Re: [OMPI devel] [OMPI svn] svn:open-mpi r21097

2009-04-28 Thread Rainer Keller
Ralph. On Tuesday 28 April 2009, Ralph Castain wrote: > Okay, I have found the correct fix. Please back this commit out and I > will submit the correct one. > > Just a missing header file in orte_init.c Thanks, Rainer -- Dr.-Ing. Ra

Re: [OMPI devel] [OMPI svn] svn:open-mpi r21097

2009-04-28 Thread Rainer Keller
Ralph, I know it's not right. That's why it got commited in a separate commit! BUT, somewhere in libopen-rte.so, they are marked as file-LOCAL, and are not known. CU,. Rainer On Tuesday 28 April 2009, Ralph Castain wrote: > Ummm...no, that definitely is wrong, Rainer. Please back this out - it >

Re: [OMPI devel] [OMPI svn] svn:open-mpi r21097

2009-04-28 Thread Ralph Castain
Okay, I have found the correct fix. Please back this commit out and I will submit the correct one. Just a missing header file in orte_init.c On Apr 28, 2009, at 7:55 PM, Ralph Castain wrote: Hmmm...even worse than I thought. We now have orte_name_wildcard and orte_name_invalid instantiated

Re: [OMPI devel] [OMPI svn] svn:open-mpi r21097

2009-04-28 Thread Ralph Castain
Hmmm...even worse than I thought. We now have orte_name_wildcard and orte_name_invalid instantiated everywhere that orte_globals.h is used. This breaks everything in my systems as the values are no longer defined in a central place - so they are now undefined everywhere. I thought you had t

Re: [OMPI devel] [OMPI svn] svn:open-mpi r21097

2009-04-28 Thread Ralph Castain
Ummm...no, that definitely is wrong, Rainer. Please back this out - it doubly allocates memory for those variables. Like the comment says, they are instantiated in orte_init.c. On Apr 28, 2009, at 7:44 PM, rusra...@osl.iu.edu wrote: Author: rusraink Date: 2009-04-28 21:44:43 EDT (Tue, 28