Enacted & Retracted blocks' transactions removed/added to transaction…

[tomusdrw]

… queue

Closes #649

[tomusdrw]

Not sure where (if) should I merge good+enacted bad+retracted.

[gavofyork]

i'd say generally "good" should become "enacted" and "bad" to "retracted". i don't think they're used for different meanings.

[tomusdrw]

I agree that good \subseteq enacted, but bad are blocks that have never made to the chain - not sure if it's ok to call them retracted. So in general the question is what to pass in SyncMessage::NewChainBlocks - I'm good with passing just enacted and retracted and doing the merge in client.

Can you confirm my assumption that \forall x | x \in good \implies x \in enacted?

[gavofyork]

what is routes?

[gavofyork]

ahh so:

• "good" actually means "imported";
• "bad" means "invalid";
• "enacted" means was not previously in the canon chain, and now is;
• "retracted" means was previously in the canon chain, and now is not.

as such, none of them are subsets of the other.

[gavofyork]

i suggest we rename good and bad to imported and invalid, and place a comprehensive description somewhere very visible in the documentation.

[tomusdrw]

But isn't good - (enacted + retracted) = 0 (all blocks that are in good are actually included in one of enacted/retracted?

[tomusdrw]

@gavofyork Is the block_queue is empty check needed here?
My understanding is that if we don't send the event - some of the blocks' transactions will not be removed from queue because of that (cause in next event the blocks won't be there)

[gavofyork]

0 -> 1 -> 2a -> 3a -> 4a
       \-> 2b

3b, 4b, 5b, 3c

good: 3b, 4b, 5b, 3c
enacted: 2b, 3b, 4b, 5b
retracted: 4a, 3a, 2a

0 -> 1 -> 2a -> 3a -> 4a
       \-> 2b -> 3b -> 4b -> 5b
              \-> 3c

(all disjoint)

[gavofyork]

Fails on:

thread 'chain::tests::should_add_transactions_to_queue' panicked at 'assertion failed: (left == right) (left: 0, right: 1)', sync/src/chain.rs:1664

[gavofyork]

now contained in #700