Back again... sorry for being a bit late
Herebelow are some (not many) comments.
1. MED dependency
MED >= 2.3.4 headers not found
compatible MED headers not found
The warning echoed by the configure script is not an issue. As written in a different thread, MED headers changed between 2.3.1 and 2.3.4 versions; so we handle them differently. This doesn't mean you cannot use MED < 2.3.4 (actually the minimum requirement must be 2.3.0 IIRC, even if I'm not sure this version has ever been released), what
Code_Saturne is only indeed a MED 2.3 mesh format.
Related comment on HDF5 installation, I would recommend to use
libhdf5-serial instead of the different MPI versions, except if needed. Everything will still work and you will avoid some difficulties.
2. METIS installation
We can actually change the function name 'log2' to 'ilog2' in the related head files and source codes under the directory, Lib.
Indeed, it is annoying to install due to prototypes redefinition, non-standard directories installation, ... You can avoid it by installing the
SCOTCH library (perhaps a bit less performant for many-processus calculation, but sufficient for standard ones), or use the internal partioner (based on a Space-Filling Curve algorithm, good-enough for few-processus calculation
e.g. several-core mono-processor computer).
3. MEI installation
MEI is now compulsory to be prepared for the compilation of ncs.
No. Only if you want to use "fonction definition" in the graphical interface... so strongly adviced!
sudo apt-get install bison flex
Not needed anymore in version 2.0-rc1 as we moved bison/flex use from the compilation stage to the boostrap stage. This will only be needed if you want to change the parser/lexer of MEI as you'll need to bootstrap MEI to modify the generated C sources accordingly. No "standard" user is going to do that
Note that explicitly specifying --with-bft=/usr/local is necessary for Python to find bft.
This part might only be needed because of the installation in /usr/local. This seems not to be needed if you install it in /usr... To be confirmed.
4. Other stuff...
Probably one has to re-create his study and case configurations again with 2.0-rc1...
Indeed, this is a good advice. We try to keep some compatibility between the versions, but generally not during the development stage. Let's say that, starting from the release candidate, the Fortran routines API and the XML files compatibility should be ensured; i.e. versions 2.0-rc1, 2.0-rcX (if needed), 2.0.0, 2.0.1, ... should be compatible.
Last but not least, I tried to 'apt-get install code_saturne' because I know Code_Saturne is already in Debian "unstable" depository
At the moment, the code-saturne package is only available in
Debian testing. It means it will make it for the future Debian "Squeeze" stable release, but not for Ubuntu 10.04 LTS as the feature freeze was before the entering of our package in Debian testing...
The apt-get command will be:
apt-get install code-saturne. The package currenly miss MPI support (this will hopefully be corrected before final release).
David