open TELEMAC-MASCARET The mathematically superior suite of solvers

Dear all
I have a problem when I run a simulation which needs to be recompiled (when including a new fortran file).
The whole installation compiles fully and correctly with executables and libraries in place in <root>/builds/...etc

When I run without a fortran file - the run completes successfully.

With a different fortran file the run fails with an error:

Warning: Nonexistent include directory "[root]/builds/linux_ompi_gfort/lib/utils/special"
Warning: Nonexistent include directory "[root]/builds/linux_ompi_gfort/lib/utils/parallel"
Warning: Nonexistent include directory "[root]/builds/linux_ompi_gfort/lib/utils/damocles"
Warning: Nonexistent include directory "[root]/builds/linux_ompi_gfort/lib/utils/bief"
Warning: Nonexistent include directory "[root]/builds/linux_ompi_gfort/lib/sisyphe"
Warning: Nonexistent include directory "[root]/builds/linux_ompi_gfort/lib/tomawac"
Warning: Nonexistent include directory "[root]/builds/linux_ompi_gfort/lib/telemac2d"
t2dfort.f:97.9:

USE BIEF
1
Fatal Error: Can't open module file 'bief.mod' for reading at (1): No such file or directory

The libraries are all in place but not being picked up by the re-compilation.
I include my cfg file.
Can any one provide any suggestions as an answer to this?

Thanks
JOhn

Hello John,

Did you by any chance rename the build directory AFTER you have compiled the sources? For example you compiled your configuration as "linux_ompi_intel" and then changed the name of the configuration name to say "linux_ompi_gfort" and renamed the directory to "linux_ompi_gfort" without recompiling?

Regards,
Costas

Hi Costas
Thanks for your reply.
No, the build and its name (linux_ompi_gfort) were created by the installation compile. I only use gfortran as I usually find installation simpler and more reliable (however this time I seem to have failed!!).
I don't understand the problem because the install compile runs perfectly.

John

Yes, gfortran is absolutely fine, I use it for windows. I mentioned the "intel" suffix just to point the change in the path, not a change in the compiler. What has happened to me one time is that I wanted to rename a configuration from say "Corei7" to "Intel" and was lazy to recompile. I just renamed the configuration in systel.cfg and the directory name in builds and ended up with a problem similar to yours...

Costas

Just thinking out loud...

Which fortran version are you using? I have encountered problems when I tried gfortran 4.9.x and had to revert back to 4.8.x. It had something to do with a new module (.mod) format from 4.9.x on.

Costas

I'm using gcc/4.8.3 with openmpi/1.6.5.

Curiously - I have just found a work round I think.

If I compile the code with the scalar gfort this is successful and then run the simulation with the parallel version - it works (haven't seen the results yet - but the code is running!!)

I wonder if there is an issue with python passing a variable to stdin (<root>). I have encountered a similar problem on this cluster which was overcome by changing the combination of gcc and openmpi.

Which combination do you use?


John

I have a windows installation. I use gcc/4.8.3 with msmpi/4.2.4400.0. That is as far as I go I am afraid.

Costas

Costas
Thanks for your responses.
John

Hi You could try replacing the line:By
In the linux_ompi_gfort configuration.
Recompile with the option --clean

The problem might be that the gfortran used by mpif90 is not the same as the used previously.
So it does not recognize the module as a proper module.

Hope it helps

Yugi
Thanks for the suggestion.
It creates an fortran error on the installation compile at:

In file /gpfs/env/e088/opentelemac_mascaret/v6p3r2/sources/utils/bief/bief_def.
f:564

TYPE(COMPO),DIMENSION(:),ALLOCATABLE::COMPO
1
Error: Attribute at (1) is not allowed in a TYPE definition
In file /gpfs/env/e088/opentelemac_mascaret/v6p3r2/sources/utils/bief/bief_def.
f:565

TYPE(COMPO),DIMENSION(:),ALLOCATABLE::HAP
1
Error: Attribute at (1) is not allowed in a TYPE definition
... The following command failed for the reason above
/gpfs/grace/openmpi-1.6.5/bin/mpif90 -c -O3 -fconvert=big-endian -frecord-marker
=4 -ffixed-line-length-132 -I /gpfs/env/e088/opentelemac_mascaret/v6p3r2/builds
/linux_ompi_gfort/lib/utils/special -I /gpfs/env/e088/opentelemac_mascaret/v6p3
r2/builds/linux_ompi_gfort/lib/utils/parallel -I /gpfs/env/e088/opentelemac_mas
caret/v6p3r2/builds/linux_ompi_gfort/lib/utils/damocles -I /gpfs/env/e088/opent
elemac_mascaret/v6p3r2/builds/linux_ompi_gfort/lib/utils/bief -I /gpfs/env/e088
/opentelemac_mascaret/v6p3r2/builds/linux_ompi_gfort/lib/sisyphe /gpfs/env/e08
8/opentelemac_mascaret/v6p3r2/sources/utils/bief/bief_def.f
[e088@cn232 ~/opentelemac_mascaret/v6p3r2/configs]$

Thanks
JOhn