Bug#1102465: petsc: PETSc was configured with one MPICH/Open MPI mpi.h version but now appears to be compiling using an older version

[Andreas Tille]

Hi Adrian,

Am Fri, Oct 23, 2026 at 10:43:46PM +0300 schrieb Adrian Bunk:
> architecture-is-64-bit is not a solution for a normal bug.

ACK, fully agreed. We should fix the actual problem rather than use
architecture restrictions as a workaround.

At the same time, I wonder whether this is also an opportunity to
discuss a broader question: do we still need to provide binary packages
for all scientific software on architectures such as i386, where there
seems to be little practical demand from users?

I would not suggest introducing architecture restrictions as a quick fix
for individual bugs. But perhaps we should be more willing to reconsider
the architectures we support when maintaining a package on a particular
architecture creates a recurring burden without a clear benefit to
users.

For me, the goal should be to spend our limited maintenance resources
where they provide the greatest value, rather than maintaining packages
on every architecture simply because we can build them there.

Kind regards,
   Andreas

-- 
https://fam-tille.de