Hello Myles,

great hearing from you. I actually used your packages in the first place and got Dolfin to run (with some fenics-examples from a seminar with Anders in Finland last year). But then this black-triangle problem with Viper and some mesh-errors (cgal not included) convinced me that this could be all solved with compiling it all myself. Regarding cgal I finally won't need it as a I found out finally - GMSH will be my friend it seems and the Viper issue..well I can still not tell as Dolfin is not quite running yet. Although it seems that I am close ;).

It is good to know that pacman can use packages from opensuse, sth. learned.

Steffen


On 12/05/14 19:24, Myles English wrote:
Hello Steffen,

I understand that you have made progress on your install but would just
like to add two points for the benefit of anyone else installing on Arch
Linux.

There are packages in the Arch User repository for building the fenics
components, and PETSc, Trilinos, SLEPc, etc.  Those PKGBUILDs will make
sure that Python2 is used, etc.

There are also binary packages built four times a day (or so) on a build
server here: https://build.opensuse.org/project/monitor/home:mylese to
use these packages in pacman you can add this repository to your
/etc/pacman.conf:

[home_mylese_Arch_Extra]
SigLevel = Never
Server = http://download.opensuse.org/repositories/home:/mylese/Arch_Extra/$arch

The caveats are that I haven't updated either of these methods for a
while, and I have not yet found a use for dolfin in my daily house
husband work so I wouldn't have noticed if some maintenance was
required.  On the other hand nobody has complained (recently) that they
don't work.  However, they are at least a good starting point.

Myles

Steffen Wittek writes:

when I deleted the whole FEniCS/src directory to start from scratch. It
doesn't even go beyond Trilinos:

-- CMAKE_VERSION = 3.0.2
-- Found PythonInterp: /usr/bin/python (Required is at least version "2.4")
    File "<string>", line 1
      import sys; print sys.version.split()[0]
                          ^
SyntaxError: invalid syntax
-- Python version
CMake Error at cmake/tribits/modules/FindPythonInterp.cmake:104 (IF):
    if given arguments:

      "VERSION_LESS" "2.4"

    Unknown arguments specified
Call Stack (most recent call first):
    cmake/tribits/package_arch/TribitsFidPythonInterp.cmake:44 (FIND_PACKAGE)
    cmake/tribits/package_arch/TribitsProjectImpl.cmake:121
(TRIBITS_FIND_PYTHON)
    cmake/tribits/package_arch/TribitsProject.cmake:75 (TRIBITS_PROJECT_IMPL)
    CMakeLists.txt:88 (TRIBITS_PROJECT)


-- Configuring incomplete, errors occurred!
See also
"/home/phoenix/Work/FEniCS/src/trilinos-11.4.1-Source/dorsal_build_dir/CMakeFiles/CMakeOutput.log".

It is weird. The symlink from /usr/bin/python points to python3.4. But
what confuses me are the SyntaxErrors...do I have to change some flags
somewhere?

On 12/04/14 16:58, Steffen Wittek wrote:
thanks. I deleted the working tree and am compiling again. While this
happens the compiler complains that CGAL cannot be found:

CMake Warning at cmake/modules/FindCGAL.cmake:46 (find_package):
   Could not find a package configuration file provided by "CGAL" with
any of
   the following names:

     CGALConfig.cmake
     cgal-config.cmake

   Add the installation prefix of "CGAL" to CMAKE_PREFIX_PATH or set
   "CGAL_DIR" to a directory containing one of the above files.  If "CGAL"
   provides a separate development package or SDK, be sure it has been
   installed.
Call Stack (most recent call first):
   CMakeLists.txt:512 (find_package)


-- CGAL could not be found. Be sure to set CGAL_DIR (missing:
CGAL_LIBRARIES CGAL_INCLUDE_DIRS CGAL_TEST_RUNS CGAL_VERSION_OK)
(Required is at least version "4.1")

So it seems that there is something else wrong on top. Is CGAL just a
standalone package or is it included in another one? When I had Dolfin
installed from the tarballs the python ver. used was 2.7 . If I now
include the FFC and UFC in dorsal it directly accesses python3.4 and
throws around with errors (below). I guess it could be the mismatch as
you mentioned. The way to go is to delete the whole working tree of
FEniCS/src and let dorsal wget all packages again?

####################################
   File
"/home/phoenix/Work/FEniCS/lib/python3.4/site-packages/ferari/sigdig.py",
line 41
     print "%.16e" % (a,)
                 ^
SyntaxError: invalid syntax

   File
"/home/phoenix/Work/FEniCS/lib/python3.4/site-packages/ferari/pg.py",
line 65
     raise RuntimeError, "can't compare"
                       ^
SyntaxError: invalid syntax

Fetching ufl-1.4.0
Verifying ufl-1.4.0
ufl-1.4.0.tar.gz: OK
Unpacking ufl-1.4.0
Building ufl-1.4.0
   File "setup.py", line 24
     print version
                 ^
SyntaxError: Missing parentheses in call to 'print'

####################################

Steffen

On 12/04/14 15:27, Jan Blechta wrote:
Some form files in DOLFIN source tree like
/home/phoenix/Work/FEniCS/src/dolfin-1.4.0/dolfin/ale/Poisson3D.h
are generated by FFC and they may be outdated. Try cleaning DOLFIN
working tree or force their regeneration by

   $ ./cmake/scripts/generate-form-files

If this does not help you may have some mismatch in DOLFIN, FFC, UFC
(now part of FFC) versions.

Jan


On Thu, 04 Dec 2014 15:09:08 +0000
Steffen Wittek <[email protected]> wrote:

Dear community,

I used the tarball from myles at the AUR, but unfortunately dolfin is
compiled without cgal and hence I wanted to recompile the source
myself. On top viper had some problems and I hope to fix that on top.

I got dorsal and modified another platform-file to get my own. The
tarball that was finally downloaded was dolfin-1.4.0.tar.gz
<https://bitbucket.org/fenics-project/dolfin/downloads/dolfin-1.4.0.tar.gz>.

The compilation stops however in the file MeshDisplacement.cpp with
some dozens error messages that all read similiar. The last one reads:

############################################
In file included from
/home/phoenix/Work/FEniCS/src/dolfin-1.4.0/dolfin/ale/MeshDisplacement.cpp:25:0:

/home/phoenix/Work/FEniCS/src/dolfin-1.4.0/dolfin/ale/Poisson3D.h: In
constructor 'Poisson3D::Form_a::Form_a(std::shared_ptr<const
dolfin::FunctionSpace>, std::shared_ptr<const
dolfin::FunctionSpace>)':
/home/phoenix/Work/FEniCS/src/dolfin-1.4.0/dolfin/ale/Poisson3D.h:2010:71:
error: invalid new-expression of abstract class type
'poisson3d_form_0' _ufc_form = std::shared_ptr<const ufc::form>(new
poisson3d_form_0()); ^
dolfin/CMakeFiles/dolfin.dir/build.make:2676: recipe for target
'dolfin/CMakeFiles/dolfin.dir/ale/MeshDisplacement.cpp.o' failed
############################################

I thought the pure fact that dorsal is starting with the compilation
means that all other packages that were needed were indeed found and
it is an internal problem in dolfin.
Does anyone have some input on this? If there is any further input
needed just let me know.

Thank you already,
Steffen


_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics
_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics

_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics

Reply via email to