How do I catch the possible unconfirmed transactions, so the ones that
got saved in the db when were first posted, but in the process were
rejected/unconfirmed in the next 6 blocks.
There are two ways a transaction (TXID) is first introduced by walletnotify:
- 1) Acceptance in mempool
- 2) Confirmation in a block
A transaction in mempool has no guarantee of being mined, and a confirmed transaction may have never been previously seen in your local mempool. Therefore, you need to be able to differentiate between the two scenarios, and only store the transactions which actually receive confirmation.
Reorgs: Furthermore, there is no guarantee that a previously confirmed transaction receives confirmation from subsequent blocks. It is possible that subsequent blocks form a separate branch, so that your previously confirmed transaction is no longer in the strong chain. This is where incrementing confirmation with each new block as you describe will fail.
Better to verify the longest header chain (they must link together via prev header hash) and ensure that the transaction is included in this longest chain when computing the level of transaction confirmation. If this longest header chain reorganises, you can update the confirmations of all transactions accordingly.
So how would a transaction that was rejected or wasn't confirmed in the 'subsequent blocks' differ from the confirmed one in the output of gettransaction <txid> – zabbir – 2019-01-26T19:34:55.010
I am not sure whether gettransaction <txid> returns transactions with confirmation field = 0. If it doesn't, that also tells you it isn't confirmed, but was accepted in your local mem-pool. – James C. – 2019-01-27T08:51:19.753
It does, http://rgho.st/8rhMpLRsh. What if a transaction gets 2 confirmations and then gets rejected, how would it appear at gettransaction <txid> ?
– zabbir – 2019-01-28T04:38:34.920Good question. Either 0 conf or not at all (TXID not found). The problem is you don't get any notification of a reorganisation, so you need to determine a reorganisation-event based on your header-chain, when new blocks are announced which don't extend the previous chain. – James C. – 2019-01-28T06:42:54.153
Why would I want to catch these blocks? If transactions in them never get confirmed in my db: how would they be harmful? – zabbir – 2019-01-30T13:59:13.020
If you have TX's stored in your database which are confirmed, they may become unconfirmed (Prevout unspent) or even invalid (Prevout already spent) during a chain reorganisation. That is why you need to recognise a chain reorganisation. This cannot be accomplished simply through tracking chain height. Different competing chains may have the same tip height. – James C. – 2019-01-30T14:21:01.773
But what if I simply check the latest status of confirmations with gettransction <tx> for every notification of blocknotify and update all of my transactions below 6 confirmations. And stop this process for every transaction that has 6 + confs on new block. That's why I was interested how a rejected tx would appear so I can set confirmed to false in my db and it doesn't check it forever. – zabbir – 2019-01-30T15:58:09.407
With rejected tx you mean a previously confirmed transaction which has been reorganised out of the strong chain? Assuming no deeper reorgs than 6 blocks, your 6-block window would make sense to me. – James C. – 2019-01-30T16:05:37.357
Do reorganizations deeper 6 blocks ever happen? – zabbir – 2019-01-30T16:10:15.810
Today? Unlikely, block propagation times down to zero, little to no orphaning and very expensive to create 6-deep competing branch. Historically? I don’t know. – James C. – 2019-01-30T16:11:56.113