This repo is not forked from upstream

[mwoehlke-kitware]

This repository is not forked off of what appears to be the (current) canonical upstream, https://github.com/libbot2/libbot, and the history is out of sync.

I propose that someone with sufficient permissions:

• Push a copy of the current master to some new branch name (to prevent garbage collection from pruning the old history until forks can be updates).
• Reset master to mwoehlke-kitware/libbot@e7ff4150c275, which has the same content as current master (@a889c5f10f31), but meshes with the upstream history.
• Ask github to attach this repo as a fork of https://github.com/libbot2/libbot so that PR's can be made between us and upstream.

[jamiesnape]

Pinging @david-german-tri.

[david-german-tri]

Hmm. Alternatively, could we just fork the canonical upstream into a new repo, push our patches to it, cut Drake & co over, and then delete this repo once we're really sure it has no users? (Or perhaps just stick a loud warning in the README and forget it exists?) That seems like it has a few advantages:

• No rewriting public history
• Similar amounts of work, since you already have the stack of commits in your fork
• Canonicalize on the new libbot2 name
• Don't have to talk to GitHub support (can they even make the "fork" link post facto?)

[jamiesnape]

I like that plan. Quite convenient the upstream repo is not called libbot.

[mwoehlke-kitware]

Alternatively, could we just fork the canonical upstream into a new repo

This just pushes the problem downstream, by detaching us from the (non-trivial) network of forks of this repo. We'd force everyone else to do the same, and we'd lose history (all issues and PR's in all downstream repos) in the process.

No rewriting public history

Either way, we need to point master at the aforementioned SHA (which is not in upstream). Anyway, I proposed to keep the old history, maybe even indefinitely (so old PR's will still point somewhere).

Don't have to talk to GitHub support (can they even make the "fork" link post facto?)

I believe so; at least, I found a page somewhere (SO?) that suggests they can. Possibly we could even ask first, as I think adding the fork relationship doesn't actually require shared history.

Canonicalize on the new libbot2 name

If that's desirable, it's trivial and less disruptive / destructive to just rename this repo.

[david-german-tri]

OK, those are good points. I'll contact GitHub support.

[david-german-tri]

They say:

Cross-network reparenting isn’t possible at this time. If the two repository existed in the same network, we may have been able to assist with this, but looking at them on our side, this wouldn't be possible.

[david-german-tri]

I just talked with @gizatt about this, and he says that if we switch to a new repo, it won't cause significant pain for his forks.

@patmarion, @wxmerkt, would it cause problems for you if Drake switched to a new RobotLocomotion/libbot2 repository, which is a fork of upstream https://github.com/libbot2/libbot? master in the new repo would have the exact same source code as this repo does right now, but a history that was consistent with upstream. This repo could continue to exist indefinitely.

[patmarion]

fine with me, thanks for working on it!

[mwoehlke-kitware]

Updated SHA (including #22) is 83d56d6.

[david-german-tri]

We probably won't hear from @wxmerkt until tomorrow AM due to time zones, so let's table until then.

[wxmerkt]

@david-german-tri No problems here, thanks for checking :)

[david-german-tri]

Fork: https://github.com/RobotLocomotion/libbot2

[david-german-tri]

@mwoehlke-kitware, please go ahead and PR your stack of commits, then cut Drake over!

[mwoehlke-kitware]

New (properly forked) repo is https://github.com/RobotLocomotion/libbot2, used by Drake as of RobotLocomotion/drake#4680