Bug#1090791: petsc: strict dependency on the mpich version that was used to build

[Drew Parsons]

On 2024-12-22 11:33, Emilio Pozuelo Monfort wrote:
> On Sat, 21 Dec 2024 23:34:42 +0100 Andreas Beckmann <anbe at debian.org> 
> wrote:
>> Control: affects -1 + src:slepc src:fenics-dolfinx src:mpich
>> 
>> On Sat, 21 Dec 2024 19:19:43 +0200 Drew Parsons <dparsons at debian.org> 
>> wrote:
>> > This is a delicate issue.  The strict version test is in the upstream
>> > headers.  I imagine they have their reasons for making it so.
>> 
>> I don't think so ...
>> 
>> Anyway, if you want to keep the strict dependency, please mirror it in 
>> the package dependencies (probably in the -dev packages) s.t. every 
>> mpich version bump triggers a mini transition, i.e. a binNMU of petsc.
> 
> I'd rather the check gets removed. In my experience, that's due to 
> overly cautious upstreams, but it's not really suitable for Debian, 
> where we track ABI breaks. But if I'm wrong and there's a good reason 
> for that, then yes, we need strict dependencies.


The strict test does seem a little strange.
I'll relax it for Debian and raise an Issue upstream for further 
discussion.

Drew