[petsc-dev] prefix installs and lib/petsc-conf

[Barry Smith]

  How about?

    share/petsc/matlab,python,julia/ non-generated stuff 
    lib/petsc/matlab,python,julia/ generated stuff for example Matlab .mex files

    lib/petsc/conf    

    include/petsc/finclude,private,mpiuni  (note some generated stuff goes in here; could put it in lib/petsc/include but that seems silly to me)

    bin/petsc/mpiexec.uni  when using uni
    bin/petsc/uninstall.py

    So moving all the petsc_xx directories to petsc/xx  


   What do you all think?

   Barry



> On Jan 14, 2015, at 7:23 AM, Lisandro Dalcin <dalcinl at gmail.com> wrote:
> 
> Barry/Jed, I would like to propose an slightly alternate layout for
> prefix installs. Let's create a $PREFIX/lib/petsc directory, inside it
> we further create a bunch of directories as required:
> 
> bin -> Put here some or all of the stuff in $PETSC_DIR/bin,

  
> 
> conf/ -> put variables,rules,test ... I would even go further and call
> the directory "make", and put there anything related to make as a
> build system.
> 
> log/ -> put here make.log, configure.log, test.log, etc. (and we
> should not install this crap)
> 
> matlab/ -> we can put here the stuff in share/petsc/matlab

 
> 
> julia/ -> we can put here the Julia stuff
> 
> python/ -> we can put here library Python code like the binary I/O stuff

  
> 
> What do you think? The idea is not not pollute the lib/ location with
> many petsc-<something>

 
> and at the same place provide a place to inject
> reusable library components for julia, matlab, python, etc.
> 
> Perhaps you want to reserve 'lib/' for platform-dependent (e.g.
> binary) files. In such a case, share/petsc/matlab is fine, we could
> also have share/petsc/julia, share/petsc/python, etc.

  

> But still, we
> could do lib/petsc/conf instead of lib/petsc-conf

   

> 
> What do you think in general? And please tell me: What do you think we
> should install in 'lib'? Should 'lib' be reserved for
> platform-dependent stuff? In such a case we should put there
> petscconf.h, Fortran modules, etc.

  


> 
> 
> 
> -- 
> Lisandro Dalcin
> ============
> Research Scientist
> Computer, Electrical and Mathematical Sciences & Engineering (CEMSE)
> Numerical Porous Media Center (NumPor)
> King Abdullah University of Science and Technology (KAUST)
> http://numpor.kaust.edu.sa/
> 
> 4700 King Abdullah University of Science and Technology
> al-Khawarizmi Bldg (Bldg 1), Office # 4332
> Thuwal 23955-6900, Kingdom of Saudi Arabia
> http://www.kaust.edu.sa
> 
> Office Phone: +966 12 808-0459