[petsc-dev] metis config error on XT4

Sean Farley sean at mcs.anl.gov
Wed Nov 30 19:21:11 CST 2011


>
> Luckily CMake is much easier than BuildSystem. This should be a snap!


Oh, hey there Mr. Passive Aggressive, I didn't see you come in. Well, if
you see Matt, tell him that the problem took less time to solve than it
took for BuildSystem to run (on a local disk).

Mark, the problem is that Franklin is using an old version of 'make' (3.80)
and George is using a feature from 3.81. Well, to be fair, the real problem
is that George can't decide if he wants to use cmake or not. He has set the
cmake variables for c++ flags ... but doesn't use them. So, the whole
business of loading gcc modules but using PGI compilers shouldn't be
necessary. Since George doesn't have a lick of c++ code, I think the right
thing to do is take out the c++ settings in cmake for metis / parmetis.

As for this business with George using $(abspath GKlib) in his makefile, I
think I will patch that too using Jack's suggestion. I'll roll a new
tarball tomorrow for this and email George the patches.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20111130/570d6dec/attachment.html>


More information about the petsc-dev mailing list