Tim Jenness – since Josh is away, would you have time to look at this please? It's just a single commit on tickets/DM-4733 in lsst_build.

This works for me, but there are a few things to note:

• It's not clear why this is necessary – is there a downside to simply globally enabling LFS for all of the repositories?
• I don't understand the gymnastics with GIT_ and SSH_ASKPASS. These were previously being removed from the environment at clone time but not on subsequent fetches, which seems odd. I've attempted to preserve the logic, but since I didn't follow the intent I'm not confident I've properly captured it.
• The code in this package doesn't seem to follow the coding standards. I've attempted to follow the local conventions rather than the LSST rules.

Sorry but I don't know enough about that bit of code to review it. I think we will have to wait for Joshua Hoblitt.

Ok, thanks. We should try to avoid single points of failure, though – is there anybody else who could tackle this (maybe Frossie Economou has an idea)?

John Swinbank I believe your branch should work but I want to do some manual testing before merging it. I vaguely recall having seen issues with lfs pull on existing clones not always replacing checked out files from non-lfs checkouts but I suspect it's long since been fixed. An alternative is to delete the repo and re-reclone it.

In face-to-face discussion, Josh suggests reworking this so that it simply deletes and re-clones rather than attempting to enable LFS in an already checked-out repository. Should be straightforward to do that.

This is a blocker for transitioning at least two sims packages (sims_data and sims_sed_library) that are currently starting to cause timeouts in conda installations.
Is there a resolution coming for this issue?

I think this work is now obsolete, given the lack of activity over several years. Lynne Jones, I'm sorry for the lack of an earlier response — is it still causing problems for you? If so, give me a shout and we'll reopen this (or file another ticket).