[petsc-dev] reorganizing src/sys

[Barry Smith]

  Dmitry,

     If it has a possible future then it doesn't have to be removed (though it does need to be renamed :-)). I have no problem keeping it so long as there is a comment in the code about its purpose and current state.

     Barry

On Dec 11, 2012, at 7:43 AM, Dmitry Karpeev <karpeev at mcs.anl.gov> wrote:

> sys/shell is mine. It's a badly named attempt at simple language interoperability. Since it's not really being used, as far as I know, it can go, although the corresponding parts of petsc4py would have to be removed, too. I can clean it out this week along with some other long dead experimental code.
> Dmitry.
> 
> On Dec 11, 2012 6:36 AM, "Barry Smith" <bsmith at mcs.anl.gov> wrote:
> 
>    Removed. Thanks, Barry
> 
> On Dec 10, 2012, at 11:00 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:
> 
> > comm/pami was mine, but it can be removed because that code only works on COMM_WORLD anyway (lame interfaces in the PAMI/MPI implementation)
> >
> > On Dec 9, 2012 8:36 PM, "Barry Smith" <bsmith at mcs.anl.gov> wrote:
> >
> >    The src/sys directory is getting rather clogged with an odd collection of stuff in seemly randomly and bizarrely named directories with no hierarchy (yes Jed I know we don't need no stinky hierarchy, we only need tags but that won't help hz50241027 who has "No tags file").
> >
> > Barrys-MacBook-Pro:sys barrysmith$ pwd
> > /Users/barrysmith/Src/petsc-dev/src/sys
> > Barrys-MacBook-Pro:sys barrysmith$ ls
> > adic          comm          error         f90-src       ftn-custom    makefile.html mpiuni        python        shell         totalview     viewer
> > ams           dll           examples      fileio        index.html    matlabengine  objects       random        threadcomm    utils         yaml
> > bag           draw          f90-mod       fsrc          makefile      memory        plog          sf            time          verbose
> >
> >
> > I'd like to organize it with more structure, first putting all the "system" stuff that does NOT know about PetscObject (only depends on petscerror handling, info, and malloc, note does not depend on logging) together (truesys), all the stuff that defines the PetscObject model and logging together (petscobject), and all the stuff that builds higher level infrastructure on top of PetscObjects (topobjects). (names subject to improvement).
> >
> > The truesys is mostly wrappers for non-portable system routines, things like PetscSortInt() etc.
> >
> > topobjects includes viewer, random, draw, sf
> >
> > There is some weird stuff like shell?, comm/pami? other?  What are they? Who owns them? Should they be removed?
> >
> >   As with most changes in PETSc I'd like to do this quickly but evolutionarily, moving things around a bit at a time to get to the new form.
> >
> > Thoughts?
> >
> >    Barry
> >
> >
> >
>