[gmx-developers] LINCS broken in CVS?

[Berk Hess]

Justin MacCallum wrote:

> Hi,
>
> I'm currently working on implementing a parallel version of the 
> Adaptive Force Bias method in GROMACS. In my testing I've noticed that 
> LINCS and shake seem to be broken in the current CVS. When I take a 
> topology that worked with 3.2.1 and re-run grompp and mdrun with CVS I 
> get lincs errors. According to the output it appears that LINCS thinks 
> the constraint length is zero, despite the fact that gmxdump gives the 
> correct distances.
>
I have fixed it.
In my previous commit of clincs.c I put a 0 on the wrong side of an 
equal sign.
(stupid enough I only tested the code with zero length constaints = 
flexible constraints).

Thanks for reporting this,

Berk.