Change AC_CHECK_FILE to AC_CHECK_HEADER in all m4 files

[libmesh]

Change AC_CHECK_FILE to AC_CHECK_HEADER in all m4 files

[benkirk]

Right - the former only checks if it exists, the latter makes sure the compiler & current compile options can use it...

[benkirk]

Hey, looks like trilinos v11 can pull in PETSc as well:

configure:28876: checking /software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_include.h usability
configure:28876: mpicxx -c   conftest.cpp >&5
In file included from /software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_common.h:56:0,
                 from /software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_include.h:16,
                 from conftest.cpp:132:
/software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_config.h:120:0: warning: "HAVE_MPI" redefined [enabled by default]
conftest.cpp:93:0: note: this is the location of the previous definition
In file included from /software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_struct.h:308:0,
                 from /software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_include.h:18,
                 from conftest.cpp:132:
/software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_petsc.h:26:22: fatal error: petscksp.h: No such file or directory
compilation terminated.

So it looks like if PETSC is enabled and there is a ml_petsc.h perhaps we need to pass the PETSC includes into trilinos??

[jwpeterson]

On Wed, Dec 19, 2012 at 12:30 PM, Benjamin S. Kirk
[email protected] wrote:

Hey, looks like trilinos v11 can pull in PETSc as well:

OK, so now all we need is for PETSc to be able to include Trilinos as
a package, and then a black hole will open up a swallow the entire
universe...

configure:28876: checking /software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_include.h usability
configure:28876: mpicxx -c conftest.cpp >&5
In file included from /software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_common.h:56:0,
from /software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_include.h:16,
from conftest.cpp:132:
/software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_config.h:120:0: warning: "HAVE_MPI" redefined [enabled by default]
conftest.cpp:93:0: note: this is the location of the previous definition
In file included from /software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_struct.h:308:0,
from /software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_include.h:18,
from conftest.cpp:132:
/software/x86_64/trilinos/11.0.3-openmpi-1.4-gcc-4.6/include/ml_petsc.h:26:22: fatal error: petscksp.h: No such file or directory
compilation terminated.

So it looks like if PETSC is enabled and there is a ml_petsc.h perhaps we need to pass the PETSC includes into trilinos??

This is the problem that I've been seeing with AC_CHECK_HEADER vs.
AC_CHECK_FILE... things get crazy when you actually try to compile
stuff.

Since 'test -r' is apparently an officially-sanctioned thing [0], it
may be simpler to just use that to see if ml_include.h exists.

Of course, you're welcome to make these configure tests as complicated
as you'd like!

John

[0] http://www.gnu.org/software/autoconf/manual/autoconf.html

"To enable configure scripts to support cross-compilation, they
shouldn't do anything that tests features of the build system instead
of the host system. But occasionally you may find it necessary to
check whether some arbitrary file exists. To do so, use ‘test -f’,
‘test -r’, or ‘test -x’."