Yes, we changed the macros in recent versions of the back-end trillium
implementation (can't remember offhand if this happened in the change to
7.0 or sometime after that). The name changed to HOST2COMPNODE.
We added a configure test to XMPI to check for which macro is defined in
the latest beta (2.2.3b8):
-----
AC_MSG_CHECKING([for HOST2COMP*])
AH_TEMPLATE(LAM_HOST2COMP, [What symbolic constant to use for HOST2COMP*])
AC_TRY_COMPILE([#include <stdlib.h>
#include <net.h>],
[int i = HOST2COMP;], HAVE_HOST2COMP=yes, HAVE_HOST2COMP=no)
if test "$HAVE_HOST2COMP" = "yes"; then
AC_DEFINE(LAM_HOST2COMP, HOST2COMP)
AC_MSG_RESULT([HOST2COMP])
else
AC_TRY_COMPILE([#include <stdlib.h>
#include <net.h>],
[int i = HOST2COMPNODE;], HAPPY=yes, HAPPY=no)
if test "$HAPPY" = "yes"; then
AC_DEFINE(LAM_HOST2COMP, HOST2COMPNODE)
AC_MSG_RESULT([HOST2COMPNODE])
else
AC_MSG_RESULT([could not find!])
AC_MSG_ERROR([*** Cannot continue])
fi
fi
-----
On Thu, 22 Jan 2004, Paulo Tribolet Abreu wrote:
> Hi all:
>
> I've been happily using xmpi under MacOSX. Today I tried it on a
> Linux machine and the make stopped with:
>
> /bin/sh ../../libtool --mode=compile mpiCC -DHAVE_CONFIG_H -I. -I.
> -I../../src/xmpi -I../../src -I/usr/X11R6/include -g -O2 -c
> ndi_resolve.cc
> mpiCC -DHAVE_CONFIG_H -I. -I. -I../../src/xmpi -I../../src
> -I/usr/X11R6/include -g -O2 -c ndi_resolve.cc -o ndi_resolve.o
> ndi_resolve.cc: In function `struct LIST * ndi_resolve(LIST *)':
> ndi_resolve.cc:69: `HOST2COMP' undeclared (first use this function)
>
> Here's my system config:
> Linux paulopc 2.2.17-21mdk #1 Sex Jun 1 10:50:46 WEST 2001 i586 unknown
> I have 48MB RAM and 256MB swap.
>
> My gcc is 2.95.3
>
> My LAM installation is 7.0.4 in /usr/local.
>
> Any ideas what might be wrong?
>
> Thanks,
>
> Paulo
> _______________________________________________
> This list is archived at
> http://www.lam-mpi.org/MailArchives/xmpi/
>
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
|
