[petsc-dev] wrong makefile variables?

Barry Smith bsmith at mcs.anl.gov
Sat Oct 16 17:18:17 CDT 2010

On Oct 16, 2010, at 3:54 PM, Lisandro Dalcin wrote:

> On 16 October 2010 23:13, Jed Brown <jed at 59a2.org> wrote:
>> On Sat, Oct 16, 2010 at 22:02, Barry Smith <bsmith at mcs.anl.gov> wrote:
>>> Ok, config/PETSc/Configure.py will automatically put PETSC_LIB_BASIC =
>>> -lpetsc into $PETC_ARCH/conf/petscvariables when run with single library.
>> So you're not going to have PETSC_LIB_BASIC = -lpetscts ... -lpetscsys?
>>  This is useful because the list of library names is not constant and
>> libpetsc changed to libpetscsys in 3.1.  Lisandro (and everyone else that
>> needs to find PETSc from configure tests) could write special code to check
>> and use whatever is necessary, but why force that? In particular, putting
>> them in PETSC_LIB_BASIC means that old versions of the configure scripts
>> will still work, and new versions won't have to have a ton of branches to
>> work with old versions of PETSc.
>> Jed
> By default, PETSC_LIB_BASIC hardwires the full list, it is defined in
> conf/variables. But it is wrong for single-lib builds, so Barry just
> need to re-define it in $PETSC_ARCH/conf/petscvariables to override
> the default value, only in case of single-libs builds.
   Correct. Everything is fine I think.


  Eventually it is likely it will be generated automatically for all cases, but one slow step at a time (since each step results in a yell :-)

> -- 
> Lisandro Dalcin
> ---------------
> Predio CONICET-Santa Fe
> Colectora RN 168 Km 472, Paraje El Pozo
> Tel: +54-342-4511594 (ext 1011)
> Tel/Fax: +54-342-4511169

More information about the petsc-dev mailing list