have tinkered with the exchange-correlation functionals.

I'll try to push the fixes back upstreams when I'm ready if they'll accept them; otherwise I'll create my own github/sourceforge repo.

Building ECCE on debian wheezy was a breeze. Building ECCE on debian jessie was painful.

In the end it boiled down to two things:
*freetype headers are no longer in freetype2/freetype/ but in freetype2/

*-Wformat-security is turned on by default

mkdir ~/tmp/ecce_compile -p
cd ~/tmp/ecce_compile
sudo apt-get install bzip2 build-essential autoconf libtool ant pkg-config
sudo apt-get install gtk+-2.0-dev libxt-dev csh gfortran openjdk-7-jdk python-dev
sudo apt-get install libjpeg-dev imagemagick xterm libfreetype6-dev libfl-dev libtool-bin

As usual I'm not 100% sure when it comes to the necessary packages. libfl-dev might not be needed.

Download the ECCE source code and put the ecce-v7.0-src.tar.bz2 file in ~/tmp/ecce_compile. Put the patch_script.sh file (see below in this post for the code) in ~/tmp/ecce_compile. Then do
tar xvf ecce-v7.0-src.tar.bz2 
cd ecce-v7.0/
export ECCE_HOME=`pwd`
cd build/

You'll now step through a list over programs and libraries that are needed and what ECCE can find. If you're having issues with e.g. javac and java being different versions, use sudo update-alternative --config javac.

At the end you'll be asked whether to skip these checks next time -- answer y(es).

Next do
./build_ecce|tee xerxes.log && ./build_ecce |tee mesa.log && ./build_ecce |tee wxwidgets.log
sh ../../patch_script.sh && ./build_ecce|tee wxpython.log 
./build_ecce|tee httpd.log && ./build_ecce|tee ecce.log

If all went well you've managed to build the ecce binaries. If not, go through wxpython.log and check for errors relating to format-security statements. Then go through ecce.log and look for issues with freetype.

What to do with the binaries? Follow one of the many ECCE installation posts on this blog, e.g. http://verahill.blogspot.com.au/2013/08/487-version-70-of-ecce-out-now.html

NOTE that if you get ''Invalid null command." trying to execute install_ecce.v7.0.csh, install tcsh and do
tcsh install_ecce.v7.0.csh

The patch_script.sh file -- when copying, make sure to check that the lines end where they are supposed to and not broken up.
#!/bin/bash cp ${ECCE_HOME}/build/3rdparty-dists/wxPython-src- ${ECCE_HOME}/3rdparty/build/ cd ${ECCE_HOME}/3rdparty/build/ tar xvf wxPython-src- rm wxPython-src- cd ${ECCE_HOME}/3rdparty/build/wxPython-src- grep -rsl "PyErr_Format(PyExc_RuntimeError, mesg)" *|xargs -I {} sed -i 's/PyErr_Format(PyExc_RuntimeError, mesg)/PyErr_Format(PyExc_RuntimeError, "%s", mesg)/g' {} cd ${ECCE_HOME}/3rdparty/build/wxPython-src- sed -i 's/wxLogFatalError(m)/wxLogFatalError("%s", m.c_str())/g' _misc_wrap.cpp sed -i 's/wxLogError(m)/wxLogError("%s", m.c_str())/g' _misc_wrap.cpp sed -i 's/wxLogWarning(m)/wxLogWarning("%s", m.c_str())/g' _misc_wrap.cpp sed -i 's/wxLogMessage(m)/wxLogMessage("%s", m.c_str())/g' _misc_wrap.cpp sed -i 's/wxLogInfo(m)/wxLogInfo("%s", m.c_str())/g' _misc_wrap.cpp sed -i 's/wxLogDebug(m)/wxLogDebug("%s", m.c_str())/g' _misc_wrap.cpp sed -i 's/wxLogVerbose(m)/wxLogVerbose("%s", m.c_str())/g' _misc_wrap.cpp sed -i 's/wxLogStatus(pFrame, m)/wxLogStatus(pFrame, "%s", m.c_str())/g' _misc_wrap.cpp sed -i 's/wxLogStatus(m)/wxLogStatus("%s", m.c_str())/g' _misc_wrap.cpp sed -i 's/wxLogSysError(m)/wxLogSysError("%s", m.c_str())/g' _misc_wrap.cpp sed -i 's/wxLogGeneric(level, m)/wxLogGeneric(level, "%s", m.c_str())/g' _misc_wrap.cpp sed -i 's/wxLogTrace(mask, m)/wxLogTrace(mask, "%s", m.c_str())/g' _misc_wrap.cpp cd ${ECCE_HOME}/src grep -srl "<freetype/freetype.h>" |xargs -I {} sed -i 's,<freetype/freetype.h>,FT_FREETYPE_H,g' {} grep -srl "freetype/" |xargs -I {} sed -i 's,freetype/,,g' {} cd ${ECCE_HOME}/build

