Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |  

This web mail archive is frozen.

This page is part of a frozen web archive of this mailing list.

You can still navigate around this archive, but know that no new mails have been added to it since July of 2016.

Click here to be taken to the new web archives of this list; it includes all the mails that are in this frozen archive plus all new mails that have been sent to the list since it was migrated to the new archives.

Subject: [OMPI devel] OMPI 1.4.5rc1 tested: MacOS/ppc and *BSD/amd64
From: Paul H. Hargrove (PHHargrove_at_[hidden])
Date: 2011-12-13 22:58:13


I have tried 8 non-mainstream test platforms, using their "/usr/bin/gcc"
compilers:

MacOS 10.3 (Panther) on PPC:
    gcc (GCC) 3.3 20030304 (Apple Computer, Inc. build 1666)
MacOS 10.4 (Tiger) on PPC:
    powerpc-apple-darwin8-gcc-4.0.1 (GCC) 4.0.1 (Apple Computer, Inc.
build 5341)
MacOS 10.5 (Leopard) on PPC:
    powerpc-apple-darwin9-gcc-4.0.1 (GCC) 4.0.1 (Apple Inc. build 5488)
FreeBSD-8.2-RELEASE on amd64:
    gcc (GCC) 4.2.1 20070719 [FreeBSD]
FreeBSD-7.2-RELEASE on amd64:
    gcc (GCC) 4.2.1 20070719 [FreeBSD]
FreeBSD-6.3-RELEASE on amd64:
    gcc (GCC) 3.4.6 [FreeBSD] 20060305
NetBSD-5.1 on amd64:
    gcc (GCC) 4.1.3 20080704 prerelease (NetBSD nb2 20081120)
OpenBSD-5.0 on amd64:
    gcc (GCC) 4.2.1 20070719

On all but two I was able to configure w/ only a --prefix option, and
then "make", "make install" and "make check" w/o errors (where "gmake"
replaces "make" on the BSD hosts).

On OpenBSD I had to configure with --disable-io-romio, because the build
failed otherwise.
The errors are the same as I experienced with an older OpenBSD-4.6 on
ompi-1.4.3rc1:
http://www.open-mpi.org/community/lists/devel/2010/08/8338.php

On MacOS 10.3 (Panther) I experienced the following failure at "make" time:
> libtool: compile: gcc -DHAVE_CONFIG_H -I. -I../../../opal/util
> -I../../opal/include -I../../orte/include -I../../ompi/include
> -I../../opal/mca/paffinity/linux/plpa/src/libplpa -I../../.. -I../..
> -I../../../opal/include -I../../../orte/include
> -I../../../ompi/include -D_REENTRANT -O3 -DNDEBUG -finline-functions
> -fno-strict-aliasing -MT path.lo -MD -MP -MF .deps/path.Tpo -c
> ../../../opal/util/path.c -fno-common -DPIC -o .libs/path.o
> ../../../opal/util/path.c: In function `opal_path_nfs':
> ../../../opal/util/path.c:508: error: `MFSTYPENAMELEN' undeclared
> (first use in this function)
> ../../../opal/util/path.c:508: error: (Each undeclared identifier is
> reported only once
> ../../../opal/util/path.c:508: error: for each function it appears in.)
Which is from the following in opal/util/path.c:
> #elif (defined(__APPLE__) && defined(__MACH__))
> for (i = 0; i < FS_TYPES_NUM; i++)
> if (0 == strncasecmp (fs_types[i].f_fsname, buf.f_fstypename,
> MFSTYPENAMELEN))
> goto found;
On this system, it appears that MFSNAMELEN is the proper length (as w/
FreeBSD):
> $ man statfs | grep NAMELEN
> #define MFSNAMELEN 15 /* length of fs type name, not inc. nul */
> #define MNAMELEN 90 /* length of buffer for returned name */
> char f_fstypename[MFSNAMELEN]; /* fs type name */
> char f_mntonname[MNAMELEN]; /* directory on which
> mounted */
> char f_mntfromname[MNAMELEN]; /* mounted file system */
> $ grep -rl MFSNAMELEN /usr/include/
> /usr/include/sys/mount.h
By editing opal/util/path.c to make that change, I was STILL able to
complete the "make", reaching the following error linking ompi_info:
> /bin/sh ../../../libtool --tag=CXX --mode=link g++ -O3 -DNDEBUG
> -finline-functions -export-dynamic -o ompi_info components.o
> ompi_info.o output.o param.o version.o ../../../ompi/libmpi.la
> libtool: link: g++ -O3 -DNDEBUG -finline-functions -o .libs/ompi_info
> components.o ompi_info.o output.o param.o version.o -Wl,-bind_at_load
> ../../../ompi/.libs/libmpi.dylib
> /home/phargrove/openmpi-1.4.5rc1/BUILD-dftl/orte/.libs/libopen-rte.dylib
> /home/phargrove/openmpi-1.4.5rc1/BUILD-dftl/opal/.libs/libopen-pal.dylib
> ld: Undefined symbols:
> _pthread_atfork
Clearly -lpthread is needed "somewhere", but I didn't pursue this
platform any deeper.

In the unlikely event anybody wants to follow up on either failure, let
me know.

There were a fair number of warnings at build time on several of the builds.
I will report any that I find alarming IF/when I've had a chance to pick
through them.

Note that I've not *run* any MPI applications yet.

-Paul

On 12/8/2011 2:13 PM, Jeff Squyres wrote:
> 1.4.5rc1 is now posted in the usual place:
>
> http://www.open-mpi.org/software/ompi/v1.4/
>
> Gearing up for a pre-Christmas release -- please test! There have only been a few bug fixes since 1.4.4. See http://svn.open-mpi.org/svn/ompi/branches/v1.4/NEWS for details.
>
> (note that we'll likely have a 1.5.5rc1 shortly, too)
>

-- 
Paul H. Hargrove                          PHHargrove_at_[hidden]
Future Technologies Group
HPC Research Department                   Tel: +1-510-495-2352
Lawrence Berkeley National Laboratory     Fax: +1-510-486-6900