Bug#1080198: h5py: autopkgtest regressions with mpich as default provider on 32 bit architectures

[Chris Hofstaedtler]

Hi Francesco,

* Francesco Ballarin <francesco.ballarin at unicatt.it> [241109 13:57]:
> please give us a grace period of further 30 days before pinging us again because
> the current migration issue is due to openmpi 5 being broken in unstable on all
> the arches on which you see the failure.
> 
> There is basically nothing we can do until that gets sorted out.

I'll note that without the ping, h5py would have been autoremoved on
Nov 13.

The autoremoval process in Debian is more or less designed in a way
where individual package maintainers can ignore neither their
dependencies, or their reverse-dependencies.

> Also note that the original report was about 32 bit arches (which use mpich), 
> while the current migration problems are on 64 bit arches. In this sense,
> it is still correct that this bug should be remained marked as done, since the
> current failed migration is unrelated to the previous ones.

Not doubting it, but it still is unfixed in testing.

Chris