Re: [Libmesh-users] nonlinear solver in libmesh library

[John P. <jwp...@gm...>]

On Wed, Mar 5, 2008 at 8:17 AM, TCHOUANMO Stephane DOCTORANT
<ste...@ir...> wrote:
> Hi,
>
>  Im back again with the petsc non linear solver in Libmesh. I'm trying to make it run under a Red Hat WS 4.
>  First I installed one of the last different versions of Petsc (2.3.3, 2.3.2 and  2.3.1) in the Libmesh directory and it worked. After that, I changed the "Make.common" in Libmesh to enable the option package Petsc, giving infos about the version, minor-major, petsc-arch and petsc-dir. Then I got into "libmesh.config" to define petsc as a solver to be used by setting: "#define HAVE_PETSC 1". Finally I configured Libmesh and compiled it with the sequences: "./configure" and "make".

OK, that is a problem: Once you re-run configure, you blow away any
changes you might have made to Make.common or libmesh_config.h.  Those
2 files are auto-generated during the configure process.

-J



Everything went well and the file "libmesh.so" got created.
>  Now Im trying to run a basic example given in Libmesh like "ex3". When I lauch "make run", I get a warning and one error which are:
>  -------------------------------------------------------------------------------------------------------
>   /home/tchouanm/libmesh-lma/contrib/tecplot/lib/i686-pc-linux-gnu/tecio.a(tecxxx.o)(.text+0x1a7): In function `tecini':
>  : warning: the use of `mktemp' is dangerous, better use `mkstemp'
>
>  /home/tchouanm/libmesh-lma/lib/i686-pc-linux-gnu_opt/libmesh.so: undefined reference to `std::basic_istream<char, std::char_traits<char> >::ignore()@GLIBCXX_3.4.5'
>  collect2: ld returned 1 exit status
>  make: *** [ex3] Erreur 1
>  -------------------------------------------------------------------------------------------------------
>
>   It seems like there's something wrong in "libmesh.so" but I cant understand why. What's the cause of the "undefined reference..." ?
>
>  Thank's a lot for your help.
>  Regards,
>
>  Stephane.
>
>
>
>  -----Message d'origine-----
>  De : John Peterson [mailto:pet...@cf...]
>  Envoyé : jeudi 21 février 2008 17:20
>  À : TCHOUANMO Stephane DOCTORANT; lib...@li...
>  Objet : nonlinear solver in libmesh library
>
>  Hi,
>
>  Yes, we have run the PETSc nonlinear solver before.  Before we can
>  help you, it seems you will need to ask a specific question, provide
>  the exact error messages you are getting (translated to English where
>  appropriate) and also tell us what you have tried doing to fix the
>  problem so far.
>
>  Also, I copied your email to the libmesh-users list.  This is the best
>  place to get answers (better than just asking 1 of the developers) and
>  it automatically archives the discussions so they can potentially help
>  out others later.
>
>  -J
>
>  TCHOUANMO Stephane DOCTORANT writes:
>   > Hi John,
>   >
>   > Once again I'm coming back to you to have some infos on Libmesh.
>   > Have you ever tried to run the non linear solver PETSC or any other
>   > nonlinear solver in Libmesh? I know that the code is supposed to
>   > work with PETSC's non linear solver and tried to make it run in
>   > vain. So I was wondering if you had experienced it in the past so
>   > you could help me out.  I have errors basically on MPI in Libmesh
>   > and I really don't understand what the deal with PETSC is. Do you
>   > have any idea about running PETSC in the code?
>   >
>   > Thanks a lot.
>   >
>   > Stéphane
>   >
>