Repository navigation
Group updates to have consistent versions #15923
colin-janke
started this conversation in
Suggest an Idea
Replies: 1 comment 10 replies
|
One challenge is that if you intentionally filter/withhold updates because some package doesn't support it yet, you could get stuck forever and be unaware of it. i.e. you may not even know that 1.2 exists and that you're missing it, if Renovate only proposes 1.1. In fact it could be up to 1.8 and you're a year behind and you still don't notice, because you don't get any PR. Due to this today we err on the side of visibility (preferring to raise a PR and you can see there's incompatible versions) than withholding, but ideally we'd find a way to achieve both. |
10 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Something of a speculative idea:
We have a number of dependencies which share a common root dependency. Eg:
We use a combination of package rules (groupings), the enforcer plugin and mvn dependency exclusions to try and control what version of different projects we use and to try and keep things consistent. However, this is getting more complex, can give surprising errors when the exclusions cause us to use incompatible code and can make the grouped MR take a long time to get merged as we get renovate-generated patch releases of the low-level library while trying to assemble things.
For example, this can result in a case where the latest version of C is 1.2 but the latest versions of A and B depend on C v1.1. We then get a renovate MR which tries to update to the latest versions of A, B and C, which then fails the enforcer plugin checks.
What I think would be nice would be some extension of the grouping rules which would allow renovate to generate compatible sets of versions. So if the latest version of C is 1.2 but the latest release of A and B still depend on C v 1.1, we would get a renovate MR bringing us up to the latest versions of A and B with C at v1.1. This could then automerge and would keep us generally closer to latest on all of these dependencies as we wouldn't have so many failed MRs.
What I think would be additionally useful would be some kind of indication of what is being blocked. Maybe in the form of an additional MR with the latest version of C so we can see that we need to get a version of A and B to depend on it.
All reactions