Reject any custom receiver address which wasn't encoded as Bech32m if
the witness program version is greater than zero. These are currently
accepted by bitcoinj but are now invalid and would fail to parse if our
fork was updated to understand Bech32m, to support sending to P2TR
addresses, which the upstream version appears to. (Thus, the presence of
such malformed receivers would not be an issue at present, but might
cause complications in the future.)
Set the burn cap of a candidate to zero if he has an invalid receiver
address, that is, one that bitcoinj cannot parse. This prevents trade
failure when creating the DPT, by making such BM inactive and
distributing their share to the other BM. (Setting the burn cap to zero
is a little more robust than simply filtering out such candidates, as
'BurningManService' handles subsequent share redistribution better than
'(DelayedPayoutTx|BtcFee)ReceiverService'.)
While this case should normally never occur, due to UI validation of the
custom receiver address, there are at least two ways a BM could
invalidate his own receiver address if so inclined:
1) He could simply bypass the UI validation;
2) He could manually create a compensation issuance tx with a change
address type unrecognised by bitcoinj, such as P2TR, as the address
field is pulled straight from the RPC JSON by each full DAO node.
Thus, it is necessary to check both change and custom addresses.
These are both redundant now and will always return true. Also add a
missing past check for Proposal 412 activation to 'RefundManager',
instead of just defaulting to the current date, in case of any very old
disputes involving DPTs created prior to the activation.
Handle the exceptional case of a receiver address chosen from a cycle
where the candidate somehow got more than one compensation proposal
accepted. Either the last custom address or first issuance change
address is supposed to be chosen for the receiver address, but in case
of a tie at that vote result height, take the address that comes
first in lexicographic order.
This avoids needless hashing & equality comparisons of instances of
'BurningManCandidate', which are quite large mutable objects (so should
probably use reference equality anyway, and not be used as keys).
Also rearrange a couple of (package) private methods.
Avoid streaming over the entire proposals list to find a matching txId,
for every 'Issuance' & 'CompensationProposal' pair used to construct and
add a compensation model to the burn output model of each candidate.
Instead, stream over the proposals list once, doing lookups by txId of
each matching issuance, which uses the TreeMap 'DaoState.issuanceMap',
thereby taking O(n*log(n)) time.
