[Libmesh-users] Odd compiling problem with libmesh 0.4.1

[John P. <pet...@cf...>]

Hi.  In fact we have seen something like that before.  Could you
post the result of the entire $(INCLUDES) Make variable?  In the
previous case, the $(INCLUDES) looked something like:

-I/valid/path/one -I/valid/path/two -I -I/all/remaining/paths/invalid

where the "naked" -I causes all the following include directories
to be messed up.  That might be the problem in your case...

-John

Kevin Ellwood writes:
 > Hi All,
 > 
 > I am trying to install libmesh at home and I am having
 > a problem that has me befuddled.  I tried many times
 > but the basic:
 > 
 > ./configure
 > make
 > 
 > should work, I think.  I have PETSC installed in
 > /usr/local/petsc-2.1.6 and PETSC_DIR is set as is
 > PETSC_ARCH=linux-gnu.  I use gentoo linux and gcc is
 > version 3.2.2.
 > 
 > "make" chugs along without issue until:
 > Compiling C++ (in optimized mode) src/mesh/mesh.C...
 > src/mesh/mesh.C:35:22: sfcurves.h: No such file or
 > directory
 > make: *** [src/mesh/mesh.i686-pc-linux-gnu.o] Error 1
 > 
 > The odd thing is that the actual compile command has
 > -I/home/kellwood/libmesh-0.4.1/contrib/sfcurves and:
 > 
 > ls /home/kellwood/libmesh-0.4.1/contrib/sfcurves
 > cmp.c  hilbert.c  Makefile  morton.c  README 
 > sfcurves.h  sfcurves_internal.h
 > 
 > Finally, when I copy sfcurves.h to src/mesh, it
 > proceeds past mesh.C without problems.
 > 
 > Has anyone experience such oddness?  I know it is not
 > a libmesh issue but it is very strange.  Incidentally,
 > I tried it on a linux box at work (RH9) and exactly
 > the same thing happened.