Work around pool bugs (drop duplicate jobs)

[jean-m-cyr]

There are two known cases where pools will send duplicate jobs
leading to rejected shares.

1 - Ethermine will occasionally incorrectly send a duplicate job.

job A
job B
job C
job B <- undexpected dup.
job D

This occurs for reason unknown (pool bug?) causing the miner to
restart work on job B and finding the same solutions as the 1st
time through, these then rejected as duplicates.

Furthermore, this prevents any further results which might have
been found for job C.

2 - Protocol startum2 (nicehash) resends job after difficulty
change.

job A
job B
new diff
job B <- incorrectly restarts job B.

Some nicehash pools (counfoundry) resend the previous job after
sending a difficulty update notification. Causes the same problem
as in case 1.

Solution.

Keep a list of for last 4 jobs and drop any incoming duplicates.

[AndreaLanfranchi]

Looks good to me.

I only have a question about that. Will the window of 4 jobs be enough ? Sometimes I see ethermine.org to push 2 or more jobs in fast sequence (within 1 second).

[piotr-dobrogost]

Why is 4 the magic number here? :)