[petsc-dev] gfortran-4.2 + OS X + petsc-3.2+ no longer works

[Aron Ahmadia]

I need to discuss this with the pyclaw guys, as there are serious
performance differences between 4.2 and 4.6.

I'll take a look at the gfortran symbols when I get a chance.

A

On Sun, Oct 2, 2011 at 3:17 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:

> Maybe you have spent much more time on this, but it seems premature to give
> up on a compiler series. I don't have a Mac to test on, but some notes from
> the output.
>
> On Sun, Oct 2, 2011 at 02:44, Aron Ahmadia <aron.ahmadia at kaust.edu.sa>wrote:
>
>> CMakeFiles/petsc.dir/src/sys/fsrc/somefort.F.old: warning: for symbol
>> _petscfortran9_ tentative definition of size 16 from
>> CMakeFiles/petsc.dir/src/vec/f90-mod/petscvecmod.F.o is is smaller than the
>> real definition of size 4 from
>> CMakeFiles/petsc.dir/src/sys/fsrc/somefort.F.old: warning: for symbol
>> _petscfortran9_ tentative definition of size 16 from
>> CMakeFiles/petsc.dir/src/mat/f90-mod/petscmatmod.F.o is is smaller than the
>> real definition of size 4 from
>> CMakeFiles/petsc.dir/src/sys/fsrc/somefort.F.old: warning: for symbol
>> _petscfortran9_ tentative definition of size 16 from
>> CMakeFiles/petsc.dir/src/dm/f90-mod/petscdmmod.F.o is is smaller than the
>> real definition of size 4 from
>> CMakeFiles/petsc.dir/src/sys/fsrc/somefort.F.old: warning: for symbol
>> _petscfortran9_ tentative definition of size 16 from
>> CMakeFiles/petsc.dir/src/ksp/f90-mod/petsckspmod.F.o is is smaller than the
>> real definition of size 4 from
>> CMakeFiles/petsc.dir/src/sys/fsrc/somefort.F.o
>>
>> ld: warning: for symbol _petscfortran9_ tentative definition of size 16
>> from CMakeFiles/petsc.dir/src/snes/f90-mod/petscsnesmod.F.o is is smaller
>> than the real definition of size 4 from
>> CMakeFiles/petsc.dir/src/sys/fsrc/somefort.F.o
>>
>
> It looks like this only affects PETSC_COMM_WORLD.
>
> include/finclude/petscsys.h:      common /petscfortran9/ PETSC_COMM_WORLD
> include/finclude/petscsys.h:      common /petscfortran10/ PETSC_COMM_SELF
>
>
>> Undefined symbols:
>>
>>   "_getarg_", referenced from:
>>
>>       _PETScParseFortranArgs_Private in zstart.c.o
>>
>>       _petscinitialize_ in zstart.c.o
>>
>>   "_iargc_", referenced from:
>>
>>       _PETScParseFortranArgs_Private in zstart.c.o
>>
>> ld: symbol(s) not found
>>
> Please send configure.log. If you look at src/sys/ftn-custom/zstart.c, you
> will see some logic for special cases of getarg and iargc, e.g.
>
> #define iargc_  _gfortran_iargc
> #define getarg_ _gfortran_getarg_i4
>
> What symbols are actually in your libgfortran?
>
> Also, I'm worried that configure is seeing a different libgfortran, perhaps
> because of more paths in the compilation line.
>
> And does "make all-legacy" work? (It's possible that the order of all flags
> is not preserved across languages when we push the flags to CMake.)
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20111002/1c723690/attachment.html>