← Back to team overview

fenics team mailing list archive

Re: FEniCS 0.1.0-pre2

 

With the discussion about experiments to selecting a cross-platform build
system, this whole problem of binary releases might get rolled into one and
the same effort. If there is a cross-platform build process than we'll also
need a cross-platform naming convention and QA process. I have been manually
constructing DevStudio build projects because I like native win32 processes
on Windows and if the FEniCS source management was slightly different the
build processes would simplify for all platforms (basically avoid the moving
of header files after component builds so that you can take advantage of a
single include path for compilation). However, this makes no statement about
configuration for which I know nothing but see lots of complexity in the
dolfin-config scripts. 

I still believe that there might be some use in using BuildSystem as the
configuration system for scons, or maybe the other way around, integrate the
Env functionality of scons into BuildSystem. Unfortunately, to take
advantage of the code browsing functionality of DevStudio it needs a
solution file. If this is not included in the requirements of the build
system then we'll always have a non-QAable situation in the sense that the
build descriptions to build the system might not be in sync with the build
description that would be used by some developers and vice versa. The
DevStudio IDE is amazingly productive for code comprehension: providing you
the functionality to look up definition/declaration of any symbol and macro.

In the mean time I'll continue to work on a native DevStudio project
description for FEniCS since I seem to be the only one that develops on
Windows. 

Theo

-----Original Message-----
From: fenics-dev-bounces@xxxxxxxxxx [mailto:fenics-dev-bounces@xxxxxxxxxx]
On Behalf Of Robert C. Kirby
Sent: Monday, September 18, 2006 3:18 PM
To: 'Anders Logg'
Cc: 'Discussion of DOLFIN development'; fenics-dev@xxxxxxxxxx
Subject: Re: [FEniCS-dev] FEniCS 0.1.0-pre2

I haven't gotten into building DOLFIN on my new machine, but it is
64-bit Ubuntu.  I was able to build Trilinos and Sundance fairly simply
by adding -fPIC to the C++, C, and FFLAGS in both libraries.  This was
only required for the Python bindings to each one.  The C++ versions
would build fine without and used about the same build scripts I used on
a 32-bit machine (up to library & mpi locations).

This is intended as a suggestion to look at building 64-bit things in
ubuntu if that is an option rather than suggesting some superiority of
competitive projects.  There have been some headaches getting those
libraries built on 64-bit red hat.

