> I do not believe this solves the scenario from my point of view. I > was not clear about the fact that the app binary is not recompiled. > The customer may be running older or newer OFED stack where > the changes described in my original scenario may have happened.
I'm not sure I follow. If you build a binary against some version of OFED where librdmacm is linked against libibverbs 1.1, then everything should work fine against any older version of OFED where all the needed symbols are available. I can't predict what might happen with unreleased future versions of OFED but I would guess that should work. I guess if we ever change the libibverbs binary in to 1.2 or something, and then librdmacm links against that in a future OFED release, we'll have to figure out a solution for applications linked against old librdmacm/libibverbs-1.1, but that is all theoretical at the moment. I think the only way you get in trouble is if you go out of your way to break things using dlvsym() tricks or something like that. - R. _______________________________________________ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
