petsc-dev directory structure questions

Satish Balay balay at
Fri Dec 14 13:57:54 CST 2007

On Fri, 14 Dec 2007, Barry Smith wrote:

> > 
> > There is a difference between pythonmodules [i.e old PETSC_DIR/python]
> > and python-usercode [i.e]. And I certainly don't like
> > dumping all 3 [petscconf, & python-modules] in the same
> > location.
>   It is ALL python user code; all used for one purpose: to generate the
> petscconf, petscconf.h petscrules. Whether it is all in
> or in several python files makes NO difference. Since it is all for the same
> purpose and only that purpose it makes sense to go together and
> no sense to go apart. Note there is lots of code in config/
> that could be put into modules but we have not;

I'll argue that its not the same. python-modules are like the c
sources in PETSc/src, and, etc are user
invokable applications like ex1.c ex2.c ex3.c. Yeah they serve a
single purpose similar to the way PETSc libraries serve a single

> > > src traditional means code that gets compiled up and put in a
> > > library or executable: the old PETSC_DIR/python stuff definitely
> > > does not fit in that category.
> > 
> > Not really - src is source for anything. For eg: we have
> > src/docs/website src/docs/manual which are sources for other
> > derivative stuff [aka manual.pdf,]
> > 
> > 
> > If we stick to your argument - then we'll have to find new places for
> > this [i.e go back to PETSC_DIR/src where sources & generated docs were
> > dumped in the same location with too much confusion about website?]
>  Absolutely WRONG! The src/docs is the SOURCE code for generating
> the docs (it is in latex and is source code). Note that it is NOT
> the utilities that generate the docs, it is only the source code,
> the utilities for generating the docs are in bin

My point was 'its not just sources that get compiled to
c-libraries. We have docs sources, website sources aswell.

Just because python sources don't need to be compiled, doesn't mean
that they can't be in src.


More information about the petsc-dev mailing list