Closed JakobAsslaender closed 1 month ago
It seems that the CI failures are unrelated to this compat update, they are the same failures as the master branch. I tested the subpackage locally and it passed all tests.
@nHackel: could you comment on this? I recall that we are currently in a bad state and it requires merging the gpu branch?
The GPU branch is ready to be merged and increases the compat for LinearOperatorCollection to 2.0
I think the bad state was partially that in the current master we are not yet using the correct keyword for the normal operator. I don't recall all the details, but things should be fixed in the gpu branch
ok, merged. I would then directly release 0.9, right?
Hmm it's probably for the best. MRIOperators definitely had breaking changes and increased to 0.3.
From an interface perspective I think MRIReco only had bug fixes and the new arrayType keyword, but we increased the requirements for MRIOperators to the new version...which I think would count as a breaking change since we increased our lower compat bound
MRIBase/Operators is already triggered...
@tknopp: can you merge and release this change? I still don't understand, how to release these sub-packages without breaking everything.
Another quick question: the compat helper doesn't work with sub-projects, does it? It would be nice to get that working to avoid the manual labor of updating packages.