changes to PETSc-dev bmake system and library locations

[Barry Smith]

On Mon, 11 Jun 2007, Lisandro Dalcin wrote:

> I think 'conf/variables' needs a bit more of work. I am not sure about
> what to do, the problem is here:
> 
> PETSC_LIB_DIR	    = ${PETSC_DIR}/${PETSC_ARCH}/lib
> PETSC_INCLUDE	    = -I${PETSC_DIR}
> -I${PETSC_DIR}/${PETSC_ARCH}/include -I${PETSC_DIR}/include \
>                        ${PACKAGES_INCLUDES} ${TAU_DEFS}
> ${TAU_INCLUDE} ${PETSC_BLASLAPACK_FLAGS}
> 
> 
> As you can see, there are references to $PESC_ARCH.  So there are two
> potential problems using a 'install' version:
> 
> 1.-  If $PETSC_ARCH is not defined, some paths are of the form
> '/usr/local/petsc//lib'. Not sure is the double '//' could be
> problematic.

  My working assumption was that Unix would always handle the // properly;
In 21 years of using Unix I've never seen this to be a problem, but then
I always learn new things.

> 
> 2.- If $PETSC_ARCH is defined somewhere (user makefiles or even
> environment), bad paths will be generated,

  This is true, if using PETSc makefiles and an "installed" PETSc,
things will be messed up. My assumption was that people who used
"installed" PETSc would not know or care about PETSC_ARCH

>  and ${PETSC_LIB_DIR} is
> definitely broken.

   Not sure what you mean here? In my limited tests setting
setenv PETSC_LIB_DIR /usr/local/lib for example,
worked fine.

  Barry

> 
> Is all this OK? Or my hg pull + hg update failed?
> 
> 
> On 6/11/07, Lisandro Dalcin <dalcinl at gmail.com> wrote:
> > On 6/11/07, Satish Balay <balay at mcs.anl.gov> wrote:
> > > I should add that one motivation to support gnu/autoconf? 'make
> > > insall' model as closly as possible - is to make it easy for
> > > 'packagers' [deb,rpm etc..] to package PETSc
> > 
> > OK, this make sense... I'll try this when I have some free time, I've
> > been playing a bit generating rpm's for Fedora.
> > 
> > 
> > [atleast the default
> > > config - whatever that means] and make it available for their
> > > respective distributions. Currently we only have only debian/freebsd
> > > packages - and nothing more.
> > >
> > > And I'm hoping that these packagers have their own models for
> > > installing multiple configs [via different prefix], so potentially
> > > multiple variants of PETSc config can be packaged [if they chose to do
> > > it].
> > >
> > > However I think the requirement of 'PETSC_DIR' might still be a minor
> > > hurdle for this mode. [the alternative would be providing compiler
> > > wrappers but thats a different discussion]
> > >
> > > Satish
> > >
> > >
> > 
> > 
> > --
> > Lisandro Dalcín
> > ---------------
> > Centro Internacional de Métodos Computacionales en Ingeniería (CIMEC)
> > Instituto de Desarrollo Tecnológico para la Industria Química (INTEC)
> > Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET)
> > PTLC - Güemes 3450, (3000) Santa Fe, Argentina
> > Tel/Fax: +54-(0)342-451.1594
> > 
> 
> 
>