What I found during troubleshooting:

In files such as:
there are sections which look like this:
863 } else { 864 PyErr_Format(PyExc_RuntimeError, mesg); 865 }
Compiling with -Wformat-security means that you'll have to patch all those expression to
863 } else { 864 PyErr_Format(PyExc_RuntimeError, "%s", mesg); 865 }
There were similar issue with wxLog*(m) statements in other files, e.g.
3rdparty/build/wxPython-src- -> ("%s", m.c_str()) 3093 m.Replace(wxT("%"), wxT("%%")); 3094 wxLogFatalError(m); 3095 } .. 3177 m.Replace(wxT("%"), wxT("%%")); 3178 wxLogTrace(mask, m); 3179 }

04 June 2015

610. Opening G09 output files in ECCE. A rough method

Update: Note that one thing that's not recognised at the moment are the MOs. Somehow I don't think that should be too difficult. Likewise, I think one should be able to import nwchem files by a little bit of editing like below.

Original post:
When opening a gaussian output file with the ECCE viewer:
* a symlink to the file is put in a subdirectory of /tmp
* the file is parsed for indications as to what the file type is:
+go+cd /tmp/ecce_me/jobs/Gaussian03__HEa24c +go+ln -s /home/me/calcs/test/Outputs/g03.g03out g03.g03out; echo CMDSTAT=$status CMDSTAT=0 +go+grep "Gaussian 98, Revision" g03.g03out; echo CMDSTAT=$status CMDSTAT=1 +go+grep "Gaussian 94, Revision" g03.g03out; echo CMDSTAT=$status CMDSTAT=1 +go+grep "Gaussian 03, Revision" g03.g03out; echo CMDSTAT=$status CMDSTAT=1 +go+grep "%begin%input" g03.g03out; echo CMDSTAT=$status CMDSTAT=1 +go+grep "%begin%input" g03.g03out; echo CMDSTAT=$status CMDSTAT=1 +go+grep "Northwest Computational Chemistry Package" /home/me/test/Outputs/g03.g03out; echo CMDSTAT=$status
That's easy enough to fool by simply putting a Gaussian 03 line in the output (assuming that the G09 and G03 output are similar enough).

Here's a successful example, in the sense that ECCE found that it was a Gaussian 03 file:
CMDSTAT=0 +go+grep "Gaussian 98, Revision" g03.g03out; echo CMDSTAT=$status CMDSTAT=1 +go+grep "Gaussian 94, Revision" g03.g03out; echo CMDSTAT=$status CMDSTAT=1 +go+grep "Gaussian 03, Revision" g03.g03out; echo CMDSTAT=$status Gaussian 03, Revision CMDSTAT=0 +go+if (-w /tmp/ecce_me/jobs/Gaussian03__7491Sb) echo TRUE TRUE +go+echo $PATH; echo CMDSTAT=$status Word too long. +go+if (-x Gaussian-03.expt) echo TRUE +go+exit; echo GOODBYE exit; echo GOODBYE

However, it fails to detect that Gaussian-03.expt is present and executable (both of which are true).


