From: Cody P. <cod...@gm...> - 2013-04-02 22:48:05
|
On Tue, Apr 2, 2013 at 4:42 PM, Kirk, Benjamin (JSC-EG311) < ben...@na...> wrote: > On Apr 2, 2013, at 5:22 PM, Cody Permann <cod...@gm...> wrote: > > > > > RemoteElem should be being created from LibMeshInit (see the > Singleton::create() call and related bits in remote_elem.C) but apparently > it isn't? > > > > > > Yeah - bad news... RemoteElem::create() is _not_ being called on several > of our systems. Of course LibMeshInit is, but for some > > reason it's not getting all the way down into those new functions. > > > > On this particular system I'm using GCC 4.5.1 which is a terrible > compiler, but that's one data point. The system that I have here where it > is working is GCC 4.6.3. Our build boxes are 4.7.2 and they are also > failing so.... hard to draw conclusions other than something isn't working > quite right. ;) > > > Argh - > > the easy fix is just put a naked > > RemoteElem::create(); > > call inside libmesh.C right after Singleton::setup(); > > problem is, this is the same mechanism used for Singleton::cleanup(), so > that may fail to delete the remote_elem. > > But start there - it should run but may complain of leaked memory if the > cleanup() mechanism is misbehaving?? > > Weird - *all* my compilers on linux worked fine - gcc 4.4, 4.5, 4.6, 4.7, > icc 12.0, 12.1, 13.0, pgi's, open solaris… > > Alright, I'll give it a shot, I can't repeat the bad behavior on our Macs, even with the identical 4.7.2 compiler so it is a complicated problem indeed. > -Ben |