[gmx-users] problem with g_bar

[Tom Kirchner]

Dear all,

Recently I posted my problem with g_bar on this mailing list but got no 
solution for it. After additional testing I decided to send an enhanced 
description of my problem.

I am doing Free Energy calculation. The dhdl values are put in separate 
.xvg files. The routine g_bar is then started using:

g_bar -b 100 -f md_*.xvg -o -oi -oh

With the result:

Program g_bar, VERSION 4.5.5
Source code file: .../src/gromacs-4.5.5/src/tools/gmx_bar.c, line: 2054

Fatal error:
Temperature in file md_0.05.xvg different from earlier files or setting


Checking the header of all .xvg files showed the same line for the 
temperature for all files:

@ subtitle "T = 298.15 (K), \xl\f{} = 0.025"

Changing the temperature manually to any integer (but still with .0) let 
g_bar run fine.

sed -i 's/T\ \=\ 298.15/T\ \=\ 298.0/g' md*

Do you have any suggestions what's wrong here?

I'm not that familiar with C, but my suggestion would be an error with 
floating point arithmetic in line 2052 of gmx_bar.c

     if ( ( *temp != barsim->temp) && (*temp > 0) )

where *temp and barsim->temp are compared. Maybe it has something to do 
with the input as long double (%lf) and statement "real *temp" in the 
function header?

I appreciate any hints.

Best regards
Tom