Rather manually listing all debug categories, with the risking of
forgetting to update the help text everytime we add a new debug
categories, use Logging::LogGetAllString() instead.

• This is not really needed since we already have all the necessary locks. The only place we need the CORRAL is when processing a block where we don't want transactions entering the mempool while we're removing conflicts and trasactions from the mempool. Further to that, we have for a long time not taken the CORRAL in the wallet when looking for new coins under certain loadtest conditions and without any issues there.

  • One concerin with removing the CORRAL here is that a child transaction could be attempting to enter the mempool just before we try to commit transactions and thus the child would end up in the orphan pool, however, as soon as the parent is fully commited then we follow by processing the orphan pool. That child txn in the orphan pool would then get immediately processed and find it's way into the mempool. So there is no issue here, only a minor time spent as an orphan but this would be a very rare event.

After reading some of the comments I'm just going to say it. Nexa wallet is unreliable on Windows and shouldn't be used. Period. Compiled from source TWICE on Windows using MinGW and a perfect Berkeley DB 5.3 both of which built without warnings of any kind. Mind you I have done this with about a dozen other BTC-qt wallet clones so I know it works. Nexa just doesn't, and it's the Nexa blockchain, not the wallet.

One day I tried to run it and it started syncing out of the blue -> real fast too. Then just stopped at 60% and wouldn't restart -> definitely some kind of weird blockchain peering issue. After loss of contact with the blockchain for more than five minutes or so Nexa just quietly closes of it's own accord, nothing in the logs except a few blockchain updates (!!) Manually adding peers per the wiki doesn't change jack.

I have been reading a lot of VERY BAD ADVICE on this project, like DISABLE FIREWALL, ANTIVIRUS, even port-forwarding to make it work. All nonsense. Nobody else needs any of that, it's unsafe, and hardly something the user of an anonymous wallet would find desirable. Too bad but I think this project has serious (and leadership?) blockchain issues that need addressed.

• this way we don't have to do a relatively expensive coin cache lookup and loop through the inputs twice.

Just use IsFork1Enabled that returns true for every block
for which GetMedianTimePast() >= activation time, ie if blocks
belong to the interval [x-1, +inf).

where:

 x-1 = first block for which median time past >= activation time
 x = first block where the new consensus rules are enforced

if you need to know if the current block is x-1 (ie activation
block (*)) you can use this construct:

 IsFork1Enabled(tip) && !IsFork1Enabled(tip->pprev)

(*) this would be useful to start accepting transaction that use
new consensus rules in the mempool before being able to mine
a block with the new rules activated.

• In RemoveForBlock() we can skip looking of ancestors if we already know we don't have any. This saves considerable time when processing large blocks. On a high end machine about 0.5 seconds for a 100MB block, on a low end machine 1 to 2 seconds.