[gmx-developers] CVS Version - segmentation faults in double precision

Berk Hess hessb at mpip-mainz.mpg.de
Wed Jun 29 09:30:03 CEST 2005


Marc Baaden wrote:

>Hi,
>
>sorry for the delay. I did the tests .. and still have segmentation
>faults. I did it last Friday (24th) and re-did it today (28th) each
>time checking out a new tree, bootstrap and configure/compile.
>
>Berk, I did not quite understand, you do not any more observe the
>segmentation faults, or do you ?
>
>  
>
Currently I don't get segv's, nor in the fresh tree I had checked out
last week, nor in my development tree.
My development tree crashed before.
The only thing I have done there is make clean and compile.

It is almost impossible to debug, especially since the -g compile
seems to run fine.

I don't really know what to think anymore.
It could be something with reading unallocated memory,
which can make the crash dependent on the memory state of
the machine you are running on.
But I would have hoped that efence would catch such errors.

Another strange thing is the compile problems and crashes that people
have reported with the CVS code on AMD64. In our group we have
many different compilations on many different AMD machines and
have never encountered problems.

Erik Lindahl might be able to give some useful hints on this,
but he doesn't seem to be around.

Berk.




More information about the gromacs.org_gmx-developers mailing list