Conversation
Eventually I'd like to see kubeclient get up to Faraday v2. Getting the googleauth dependency into v1 is a necessary step.
|
FYI, test_informer.rb is known flaky, we have a tracking issue for that, but for now its failures are not merge blockers. UPDATE: flakes'd been fixed since then, closed & reopened to re-run CI with current master => all green now 🍏 |
|
Thanks! Sorry, forgot this for no good reason :-( LGTM and reading their changelog, sounds "probably non-breaking". But I really don't know almost anything about google auth, and that's some ~30 releases they've had (most before adopting semver)... @lucasmazza @timothysmith0609 @jeremywadsack can any of you comment on the safety of this upgrade, and/or test it? |
|
@cben The latest version of @dhstewart If you just need Faraday v2, can you use the spec |
We've been using |
|
I think I'm gonna merge this and see if anybody complains. Could relax the constaint to 1.1+ later if needed. #606 is going towards some "how exactly do these libraries cache/expire tokens" questions, and reliably answering that for a wide range of versions is harder; if we can get away with only supporting recent ones that's simpler 😅 |
Why
I did some research on this issue #564
It looks like to get openid connect up to date we'll need to get to Faraday v2, but our current version of googleauth constrains below that.
So it seems that getting the googleauth dependency into v1 is a necessary first step to addressing the issue.
What
If a more incremental approach to dependency updates are desired, I'm happy to approach this differently.
Let me know what you think 👍