[petsc-users] petsc4py Build Problem

[Lisandro Dalcin]

Dear Mikhail, Satish helped me and we managed to get petsc4py working
with static builds and Cygwin's Open MPI. Please git pull and try
again.

On 11 June 2015 at 14:13, Mikhail Khodak <mkhodak at princeton.edu> wrote:
> I have attached the log files.
> I tried using Open MPI before but received the same result. I will try a
> static build later and otherwise will get a VM.
> Thanks for looking at this,
> Mikhail
>
>
> On Thu, Jun 11, 2015 at 10:48 AM, Satish Balay <balay at mcs.anl.gov> wrote:
>>
>> Also - should mention:
>>
>> - cygwin has OpenMPI pacakged - so you could also try that.
>> - most of my builds are static [so never tried dll builds]
>>
>> Satish
>>
>> On Thu, 11 Jun 2015, Satish Balay wrote:
>>
>> > If you get errors when running basic petsc examples - send us the
>> > relavant petsc logs [configure.log, make.log, test.log etc..]
>> >
>> > Also - note that mpich is not supported on cygwin/windows [it
>> > generally works for us - when we try the --download-mpich build].
>> >
>> > Unless you really need a cywin build/use of petsc4py - it might be
>> > easier to install a linux VM - and use PETSc/petsc4py on it.
>> >
>> > Satish
>> >
>> > On Thu, 11 Jun 2015, Mikhail Khodak wrote:
>> >
>> > > I added these lines to the petsc4py test files (specifically
>> > > test_comm.py)
>> > > but the error remains the same. However, I have done the standard
>> > > 'which
>> > > mpiexec', 'which mpicc', 'which mpirun' and they are all in the same
>> > > folder. In fact it is the only MPI installed.
>> > > The reason I thought it might be a PETSc build problem is because one
>> > > of
>> > > the PETSc 'make test' tests (ex5f) fails with the same error, even
>> > > though
>> > > the 'make streams' test works fine with MPI processes.
>> > > Thanks,
>> > > Mikhail
>> > >
>> > > On Thu, Jun 11, 2015 at 4:21 AM, Matthew Knepley <knepley at gmail.com>
>> > > wrote:
>> > >
>> > > > On Thu, Jun 11, 2015 at 5:07 AM, Mikhail Khodak
>> > > > <mkhodak at princeton.edu>
>> > > > wrote:
>> > > >
>> > > >> Thank you for your help - the install seems to work, apart from
>> > > >> routines
>> > > >> requiring MPI, which fail due to the "Attempting to use an MPI
>> > > >> routine
>> > > >> before initializing MPI" error. This seems to be an error in the
>> > > >> PETSc
>> > > >> build itself.
>> > > >>
>> > > >
>> > > > The MPI routines will not work until after
>> > > >
>> > > > import petsc4py, sys
>> > > > petsc4py.init(sys.argv)
>> > > > from petsc4py import PETSc
>> > > >
>> > > > If they fail after this, it is usually a mismatch between the
>> > > > mpiexec
>> > > > being used and the MPI
>> > > > libraries being linked.
>> > > >
>> > > >   Thanks,
>> > > >
>> > > >     Matt
>> > > >
>> > > > Thanks again,
>> > > >> Mikhail Khodak
>> > > >>
>> > > >> On Mon, Jun 8, 2015 at 5:11 AM, Lisandro Dalcin <dalcinl at gmail.com>
>> > > >> wrote:
>> > > >>
>> > > >>> On 8 June 2015 at 02:50, Mikhail Khodak <mkhodak at princeton.edu>
>> > > >>> wrote:
>> > > >>> > Hello, I am trying to build petsc4py-3.5.1 using Cygwin on
>> > > >>> > 64-bit
>> > > >>> Windows 7.
>> > > >>> > I have built PETSc 3.5.4 with shared and dynamic libraries using
>> > > >>> > mpich2-1.2.1 and successfully ran the installation tests. I am
>> > > >>> > using
>> > > >>> Python
>> > > >>> > 2.7 and NumPy 1.9.2 and have installed mpi4py. However, when I
>> > > >>> > attempt
>> > > >>> to
>> > > >>> > install petsc4py (both with pip and distutils) I get a mpicc
>> > > >>> > compiler
>> > > >>> error
>> > > >>> > due to undefined references/symbols. I have attached the output
>> > > >>> > of
>> > > >>> running
>> > > >>> >
>> > > >>> > pip install petsc petsc4py --allow-external petsc
>> > > >>> > --allow-external
>> > > >>> petsc4py
>> > > >>> >
>> > > >>>
>> > > >>> I've never ever built or test petsc4py under Cygwin. The errors
>> > > >>> you
>> > > >>> see are expected.
>> > > >>> Perhaps you can manually workaround the issues following the
>> > > >>> following
>> > > >>> steps:
>> > > >>>
>> > > >>> 1) Download the petsc4py tarball and unpack it.
>> > > >>> 2) Open the file "src/libpetsc4py/libpetsc4py.h", add remove all
>> > > >>> occurences of DL_IMPORT, i.e, replace DL_IMPORT(XYZ) for just XYZ
>> > > >>> 3) Use pip again:
>> > > >>>
>> > > >>> pip install petsc
>> > > >>> pip install .
>> > > >>>
>> > > >>> The last line assumes your current working directory is the one
>> > > >>> having
>> > > >>> petsc4py's setup.py
>> > > >>>
>> > > >>> Finally, I do not guarantee this will work. I'm just guessing,
>> > > >>> petsc4py never explicitly supported Windows and/or Cygwin.
>> > > >>>
>> > > >>>
>> > > >>> --
>> > > >>> Lisandro Dalcin
>> > > >>> ============
>> > > >>> Research Scientist
>> > > >>> Computer, Electrical and Mathematical Sciences & Engineering
>> > > >>> (CEMSE)
>> > > >>> Numerical Porous Media Center (NumPor)
>> > > >>> King Abdullah University of Science and Technology (KAUST)
>> > > >>> http://numpor.kaust.edu.sa/
>> > > >>>
>> > > >>> 4700 King Abdullah University of Science and Technology
>> > > >>> al-Khawarizmi Bldg (Bldg 1), Office # 4332
>> > > >>> Thuwal 23955-6900, Kingdom of Saudi Arabia
>> > > >>> http://www.kaust.edu.sa
>> > > >>>
>> > > >>> Office Phone: +966 12 808-0459
>> > > >>>
>> > > >>
>> > > >>
>> > > >
>> > > >
>> > > > --
>> > > > What most experimenters take for granted before they begin their
>> > > > experiments is infinitely more interesting than any results to which
>> > > > their
>> > > > experiments lead.
>> > > > -- Norbert Wiener
>> > > >
>> > >
>> >
>> >



-- 
Lisandro Dalcin
============
Research Scientist
Computer, Electrical and Mathematical Sciences & Engineering (CEMSE)
Numerical Porous Media Center (NumPor)
King Abdullah University of Science and Technology (KAUST)
http://numpor.kaust.edu.sa/

4700 King Abdullah University of Science and Technology
al-Khawarizmi Bldg (Bldg 1), Office # 4332
Thuwal 23955-6900, Kingdom of Saudi Arabia
http://www.kaust.edu.sa

Office Phone: +966 12 808-0459