[gmx-developers] NaNs in output with GMX 5.1 when setting box with trjconv

[Ondrej Marsalek]

Dear GROMACS developers,

I have noticed the following problem when using GROMACS 5.1, while
with 5.0.2, everything works as expected. When I try to set a cubic
box using trjconv with input in XYZ format, I get NaNs for for some of
the box variables (different ones is PDB and GRO). If I do the
following:

$ babel $GMXDATA/top/tip4p.gro test.xyz
$ echo 0 | gmx trjconv -s $GMXDATA/top/tip4p.gro -f test.xyz -o
test.pdb -box 1.0 1.0 1.0
$ echo 0 | gmx trjconv -s $GMXDATA/top/tip4p.gro -f test.xyz -o
test.gro -box 1.0 1.0 1.0

I get output files that have strange numbers and NaNs:

$ tail -n 1 test.gro
   1.00000   1.00000   1.00000   0.00000   0.00000   0.00000
0.00000-50052.06641      -nan

$ grep CRYST test.pdb
CRYST1   10.000   10.000     -nan   -nan   -nan  90.00 P 1           1

Doing the same with 5.0.2 gives the expected result:

$ tail -n 1 test.gro
   1.00000   1.00000   1.00000

$ grep CRYST test.pdb
CRYST1   10.000   10.000   10.000  90.00  90.00  90.00 P 1           1

Starting from the tip4p.gro file directly does not show the problem.
My motivation is that I have an NVT trajectory in XYZ format and want
to "inject" the box so I can use it for further analysis. Regardless
of that, I thought this might be indicative of a regression between
5.0.2 and 5.1

Best regards,
Ondrej