[Pkg-scicomp-devel] fw: gfortran transition release goal proposal

[Adam C Powell IV]

On Sun, 2007-07-22 at 01:19 +0300, Riku Voipio wrote:
> Hi,
> 
> The proposal to transition from the outdated g77 to gfortran did
> not result any comments. Which rises the suspicion that the maintainers
> of affected packages are not reading debian-release or debian-toolchain.

Indeed, I don't read those, sorry.  Thanks for the follow-up.

> If you have suggestions, critique, or you are busy but accept NMU'ing
> your packages for gfortran transition.. please speak now.
> 
> Some of the packages gfortran transition affects are also affected by
> the long double transition. This includes atleast the following fortran
> packages:
> 
> lam, atlas3, fftw3, hdf5, mpich1.0, petsc, pdl

I just uploaded a new PETSc which got through NEW Saturday morning.
This is an upstream upgrade with a new library package name, which fixes
the long double issue.  It also links against OpenMPI, using mpif90 as
its fortran compiler, so g77 should no longer be an issue for PETSc,
except indirectly via blas/lapack.  Please let me know if you find
otherwise.

pysparse is the same: only depends on blas/lapack.  I'll re-upload both
of these with versioned build-depends as soon as new blas/lapack
packages are in unstable.  Feel free to NMU if I take a while to do so.

As for MPICH, that's still a problem: gfortran does not work as a g77
substitute.  Here's the relevant output when I try to configure with
-f77=gfortran:

checking for routines to access the command line from Fortran 77... searching...
checking whether gfortran   works with GETARG and IARGC... no
checking whether Fortran 77 links with 0 -lU77 -lPEPCF90 -lg2c... no
checking whether Fortran 77 routine names are case-insensitive ... yes
checking whether Fortran 77 compiler accepts option -N109... no
checking whether Fortran 77 compiler accepts option -f... no
checking whether Fortran 77 compiler accepts option -YEXT_NAMES=UCS... no
checking whether Fortran 77 compiler accepts option -YEXT_NAMES=LCS... no
checking whether Fortran 77 compiler accepts option -YCFRL=1... no
checking whether Fortran 77 compiler accepts option +U77... no
configure: WARNING: Could not find a way to access the command line from Fortran 77
configure: error: Command line access is required for MPICH
Error configuring the Fortran subsystem!
Turning off Fortran support

mpich without fortran support is not good.  Here's what that section
looks like with g77 (on PPC since I don't have the i486 log):

checking for routines to access the command line from Fortran 77... searching...
checking whether f77   works with GETARG and IARGC... yes

config.log with gfortran shows:

configure:6012: checking for routines to access the command line from Fortran 77
configure:6018: result: searching...
configure:6050: checking whether gfortran   works with GETARG and IARGC
configure:6052: gfortran -o conftest -fPIC   conftest.f   1>&5
/tmp/ccwNAvX4.o: In function `MAIN__':
conftest.f:(.text+0x4c): undefined reference to `getarg_'
collect2: ld returned 1 exit status
configure:6055: $? = 1
configure:6071: result: no

So it seems getarg, which is needed for mpich, is missing from gfortran,
as are other variants used by other fortran compilers.  Is there a
gfortran equivalent?

Thanks,
-Adam
-- 
GPG fingerprint: D54D 1AEE B11C CE9B A02B  C5DD 526F 01E8 564E E4B6

Welcome to the best software in the world today cafe!
http://www.take6.com/albums/greatesthits.html