On 1/27/11 11:20 AM, Johannes Ring wrote: > On Thu, Jan 27, 2011 at 11:09 AM, Harish Narayanan > <[email protected]> wrote: >> On 1/26/11 5:10 PM, Johannes Ring wrote: >>> On Wed, Jan 26, 2011 at 4:17 PM, Johannes Ring <[email protected]> wrote: >>>> [resending to list] >>>> >>>> On Tue, Jan 25, 2011 at 11:48 AM, Harish Narayanan >>>> <[email protected]> wrote: >>>>> Hello all, >>>>> >>>>> I am planning on doing a release soon. Since there have been numerous >>>>> changes since the previous release, I would like to ask platform >>>>> maintainers to check things out to detect any obvious problems. >>>>> >>>>> The following is a list of supported platforms and their statuses: >>>>> >>>>> Fedora 13 [untested] >>>>> Fedora 14 [untested] >>>>> Gentoo Linux [untested] >>>>> Ubuntu 10.04 LTS [untested] >>>>> Ubuntu 10.10 [untested] >>>>> openSUSE 11.3 [untested] >>>>> Red Hat Enterprise Linux 6 [untested] >>>> >>>> Debian GNU/Linux Unstable [64, dev - OK*] >>>> >>>>> Debian GNU/Linux Testing [untested] >>>>> Mac OS X 10.6 (with MacPorts) [64, dev - OK!] >>>>> >>>>> Please reply to this e-mail updating the status of whatever you decide >>>>> to test, and it will be helpful to point out whether your testing was on >>>>> a 32- or 64-bit machine and whether you were trying a stable/development >>>>> build of FEniCS. >>>> >>>> * DOLFIN was unable to detect SCOTCH because PT-SCOTCH is broken in >>>> the Debian package. >>> >>> The attached patch allows DOLFIN to detect and build with SCOTCH on >>> Debian sid. However, the change will affect all platforms so I'm not >>> sure if it is acceptable. >> >> Thank you for this fix. It fixes SCOTCH on snowleopard as well. > > OK, good. I think it is okay that Dorsal builds SCOTCH without > -DSCOTCH_PTHREAD since this requires the MPI library to support the > MPI_THREAD_MULTIPLE level. The OpenMPI library in Debian/Ubuntu and in > Mac OS X 10.6 does not support this. > > One can run "ompi_info | grep Thread" to check if the > MPI_THREAD_MULTIPLE level is supported by the OpenMPI library. It is > not supported if the output is "mpi: no" and SCOTCH should be built > without -DSCOTCH_PTHREAD.
OK. I will leave it as it is for now, and if this turns out to be an issue elsewhere, will add a conditional for the SCOTCH_PTHREAD flag. Harish _______________________________________________ Mailing list: https://launchpad.net/~dorsal Post to : [email protected] Unsubscribe : https://launchpad.net/~dorsal More help : https://help.launchpad.net/ListHelp
