>
> have to re-initialize it and kick git occasionally to get it to behave with
> accord. We'd probably need to do the same w/jamm but wrapping any submodule
> operations with retries and reinit shouldn't be too hard.
Having both jamm and accord being the same boat, and using the same .build/git/
scripts, would help address and improve this problem.
Any such erred states should be made as simple as possible to address, and
guarded against at commit time.
Can you elaborate a bit more with examples ? I haven't been aware of any such
issues happening in ci-cassandra.a.o's trunk builds, which has been running for
quite a while now with accord…
And are we missing a README.md in the .build/git/ directory, if there's tribal
knowledge here not well known …?
> The bigger issue I see is that having changes be in a separate unmaintained
> repo makes coordinating changes w/C* significantly harder.
That's the whole benefit of the git submodule, for us it is in-tree 🤷
Personally I'm not a huge fan of git submodules (see the original thread where
we discussed introducing it), but a lot of work has gone in to making it simple
and automated for us, and now that the project has it I'd rather see we align
and consolidate practices. And if it's still considered too painful maybe we
should revisit whether it's still what we want to be using it for accord.
