eth/catalyst: force sync of txpool before clearing subpools in Rollback

[dahu33]

This PR enforces a transaction pool synchronization before clearing it in the (c *SimulatedBeacon) Rollback() function.

An alternative approach would be to modify (p *TxPool) Clear() to call (p *TxPool) Sync() directly before clearing, but this PR opts to perform the synchronization explicitly in Rollback().

Issue Description

In the example below, without the enforced sync, transactions from blocks between oldHead and newHead are unexpectedly included in the newly committed block, rendering the rollback ineffective.

simulatedBeacon.Fork(newHead)
simulatedBeacon.Rollback()
simulatedBeacon.Commit()

However, introducing a time.Sleep between Fork() and Rollback() produces the expected behavior, where transactions from intermediate blocks are correctly dropped, confirming the flakey behavior:

simulatedBeacon.Fork(newHead)
time.Sleep(1 * time.Second)
simulatedBeacon.Rollback()
simulatedBeacon.Commit()

[jwasinger]

I think perhaps we could perform a Sync as the first thing within Clear, as it makes sense intuitively: Any in-limbo transactions should be removed by Clear. I have to think about this some more.

[dahu33]

I think perhaps we could perform a Sync as the first thing within Clear, as it makes sense intuitively: Any in-limbo transactions should be removed by Clear. I have to think about this some more.

Agreed, updated the commit to reflect this.

[dahu33]

@jwasinger any update on this?

[jwasinger]

I thought that it might be more proper to make SubPool.Clear sync any pending transactions, instead of doing this behavior in the global pool. But it's not trivial to implement and I don't think it actually adds any value (the subpools themselves aren't exposed).

[dahu33]

Thanks for the review @jwasinger, committed your suggested change which reverted your approval.

[MariusVanDerWijden]

SGTM

[jwasinger]

lint failing. @dahu33 can you fix them? or enable contributor pushing to this branch and I can do it?

[dahu33]

lint failing. @dahu33 can you fix them? or enable contributor pushing to this branch and I can do it?

My bad. Should have achieved perfection now :)