On Mar 2, 2016, at 1:02 AM, George Bosilca <bosi...@icl.utk.edu> wrote:
> 
> I am not sure why Jeff decided to implement this approach. If I remember 
> correctly last week we converged toward the solution described by Ralph (ie 
> disabling the debug build by default for everyone,  including developers).

Yes, we did.  But that hasn't happened yet.  I don't remember who was supposed 
to do that, offhand.  If we decide that disabling debug builds by default is a 
better approach than this one, no problem.  To me, they seem like complimentary 
ideas -- especially if debug builds are not the default, then it will be 
unusual to have debug builds, and therefore an extra warning is a not 
necessarily a bad thing.

Here's the message that is emitted from 
https://github.com/open-mpi/ompi/pull/1417:

# OPEN MPI: This is a debug build. DO NOT USE IT FOR PERFORMANCE RESULTS!
# OPEN MPI: Set MCA param "orte_nowarn_debug_build" to disable this msg

I didn't state it in the message, but you can also "mpirun 
--nowarn-debug-build" -- I only listed the thing you'd put in your environment 
to disable it just because that seemed like a more common thing to do.

> I wish we could fix all the cases. What we tried to achieve is to prevent the 
> most usual blunders, but without finding the perfect solution. Most of the 
> solutions we came up with put the burden on us, the OMPI developers.

+1

-- 
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to: 
http://www.cisco.com/web/about/doing_business/legal/cri/

Reply via email to