My app has a typical OAuth flow in terms of, 401 c...

# squarelibraries

[Colton Idle]

My app has a typical OAuth flow in terms of, 401 comes into the app, the app (via an interceptor) detects this 401, and tries to exchange the refresh token for a new set of auth and refresh tokens. This works great with okhttp interceptors... but I've noticed in cases where the user comes back to the app and 2 network calls are made at the same, I will get two 401s, and therefore (unfortunately) my interceptor will trigger twice. Anyone have any guidance/experience around this. My initial reaction is to write some sort of state to my user isAuthRefreshInProgress and short-circuit the interceptor if it's already in progress, but then I'm not sure how I'd wait for the auth refresh to be complete, so I can make that second call again. Here's a gist of my current interceptor.

[ephemient]

if a second call to refresh comes in while the first is still processing, wouldn't you prefer to wait for it to finish and use its result?

[Paul Woitaschek]

Our auth code is ancient but we have synchronized on the authenticators body

[ephemient]

perhaps with something like this. untested but it should convey the idea. start=LAZY because updateAndGet can be restarted on conflict, GlobalScope because it doesn't belong to the caller's scope if it's shared

Untitled.kt

[Colton Idle]

@Paul Woitaschek yeah I thought the synchronized I had in my gist would be enough, but I guess right now I'm likely to do a "short circuit" check in the very first line of the synchronized block so that I don't continue if I already got a new auth + refresh token set.

@ephemient I will have to play around with your code a bit as a lot of it is admittedly going over my head.

I guess in a general sense it doesn't seem like interceptors really have anything built in to help with this situation?

[ephemient]

interceptors are pretty much independent of each other and AFAIK there's nothing built-in to provide synchronization

[Colton Idle]

Cool cool. Okay. I think just checking this new state property, and short circuiting should be enough then.

I do have to try to play around with my use of `synchronized`as I don't fully understand it, but I hope that it actually makes my parallel interceptors work in serial.