[petsc-dev] MPI I/O tests failing in master after merging my branch

[Jed Brown]

Lisandro Dalcin <dalcinl at gmail.com> writes:

> On 22 March 2018 at 15:31, Matthew Knepley <knepley at gmail.com> wrote:
>> On Thu, Mar 22, 2018 at 8:28 AM, Jed Brown <jed at jedbrown.org> wrote:
>>>
>>
>> What I imagined was just mapping each Petsc typedef (do we list them all
>> anywhere),
>> to an MPI type, creating those that are missing (as we do for MPIU_INT). Is
>> there more
>> to it than that?
>>
>
> We have that (I mean, each PETSc typedef). What we are somehow missing
> is the mapping plain C 'int' -> MPI_INT. That's important for binary
> I/O with non-PETSc formats (e.g. GMSH).

There are a few public interfaces that would need to change to accept
MPI_Dataype, then update a bunch of code that uses it.  Might only take
a couple hours.  My guess is that PetscBag will be the most subtle (and
it needs to remain compatible).  One structural usage difference would
be that PetscDataType can be used in a switch statement.