The SIZE function is defined slightly lower down in configure (or, in
source code terms, config/lam_get_sizeof_fortran.m4). It's a C
function that is compiled separately but linked in with the fortran
source code to produce a final executable. Here's the m4 commands that
do this:
LAM_LOG_COMMAND([$CC $CFLAGS -I. -c conftest.c],
LAM_LOG_COMMAND([$F77 $FFLAGS conftestf.f conftest.o -o conftest],
LAM_LOG_COMMAND([./conftest],[HAPPY=1],[HAPPY=0]),
[HAPPY=0]),
[HAPPY=0])
The issue is that the name of the C function is variable -- it depends
on your fortran compiler. It will be one of:
size_
size
SIZE
The test takes this into account and should be using the appropriate
name for your fortran compiler. The LAM configure script is known to
work with both the PGI and Intel compilers -- I'm surprised that it's
failing for you.
Can you send your full config.log (please compress) showing a failed
fortran size test (i.e., don't use --without-fc)?
On Feb 27, 2005, at 11:06 PM, Kwan Wing Keung wrote:
>
> Hello,
>
> It looks like a naive question but I got stuck.
>
> We previously (1.5 years before) built the lam 6.5 and did not got any
> problem.
>
> We attempted to build the update lam-7.1.1 but failed upon size
> checking
> on various FORTRAN data types.
>
> We examined the source code within "configure" (a 45692 line monster!)
> and located below the corresponding code.
>
> cat > conftestf.f <<EOF
> program fsize
> external SIZE
> INTEGER x(2)
> call SIZE(x(1),x(2))
> end
>
> "configure" was aborted as the subroutine SIZE is un-resolved.
>
> We went back to the system level (standard Red Hat 9.0 linux) and
> attempted to test compiling this segment of code, but again failed both
> in using PGI Fortran and Intel Fortran, again due to the unresolved
> subroutine SIZE.
>
> We checked again within the source code of "configure", but we
> could not locate the source of this subroutine SIZE. So if this
> subroutine SIZE does not come along with "configure" and also NOT
> within PGI/Intel Fortran compiler, then where will it come from?
>> From the C or C++ compiler?
>
> Although we can now configure and make the lam-7.1.1 by
> temporarily bypassing this Fortran building portion upon specifying
> the flag "--without-fc", it seems many previous questions to this
> forum are also related to this SIZE problem upon building the
> Fortran module.
>
> Thanks
>
> W.K. Kwan
> Computer Centre
> University of Hong Kong
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
|
