Okay, we seem to have this fixed finally. I've confirmed that the libevent 2.0 upgrade is working. Jeff and I plan to do some further cleanup of the configure integration, but that will be transparent to operations.
So please check the system out: http:/bitbucket.org/rhc/ompi-lib2 and let me know if you see any problems. Otherwise, I will bring this to the trunk Oct 22nd. Thanks Ralph On Sep 18, 2010, at 6:12 PM, Ralph Castain wrote: > Hold off on this a little. I discovered a problem with the stupid build > system again that is causing libevent to try and install when it shouldn't. > Afraid I'll have to wait for help from the "build guru" when he returns. > > I'll update you when the build system allows us to continue. > Ralph > > > On Sep 18, 2010, at 10:53 AM, Ralph Castain wrote: > >> Hi all >> >> As mentioned on the telecons and in prior mails, I have been working on >> updating the event library to libevent 2.0. While doing so, I have >> significantly simplified the integration so that we can perform such updates >> much quicker/easier in the future. Basically, it now requires simply >> unpacking the tarball and modifying two files to include a line or two. >> >> Because of the integration change, it is possible that some things were >> missed. Therefore, it would be a REALLY GOOD IDEA for people to test my >> Mercurial branch before I bring this stuff over to the trunk. >> >> See the branch at http:/bitbucket.org/rhc/ompi-lib2 >> >> I know we have a bad history of people not doing such testing, but I'm >> asking you to make a special effort here. Otherwise, you really don't have a >> good reason to complain later. :-) >> >> This update retains all the prior behaviors - it just puts libevent 2.0 as >> the engine. Soon after this has soaked a little, we will begin implementing >> the advanced features available in this update. Specifically, we will create >> separate event bases for the MPI and ORTE layers, thus allowing us to >> progress those layers independent of each other. >> >> I will continue to do some work on the branch (cleanups, remove >> opal_progress_thread stuff since it doesn't work anyway), but will ensure >> that the branch always works for those of you testing it. Barring any >> reported problems, I will check this into the trunk in about a week. >> >> Thanks >> Ralph >> >
