On Thu, 13 Oct 2016, Antonio Trande wrote:

> I guess, we can consider all tests successfully passed.


I'm sure there will be more issues to deal with [when folks start
using these packages]

For one - with our default build - one uses portable makefile [with
PETSC_DIR/PETSC_ARCH] for development. Will have to figure out - what
the approriate use model would be [with fedora petsc packages..]
Perhaps pkg-config?

Also - I'm not sure about the utility of having a 'seq' build
packaged.. [from our point of view - its useful only when MPI is not
available - or too combersome to use].

Infact it can cause problems - as the mechanism to implement this is
to use a dummy/sequential/incomplete MPI (aka mpiuni) in the
package. And this can conflict with mixed with other packages that use
MPI or have their own dummy/sequential MPI.. [for ex - seq-mumps]

Using --with-mpiuni-fortran-binding=0 can mitigate this issue with
slightly crippled fortran functionality..

[Hm - Perhaps we should fix this - and namespace even the
mpiuni/mpif.h using preprocessing - and remove
--with-mpiuni-fortran-binding option..]


Reply via email to