On Thu, 8 Jan, 2015 at 12:37 AM, Juan Luis Cano <[email protected]>
wrote:
On 2015-01-07 08:56, Johannes Ring wrote:
On Tue, Jan 6, 2015 at 8:39 PM, Jan Blechta
<[email protected]> wrote:
This is ok. PETSc with CUSP is not needed. Actually, CUSP backend is
not well maintained and will be probably removed one day.
The problem seems to be
=========================================================================
CMake Error at cmake/modules/FindSCOTCH.cmake:124 (file):
file STRINGS file
"/home/juanlu/miniconda3/conda-bld/work/dolfin-1.4.0/SCOTCH_INCLUDE_DIRS-NOTFOUND/scotch.h"
cannot be read.
Call Stack (most recent call first):
CMakeLists.txt:396 (find_package)
CMake Error at cmake/modules/FindSCOTCH.cmake:125 (string):
string sub-command REGEX, mode MATCH needs at least 5 arguments
total to
command.
Call Stack (most recent call first):
CMakeLists.txt:396 (find_package)
CMake Error at cmake/modules/FindSCOTCH.cmake:126 (string):
string sub-command REGEX, mode MATCH needs at least 5 arguments
total to
command.
Call Stack (most recent call first):
CMakeLists.txt:396 (find_package)
=========================================================================
Yes, this problem has already been fixed in the development version
of DOLFIN.
Johannes
Thanks Johannes and Jan. You were right: after applying the changes
from dolfin-1.4.0 to current master to FindSCOTCH.cmake the problem
disappears.
The last thing I think I need to finish is related more to PETSc
itself. When building the package using `python setup.py install`,
all the shared libraries go to
"$PREFIX/lib/python2.7/site-packages/petsc/lib", and this is causing
trouble because that is not an standard path for SWIG and others to
look for shared libraries. Modifying LD_LIBRARY_PATH to include it is
tempting but not suitable probably - I even thought of making
symlinks from $PREFIX/lib at build time. If anyone has other
suggestion for making these libraries generally available it can save
me the step of asking this same question in the petsc mailing list.
The documentation for its build system is a bit dated, and setting
--prefix in some different forms had no effect.
I suggest bringing this up on the PETSc mailing list - there has been
some recent discussion on the PETSc list on the difficulties caused by
the non-standard install paths that they use.
Garth
Juan Luis
Juan Luis
_______________________________________________
fenics-support mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics-support
_______________________________________________
fenics-support mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics-support
