You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In cli/cli#9370 (comment), several people, including myself, have issues with the gh cli that is not working properly with login when the auth server responds with "slow_down" error. According to the RFC8628 section 3.5, this means the client MUST increase the interval of polling, and it should continue to poll and not crash out. I had a look at the code (I'm don't consider myself a Go developer) and I didn't see any handling of that (e.g. https://github.com/cli/oauth/blob/main/device/device_flow.go#L190). Presumably the change needs to not only check that "slow_down" being a possible "continue" case but also increase the interval or use the new interval provided.
Hey @jerr0328, you're right that this module does not handle slow_down. It's definitely something we can implement, though I'm not sure it'll really deal with the underlying issue there, as I mentioned here: cli/cli#9370 (comment).
We can definitely do it and see what the outcome is though!
Given I am logging in against a server that responds with slow_down when requesting the access token When I go through an oauth flow targeting that server Then the requests for the acess_token increase by 5 seconds per slow_down response received.
Activity
williammartin commentedon Dec 11, 2024
Hey @jerr0328, you're right that this module does not handle
slow_down. It's definitely something we can implement, though I'm not sure it'll really deal with the underlying issue there, as I mentioned here: cli/cli#9370 (comment).We can definitely do it and see what the outcome is though!
williammartin commentedon Dec 11, 2024
Acceptance
Given I am logging in against a server that responds with
slow_downwhen requesting the access tokenWhen I go through an oauth flow targeting that server
Then the requests for the acess_token increase by 5 seconds per
slow_downresponse received.