691630 - Untangle backoff handling code

Status: RESOLVED FIXED

(Firefox :: Sync, defect)

Description

[Richard Newman [:rnewman]]

---
On a further note, in a follow-up, I would like to untangle all of the backoff
code. It's spread over various functions and methods. I'm not even confident
that by fixing those two bugs backoff will work. But the tests I am writing
*right now* will tell us more.
---

Comment 1

[Philipp von Weitershausen [:philikon]]

Things that come to mind:

* Move Status.backoffInterval to SyncScheduler
* Revise or at least document MINIMUM_BACKOFF_INTERVAL and MAXIMUM_BACKOFF_INTERVAL
* Revise or at least document SyncScheduler.handleSyncError, Utils.calculateBackoff and MAX_ERROR_COUNT_BEFORE_BACKOFF

I'm not sure I actually understand some of the intentions of this code, so perhaps we should first sit down with mconnor and document inputs and outputs and then just write the code to be like that. I wouldn't at all be surprised if there were mismatches.

Comment 2

[Richard Newman [:rnewman]]

We never set Status.enforceBackoff to false.

We don't document a server-side way to turn backoff back off (heh).

Status is the wrong place for this.

/me goes shopping in the clown supply store

Comment 3

[Richard Newman [:rnewman]]

We did a lot of this with the policy rework.