To sort that out and to enable G09 detection, edit apps/data/client/cap/Gaussian-03.edml:
465 <output mimetype="chemical/x-gaussian03-output" type="parse" verifypattern="Gaussian 09, Revision">g03.g03out</output> 466 <output mimetype="chemical/x-gaussian-03-output" type="parse" verifypattern="Gaussian 09, Revision">g03.out</output> 467 <output mimetype="chemical/x-gaussian03-output" type="parse" verifypattern="Gaussian 03, Revision">g03.g03out</output> 468 <output mimetype="chemical/x-gaussian-03-output" type="parse" verifypattern="Gaussian 03, Revision">g03.out</output> .. 480 <importer>${ECCE_HOME}/scripts/parsers/Gaussian-03.expt </importer>
Your G09 files should now open properly (most of the time).

My ecce_env is fine and my runtime_setup.sh file is called by bash, but somehow it wouldn't find the Gaussian-03.expt file. Maybe it has something to do with the use of csh -f

NOTE that the file isn't imported -- it's just opened. It would've been nice if you could actually import the calculation into ECCE. Still, being able to view it is a nice start. 

609. NBO6 on a debian cluster (/w g09)

Curse blogspot and the lack of revision control and backups! I lost my post when it was almost finished.

So here's a briefer version. I have bought NBO6 and I want to integrate it with gaussian G09 rev. D (you can't use it directly with earlier binary versions)

My 'instructions' are basically copy/pasted from the NBO6 installation instructions -- this is a tl;dr version.
sudo cp nbo6.0-bin-linux-x86_64.tar.gz /opt/ cd /opt/ sudo tar xvf nbo6.0-bin-linux-x86_64.tar.gz sudo chown $USER:$USER nbo6 -R vim nbo6/bin/gaunbo6
3 set INT = i8 4 set BINDIR = /opt/nbo6/bin
In your queue file add
set path = ( /opt/nbo6/bin $path )
In my case, as I use ECCE I edited my apps/siteconfig/CONFIG.node files:
Gaussian-03Command{ set path = ( /opt/nbo6/bin $path ) setenv GAUSS_SCRDIR /home/me/scratch setenv GAUSS_EXEDIR /opt/gaussian/g09d/g09/bsd:/opt/gaussian/g09d/g09/local:/opt/gaussian/g09d/g09/extras:/opt/gaussian/g09d/g09 /opt/gaussian/g09d/g09/g09< $infile > $outfile echo 0 }
I then tested it by running a basic gaussian calculation:
%Chk=H2O_631g.chk #P rOPBE/6-31G 6D 10F SCRF=(PCM,Solvent=water) Punch=(MO) pop=(nbo6) H2O 6-31G 0 1 ! charge and multiplicity O 0.00000 0.00000 0.118491 H 0.00000 0.754898 -0.473964 H 0.00000 -0.754898 -0.473964
and got
... 411 fchk file "/home/me/scratch/Gau-24659.EFC" 412 mat. el file "/home/me/scratch/Gau-24659.EUF" 413 414 Writing Wrt12E file "/home/me/scratch/Gau-24659.EUF" 415 Gaussian matrix elements Version 1 NLab= 7 Len12L=8 Len4L=8 416 Write GAUSSIAN SCALARS from file 501 offset 0 to matrix element file. .. 429 Write ALPHA FOCK MATRIX from file 10536 offset 0 to matrix element file. 430 No 2e integrals to process. 431 Perform NBO analysis... 432 433 *********************************** NBO 6.0 *********************************** 434 N A T U R A L A T O M I C O R B I T A L A N D 435 N A T U R A L B O N D O R B I T A L A N A L Y S I S 436 ********************* Me ********************* .. 447 Filename set to /home/me/scratch/Gau-24659 .. 620 ------------------------------- 621 Total Lewis 9.99612 ( 99.9612%) 622 Valence non-Lewis 0.00029 ( 0.0029%) 623 Rydberg non-Lewis 0.00358 ( 0.0358%) 624 ------------------------------- 625 Total unit 1 10.00000 (100.0000%) 626 Charge unit 1 0.00000 627 628 $CHOOSE 629 LONE 1 2 END 630 BOND S 1 2 S 1 3 END 631 $END 632 633 Maximum scratch memory used by NBO was 62605 words 634 Maximum scratch memory used by G09NBO was 9032 words ...