Jim Bosch added a comment - 24/Oct/18 7:15 AM Ian Sullivan , could you take a look at these obs_* package dataset type changes?  I think you were the last person to attempt something like this in obs_lsstCam, and you're certainly in the best position to check dcrCoadd things. On that note, there are also some dcrCoadd definitions in obs_decam that look suspicious to me, but I'm not sure what to change them to (unless it's to delete them so the obs_base ones are used).  I have not made those changes on this ticket but would be happy to if you can point me in the right direction. PRs are: https://github.com/lsst/obs_base/pull/118 https://github.com/lsst/obs_decam/pull/95 https://github.com/lsst/obs_lsstCam/pull/34 https://github.com/lsst/obs_lsstSim/pull/78 https://github.com/lsst/obs_sdss/pull/57 https://github.com/lsst/obs_subaru/pull/161 https://github.com/lsst/obs_cfht/pull/51 I am under no illusions that this fixes all problems with dataset type definitions, but I didn't want to change anything that could be backwards incompatible (i.e. changing a template definition) unless I was certain it was already broken (aside from obs_lsstCam, where I'm explicitly trying to get things in shape for better future compatibility). I'd recommend looking at these commit-by-commit.

Ian Sullivan added a comment - 24/Oct/18 12:01 PM I've tested the changes for `obs_lsstSim` along with the additional deletions I suggested on the PR, and everything appears to work. Some metadata had before been persisted in the wrong place, and is now persisted correctly.