> > on your system where compilation fails with mpif90, you can > mpif90 --showme ... > in order to get the gfotran command line that is used by the wrapper > then you can compare it to your own gfortran command line that is working, > and then try to figure out which difference is involved in the crash.
This was a very helpful tip, thank you! This allowed me to find that the mpif90-based compile flag that leads to the crash is "-Wl,-flat_namespace". I can use the rest of the command line given by mpif90 --showme to generate a working module (no errors on print statements). Can I configure mpi to not use the -Wl,-flat_namespace flag by default? Obviously I don't want to do that if it's a critical flag to have, although my code appears to run just fine without it... On Mon, Apr 3, 2017 at 9:07 PM, Gilles Gouaillardet < gilles.gouaillar...@gmail.com> wrote: > Hi, > > on your system where compilation fails with mpif90, you can > mpif90 --showme ... > in order to get the gfotran command line that is used by the wrapper > then you can compare it to your own gfortran command line that is working, > and then try to figure out which difference is involved in the crash. > > if you use the fortran module (e.g. use mpi or use mpi_f08), then > keep in mind the gfortran used by the wrapper must be the same than > the gfortran used to build Open MPI : fortran modules might not be > compatible between different gfortran versions) > > > if the generated driver.F90 is simple enough, then you can post it to > the mailing list so some of us can have a look at it > > Cheers, > > Gilles > > On Tue, Apr 4, 2017 at 10:53 AM, Austin Herrema <aherr...@iastate.edu> > wrote: > > Hello, > > > > I have encountered a strange bug that I suspect to be related to Open MPI > > and I'm hoping somebody might be able to point me in the right direction > to > > solve it. > > > > For almost all other cases, my Open MPI installation seems to be working > > correctly. I'm on OS X El Capitan 10.11.6 and have Open MPI 2.1.0 > installed > > via Homebrew. I'm attempting to run an f2py-wrapped (ultimately compiled > > using mpif90) fortran code in Python 2.7. In a simple python script, I > can > > import my .so and run the code without problems: > > > > import module_name > > module_name.execute() > > > > and I see the expected output. However, when I also import certain > > packages... > > > > import networkx as nx > > importmodule_name > > module_name.execute() > > > > (I haven't been able to determine the rhyme or reason behind which > > packages), I get an error like: > > > > At line 79 of file driver.F90 > > Internal Error: list_formatted_write(): Bad type > > > > The line number and error always corresponds with the first write > statement > > in the fortran subroutine. The reasons I expect this to be Open > MPI-related > > are 1) the web searches I've done on this issue tend to point to the > > gfortran libraries being out of date or something like that (not the case > > for me I don't think, but still) and 2) if I instead use gfortran to > compile > > the f2py-wrapped code, then these simple tests always work (for actual > use, > > though, I'd like to use Open MPI). My test fortran code is literally > just a > > write statement. > > > > I've tried re-compiling/re-installing Open MPI via homebrew and have > tried a > > couple different python2.7 installations (one of them Anaconda). > gfortran is > > also installed via Homebrew. I can successfully run the same tests on a > > linux machine--I don't get the error even when compiling with mpif90--so > it > > appears to somehow be specific to my machine. > > > > I hope this is the correct mailing list for this question. If not, my > > apologies and I will resend elsewhere. > > > > Any thoughts at all? Thank you, > > Austin > > > > > > > > -- > > Austin Herrema > > PhD Student | Graduate Research Assistant | Iowa State University > > Wind Energy Science, Engineering, and Policy | Mechanical Engineering > > > > _______________________________________________ > > devel mailing list > > devel@lists.open-mpi.org > > https://rfd.newmexicoconsortium.org/mailman/listinfo/devel > _______________________________________________ > devel mailing list > devel@lists.open-mpi.org > https://rfd.newmexicoconsortium.org/mailman/listinfo/devel > -- *Austin Herrema* PhD Student | Graduate Research Assistant | Iowa State University Wind Energy Science, Engineering, and Policy | Mechanical Engineering
_______________________________________________ devel mailing list devel@lists.open-mpi.org https://rfd.newmexicoconsortium.org/mailman/listinfo/devel
