popm: exponential backoff overflow

[AL-CT]

Confirmation

• I have checked for similar issues.

Describe the bug

PR #612 added logic for exponential backoff to popm. However, the method used (byte-shifting), will eventually force delay to overflow, leading to delay being set to a negative number (preventing Tick() from ever continuing) or even 0:

delay := baseDelay * (1 << attempt)
if delay > maxDelay {
	delay = maxDelay
}

This is exacerbated by the jitter code, which will panic once the argument for Int64N() becomes too high:

jitter := int64(delay / 10)
delay += time.Duration(mathrand.Int64N(jitter*2) - jitter)

Expected behaviour

popm should retry indefinitely without any issues

Environment

Version:
Operating System:
Architecture:

Additional Information

No response

[marcopeereboom]

Fucking exponential holdoffs are always shit. Kill it with fire, just do something random instead.