[gmx-developers] distance restraints and CVS gromacs

[Berk Hess]

Peter Kasson wrote:
> I'm having a couple sets of errors using distance restraints with CVS 
> Gromacs.
This is about the only thing, together with orientation restraints, that 
does not work yet
with domain decomposition.
I am working on this, so we can release a beta version soon.
>
> 1.  Parallel execution--I get an immediate segfault when running in 
> parallel.  Relevant portion of output is:
>
> Loaded with Money
>
> [host:09849] *** Process received signal ***
> [host:09849] Signal: Bus error (10)
> [host:09849] Associated errno: Unknown error: 1149974564 (1149974564)
> [host:09849] Signal code:  (1149829269)
> [host:09849] Failing at address: 0x4c894824
> [host:09851] *** Process received signal ***
> [host:09847] *** Process received signal ***
> [host:09847] Signal: Bus error (10)
> [host:09847] Associated errno: Operation not permitted (1)
> [host:09847] Signal code:  (6226608)
> [host:09847] Failing at address: 0x0
> [ 1] [0xbffff158, 0x0000000c] (-P-)
> [ 2] (vfprintf_l + 0x5e) [0xbffff188, 0x900e3b50]
> [ 3] (fprintf + 0x49) [0xbffff1b8, 0x90010c49]
> [ 4] [ 1] [0xbffff158, 0x0000000c] (-P-)
> [ 2] (vfprintf_l + 0x5e) [0xbffff188, 0x900e3b50]
> [ 3] (fprintf + 0x49) [0xbffff1b8, 0x90010c49]
> [ 4] (init_disres + 0x281) [0x0033f740, 0x000fb2bf] (F--)
> [host:09847] *** End of error message ***
> (init_disres + 0x281) [0x0033f740, 0x000fb2bf] (F--)
> [host:09849] *** End of error message ***
>
> 2.  Single-processor execution:  My test run finished successfully, 
> but the box vectors appeared to get set to 0 on the last step of an 
> NVT run.  I am currently trying to reproduce this.  The coordinates 
> still look good, which is odd.
I fixed this bug in checkpoint.c:
revision 1.9
date: 2008/06/27 15:09:17;  author: hess;  state: Exp;  lines: +2 -11
fixed bug which set matrices to zero on checkpoint writing, introduced 
in the previous revision

I guess your version of checkpoint.c is of the week before 6/27?

Berk.