[gmx-users] Lingering problems with the test set

[Mark Abraham]

Justin A. Lemkul wrote:
> 
> Hi All,
> 
> I just grabbed the latest version of the test set from CVS, noting the 
> updates and improvements to the package.  I ran gmxtest.pl, and all 

Actually there were a bunch more improvements that I'd made which hadn't 
been correctly added to CVS. Erik's fixed that now. Road-testing would 
be welcome. There aren't any more bd-temp issues, and he seems to have 
taken up your suggestion below also.

Mark

> simple, complex, and kernel tests passed, but pdb2gmx failed.  Looking 
> into the pdb2gmx.log file, grompp complains about "Unknown or double 
> left-hand 'bd-temp' in parameter file."  Commenting this line out in the 
> em.mdp file circumvents the problem, but leads to multiple failures of 
> the ffoplsaa+tip4p/tip5p (-vsite none) tests involving the 1pga.pdb 
> structure.  The problem is naming inconsistency between MW-HW3 in the 
> case of tip4p, and LP1/LP2-HW3/HW4 in the case of tip5p.  Of course, 
> these errors come from pdb2gmx processing of the crystal waters and use 
> of the hydrogen database.
> 
> To circumvent all of these problems in one move, I made the following 
> change to the gmxtest.pl script:
> 
> 355c355
> <             open(PIPE,"$grompp -f ../../../../em -c b4em 2>&1 |");
> ---
>  >             open(PIPE,"$grompp -maxwarn 10 -f ../../../../em -c b4em 
> 2>&1 |");
> 
> Perhaps this change would be appropriate for the general public as 
> well.  With this modification, every test passes as expected, since 
> grompp will correctly use the names from the topology.
> 
> It might also be appropriate to kill the bd-temp line in em.mdp, since 
> this is not a valid .mdp option any more :)
> 
> -Justin
>