The fitsthumb replacement was cherry-picked in DM-4323, so we've already lost that dependency.

Agreed with Paul that it's safe to drop psycopg2.

Pyfits is used by some of the scripts in bin.src. I've not investigated whether we need them, but this dependency costs us nothing – it's easily satisfied on the LSST side.

Neither suprime_data nor testdata_subaru exist on LSST, and I don't think they're likely to in the immediate future.

Jim Bosch: I've rebased your commits against the current master and added my own to drop psycopg2, suprime_data and testdata_subaru. Are you happy with that?

Dropping all of those is fine, though I'll note that there's still a lot of PostgreSQL-related code in obs_subaru, and it might make sense to remove those along with the EUPS dependency on psycopg2 - but I'm happy to defer to you and Paul Price on that question.

I agree on removing the PostgreSQL-related code. John Swinbank, let me know if you want me to do it.

If you have time to do it now, go for it. Otherwise, it's on my todo list for a dull moment.

Hey Lauren – do you have time to look at this please? I think all the changes are straightforward. Almost all of the churn is caused by stripping out psycopg2 as discussed above (the HSC folks will re-add this on their own future fork).

Note that obs_subaru is not currently used when building lsst_apps or lsst_distrib, so it doesn't get tested by a standard Jenkins run. However, Jenkins is capable of building it directly: see e.g. build #6685.

Moving this out of review until I have chance to drop in some working test cases. Should be back in a few hours.

Following Tim Jenness's comment, I reinstated the testdata_subaru package and provided some minimal tests to demonstrate the basic operation of obs_subaru. These make no claim to be comprehensive, but will ultimately be supplemented by ci_hsc (DM-3663, DM-4251).

As well as the changes on tickets/DM-3518 in obs_subaru, please also sanity-check the new testdata_subaru package.

Note that this currently will not cleanly run through Jenkins thanks to DM-4733, but you should be able to clone and demonstrate that the tests pass on your own system. I won't merge to master until that issue has been resolved.

I put some comments on the pull request. I still have to pull down the branch to confirm that it compiles, but I'll wait to hear your response to the comments (nothing major, but a few items would require re-checking).

Thanks for the comments – I've responded to them on the PR.

Following discussion on the PR, I've added two tiny additional commits to this ticket:

• https://github.com/lsst/obs_subaru/commit/2067d129cf45eebd5555425d094f66f4202a64e1
• https://github.com/lsst/obs_subaru/commit/5d5d421ee24f49149bc343201cf72189c40450a6

Could you please also consider them in the review?

All looks good. Feel free to merge to master (when the Jenkins integration is taken care of).

Added a further commit which addresses the pyfits API change per DM-5007. It makes sense to include that on this ticket since:

• Without the fix, the new test cases here won't run;
• The test cases here demonstrate that the fix actually works!