I've updated the obs_base branch (and the tests in ci_hsc_gen3) to always use the filters from the data IDs when they are present. I suspect this is not enough to unblock Ian Sullivan, because I don't think the data IDs in the PipelineTasks are getting the information they need yet, but it may be worth a try.

Thanks Jim Bosch! I'll test it. We have decided to simply reprocess the data from raws instead of calexps, so as long as we get through that we won't actually be blocked on this. But, since other people may wish to use the data it is still good to get this fixed.

I can confirm that that change to obs_base works. It throws a number of warnings, which seems reasonable, but then it proceeds to make warps. Thank you for the quick fix!

Thanks, Ian Sullivan. That's not what I was expecting, but I realize now that it was my expectations that were wrong, because I misread your last error report. Everything is now consistent with there being no problem with the data IDs being passed to PipelineTask, and hence this is ready to go back into review.

Thanks for fixing this. Per your PM to me, I'm going to merge now so it gets in the weekly. We can sort out the question about `warnings.warn` after the Community discussion is over. I'm still not happy about the lack of local tests, but I think I understand why they are difficult, and I hope that other ticket makes them easier to implement.