0.9.3 installation problems on GNU/Linux with lam-mpi

[Raimondo Giammanco]

On Tue, 2004-02-03 at 18:09, Rob Ross wrote:


> It's up to you.  With your modified "if" version of mpinetcdf.c and that 
> flatten.c, you might be a-ok already.  Let us know...
> 
> Regards,
> 
> Rob

Hello,

 Here is the status: with lam-7.0.4 last stable, the errors
in make testing persist. Substituting the flatten.c you
provided unfortunately did not help, this time the 
make testing stops immediately with the error message:

####################
make[1]: Entering directory
`/mnt/disposable/TMP_giamma/parallel-netcdf-0.9.3/test'
cd nc_test && make testing
make[2]: Entering directory
`/mnt/disposable/TMP_giamma/parallel-netcdf-0.9.3/test/nc_test'
./nc_test -c
MPI process rank 0 (n0, p985) caught a SIGSEGV in MPI_Type_get_envelope.
Rank (0, MPI_COMM_WORLD): Call stack within LAM:
Rank (0, MPI_COMM_WORLD):  - MPI_Type_get_envelope()
Rank (0, MPI_COMM_WORLD):  - main()
make[2]: *** [test.nc] Error 1
make[2]: Leaving directory
`/mnt/disposable/TMP_giamma/parallel-netcdf-0.9.3/test/nc_test'
make[1]: *** [testing] Error 2
make[1]: Leaving directory
`/mnt/disposable/TMP_giamma/parallel-netcdf-0.9.3/test'
make: *** [testing] Error 2
#####################

It must be considered, however, that the lamtests suite did not
execute without some tests failing; I will report 
the problem to the LAM team, and maybe I will install
mpich to study it a little (Stable 1.2.5 according to my packet
manager). Fact is, I have a little experience
with the LAM tools and I would had preferred to stick with them,
but I guess learning new things is never a bad thing.

Best Regards

-- 
Raimondo Giammanco <rongten at member.fsf.org>