I've made a quick pull request that removes gen2 and does not change the test coverage. Jenkins passes fine.

Eli Rykoff for obs_subaru gen2 removal how did you handle the gen2 tests that are just a butler.get of a calibration? Is the idea that we remove the gen2 stuff from testdata_cfht and then do a test that creates a repo, ingests the raws and calibrations and then sees that they can be retrieved?

Christopher Waters do we know anything about the CFHT-specific IsrTask? It doesn't seem to be tested and is not listed in a pipeline YAML file. Is the config/processCcd.py where all the magic happens? I assume that's a gen2-specific config that is now irrelevant. For gen3 do we have to define a pipeline that explicitly lists the CFHT IsrTask? I don't know if we test that anywhere even in gen2.

Re: the gen2 get tests for obs_subaru, I simply deleted them. I assumed that the gen2 code-paths there were covered by ci_hsc_gen3.

Note that we have similar questions for obs_decam, we filed DM-34862 to convert testdata_decam to gen3 which will allow the migration of these unit tests in DM-34863. I think we need to do the same for testdata_cfht.

One of the problems is that obs_cfht is not really usable for gen3 because no-one cares enough about it to sort out the fundamentals (see DM-21865 and DM-26681).

I think that the existing CfhtIsrTask will work in gen3 as is, as it seems to only be checking gain/read noise values.  I assume that the gain/read noise variations in the data make it important to do that to get useful variance planes, and so it's probably worth retaining.  If you make a ticket for unit test coverage of it, I'm happy to do it.

As agreed, I've ripped out all the gen2 and have not attempted to deal with porting over the gen2 calibration tests. The raw ingest tests check that gen3 is working.

Jenkins passes. No rush to review.

Thanks for doing this cleanup. See my notes: unfortunately, obs_cfht was not included in the DM-30891 work, so that block in the ups table is still necessary.