I tried rooting around the documentation but didn’t find anything which seemed to cover these two issues.
We’re considering a “merge bot” à la bors and I was looking at ng as something which might fit, but:
We have multiple maintenance branches aside from the main development branch, we’d actually want to implement ng on the maintenance branches first, as it’s more important that they are not broken than the development one.
does ng handles destination branches on PRs and merges into the correct destination?
if it does, does ng use a staging branch for each destination (being able to queue PRs for e.g. 1.1 and 1.2 concurrently and independently)
similar question for trying I guess, can ng generate multiple trying branches or is only one try valid at any one time?
Bors-NG will handle multiple target branches pretty reasonably. Only PRs that are targetted at the same branch will ever be batched together, and the batches will be pushed to the correct branch. That feature was added back in May of last year.
Nice. Though I see in the PR that there’s still a single queue/staging branch. I’m not quite sure what you mean by
project maintainers only need to whitelist
Something I missed while reading the docs?
Anyway our concerns are a bit different than Marcase’s but we basically have half a dozen release branches running concurrently with the development master. Most of the release branches see pretty little activity, but it’s still activity. How hard to you believe it would be to allow maybe partitioning the staging work over multiple branches?
I mean that their CI system needs to build the staging branch for it to work. Allowing multiple partitioned queues would require it to have multiple staging branches, which would make it more complicated to set up.