[gmx-developers] memory leaks in addconf.c

[David van der Spoel]

On 2012-05-05 00:01, Jochen Hub wrote:
>
>
> Am 5/4/12 7:37 PM, schrieb David van der Spoel:
>> On 2012-05-04 15:21, Jochen Hub wrote:
>>> Hi,
>>>
>>> the memory allocated by genbox with option -ci blows up within seconds
>>> to, say, 2 GB or so. Apparently this is due to lots of snew's in the
>>> file addconf.c (functions add_conf(), , that do not have the respective
>>> sfree. A short grep on addconf.c shows that from 17 snew's in the file
>>> only 3 have an sfree. Probably there a few more snew in init_mtop,
>>> init_t_atoms, init_mdatoms, init_nrnb, init_forcerec,
>>> init_neighbor_list, init_neighbor.
>>>
>>> I guess this is not the indented behavior? Should I file a bugzilla on
>>> this issue?
>>>
>>> Cheers,
>>> Jochen
>>>
>> Yes. Redmine that is.
>> I thought there already was one, but can not find it.
>
> OK, one question before I submit a bug report: I could also submit a
> patch to gerrit which does a simple all-to-all distance check (without
> neighbor searching) when inserting a molecule. For my example (that
> required approximately 12x20.000 distance checks per insertion) this was
> very fast. And it does not leak memory, of course. I could switch on
> this feature with an option for now.
>
> Together with this patch, I would submit a little extension to genbox
> that allows to insert molecules near given positions (instead of at
> random positions).
>
> Would such a patch have a chance to go through gerrit? I feel that this
> is not the extra-elegant but a working solution for now. Please let me
> know.
>
> Cheers,
> Jochen
>
>
>
I think it could go through since it is not mdrun, but others may think 
otherwise.

-- 
David van der Spoel, Ph.D., Professor of Biology
Dept. of Cell & Molec. Biol., Uppsala University.
Box 596, 75124 Uppsala, Sweden. Phone:	+46184714205.
spoel at xray.bmc.uu.se    http://folding.bmc.uu.se