On 18/02/11 14:42, Andrew Purkiss-Trew wrote:
> On Mon, 2011-02-14 at 17:44 +0000, Paul Emsley wrote:
>> On 14/02/11 17:21, Peter Simone wrote:
>>> I'm having a similar problem.
>> You mean this one?
>>
>>> Failure to find grid coordinate uvw = ( -14, 34, 21)
>>> Xmap: Internal map ASU error - P 2ac 2ab Nuvw = ( 68, 69, 75)
>>> terminate called after throwing an instance of 'clipper::Message_fatal'
>>
>>> I'm running Coot on Mac OS X and when I try to do Real Space Refine Zone using Ramachandran restraints I get essentially the exact same error message and the program closes. When I use Coot on Linux (CentOS) I don't have this problem. Any ideas what could be causing this? Thank you,
>> Hmm... I am not sure if that is an Apple compiler issue, a version issue
>> or a bug that is hidden/benign in CentOS, but critical under Mac OSX...
>>
>> Before I go hunting on the Mac (not a favourite occupation) let's be
>> clear if this is
>>
>> 1) a map issue (does this happen when you refine against other maps?)
>> 2) a refinement issue
>> 3) an issue when Rama restraints are turned on
>> 4) a structure-dependent issue...
> Just to say that I've got this problem as well. Running under linux
> (RHEL 5.6) using the Centos 5 download (build 3331).
>
> As Peter said, it seems to happen when refining using Ramachandran
> restraints when a residue with alternate conformations is involved.
>
>
Fixed now.
It was a pretty deep bug. Adding a Ramachandran restraint tickled a bug
in the link planar peptide restraint. The dictionary allowed one to add
atoms to a peptide restraints for atoms that already existed there. Due
to alternative conformations, the atom selection in some cases found
residues where there were only 2 atoms to be restrained - but when
checking against the dictionary, these atoms were found twice (or more)
each, thus seeming to provide the 4 or more atoms needed for a LSQ plane
restraint. However, there were of course only 2 atoms really, and
finding a least-squares plane through these atoms provoked numerical
instabilities, leading to atoms with undefined coordinates and then to
trying to access the map in undefined regions (which was the crash that
you saw).
Thanks to Andrew and Peter for providing failing cases.
0.6.2 is almost ready then, speak up now if there are still bugs! :-)
Paul.
|