Rob Kirby


On Mon, 2006-09-18 at 20:23 +0200, 'Anders Logg' wrote:
> ok, no wonder.
> 
> I'm pretty sure it has be a different build for 64bit systems. No need
> to make it even more complicated since we're not even sure it works as
> expected on standard 686 Linux 32bit systems.
> 
> /Anders
> 
> 
> On Mon, Sep 18, 2006 at 02:17:08PM -0400, tomtzigt wrote:
> > SuSE 10 is an x86_64 target. That would explain why all the binary
> > distributions have a 32bit and a 64bit distribution. Ok, do you or your
IT
> > support staff have an idea if we can fake a 32bit target on SuSE? I have
to
> > believe you can, but I can't find anything on the either the help system
or
> > even google for the moment (tried "running 32bit code on 64bit linux").
> > 
> > Theo
> > 
> > -----Original Message-----
> > From: fenics-dev-bounces@xxxxxxxxxx
[mailto:fenics-dev-bounces@xxxxxxxxxx]
> > On Behalf Of 'Anders Logg'
> > Sent: Monday, September 18, 2006 2:06 PM
> > To: fenics-dev@xxxxxxxxxx; 'Discussion of DOLFIN development'
> > Subject: Re: [FEniCS-dev] FEniCS 0.1.0-pre2
> > 
> > Is this a 686 Linux machine?
> > 
> > This really looks like a problem you would get if you compile on a
> > different architecture.
> > 
> > /Anders
> > 
> > 
> > On Mon, Sep 18, 2006 at 02:02:56PM -0400, tomtzigt wrote:
> > > Still the same problem with respect to the format of the binaries. Is
> > there
> > > a command line tool that we can use to quantify the .so format on
these
> > > different machines and compiler sets?
> > > 
> > > 
> > > 
> > > Ignoring the old RedHat box, here is the SuSE log for the new pre2
> > release:
> > > 
> > >
tomtzigt@tomtzigt-ws1:~/fenics/fenics-0.1.0-pre2/demo/dolfin/pde/poisson>
> > > make
> > > `dolfin-config --compiler` `dolfin-config --cflags` -c main.cpp
> > > `dolfin-config --linker` -o dolfin-poisson main.o `dolfin-config
--cflags`
> > > `dolfin-config --libs`
> > > libtool: link: warning: library
> > > `/home/tomtzigt/fenics/fenics-0.1.0-pre2/lib/libdolfin.la' was moved.
> > > libtool: link: warning: library
> > > `/home/tomtzigt/fenics/fenics-0.1.0-pre2/lib/libxml2.la' was moved.
> > > libtool: link: warning: library
> > > `/home/tomtzigt/fenics/fenics-0.1.0-pre2/lib/libdolfin.la' was moved.
> > > libtool: link: warning: library
> > > `/home/tomtzigt/fenics/fenics-0.1.0-pre2/lib/libxml2.la' was moved.
> > > /home/tomtzigt/fenics/fenics-0.1.0-pre2/lib//libdolfin.so.0: could not
> > read
> > > symbols: File in wrong format
> > > collect2: ld returned 1 exit status
> > > make: *** [dolfin-poisson] Error 1
> > > 
> > >
tomtzigt@tomtzigt-ws1:~/fenics/fenics-0.1.0-pre2/demo/dolfin/pde/poisson>
> > > gcc --version
> > > gcc (GCC) 4.0.2 20050901 (prerelease) (SUSE Linux)
> > > 
> > > Theo
> > > 
> > > -----Original Message-----
> > > From: fenics-dev-bounces@xxxxxxxxxx
[mailto:fenics-dev-bounces@xxxxxxxxxx]
> > > On Behalf Of Anders Logg
> > > Sent: Monday, September 18, 2006 12:19 PM
> > > To: Discussion of FEniCS development; Discussion of DOLFIN development
> > > Subject: [FEniCS-dev] FEniCS 0.1.0-pre2
> > > 
> > > Please try the new package:
> > > 
> > >     http://www.fenics.org/pub/software/fenics/v0.1/
> > > 
> > > Unpack the file, run ./install and type 'source fenics.setup'.
> > > 
> > > Then go to demo/dolfin/pde/poisson and type 'make'.
> > > 
> > > If this works, then run the program by './dolfin-poisson' and plot
> > > the solution with 'python plot.py'.
> > > 
> > > This new version has been built with apg++ from autopackage which
> > > could solve some problems. It has not been double compiled with
> > > support for old g++ versions so you need a reasonably fresh g++
> > > install.
> > > 
> > > Does it work for anyone?
> > > 
> > > /Anders
> > > 
> > > md5sum: 66d7b40f9dc8f5ed626019e7bbb4cb04
> > > _______________________________________________
> > > FEniCS-dev mailing list
> > > FEniCS-dev@xxxxxxxxxx
> > > http://www.fenics.org/mailman/listinfo/fenics-dev
> > > 
> > > 
> > > 
> > > _______________________________________________
> > > FEniCS-dev mailing list
> > > FEniCS-dev@xxxxxxxxxx
> > > http://www.fenics.org/mailman/listinfo/fenics-dev
> > _______________________________________________
> > FEniCS-dev mailing list
> > FEniCS-dev@xxxxxxxxxx
> > http://www.fenics.org/mailman/listinfo/fenics-dev
> > 
> > 
> > 
> > _______________________________________________
> > FEniCS-dev mailing list
> > FEniCS-dev@xxxxxxxxxx
> > http://www.fenics.org/mailman/listinfo/fenics-dev
> _______________________________________________
> FEniCS-dev mailing list
> FEniCS-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/fenics-dev
-- 
Robert C. Kirby
Associate Professor
Department of Mathematics and Statistics
Texas Tech University
http://www.math.ttu.edu/~kirby

"Mathematical software should be mathematical"

_______________________________________________
FEniCS-dev mailing list
FEniCS-dev@xxxxxxxxxx
http://www.fenics.org/mailman/listinfo/fenics-dev




References