Merge pull request #6216

db6047d Take the training wheels off anti-fee-sniping (Peter Todd)
This commit is contained in:
Wladimir J. van der Laan 2015-12-02 13:56:52 +01:00
commit 83f06ca937
No known key found for this signature in database
GPG Key ID: 74810B012346C9A6

View File

@ -1871,15 +1871,25 @@ bool CWallet::CreateTransaction(const vector<CRecipient>& vecSend, CWalletTx& wt
// Discourage fee sniping.
//
// However because of a off-by-one-error in previous versions we need to
// neuter it by setting nLockTime to at least one less than nBestHeight.
// Secondly currently propagation of transactions created for block heights
// corresponding to blocks that were just mined may be iffy - transactions
// aren't re-accepted into the mempool - we additionally neuter the code by
// going ten blocks back. Doesn't yet do anything for sniping, but does act
// to shake out wallet bugs like not showing nLockTime'd transactions at
// all.
txNew.nLockTime = std::max(0, chainActive.Height() - 10);
// For a large miner the value of the transactions in the best block and
// the mempool can exceed the cost of delibrately attempting to mine two
// blocks to orphan the current best block. By setting nLockTime such that
// only the next block can include the transaction, we discourage this
// practice as the height restricted and limited blocksize gives miners
// considering fee sniping fewer options for pulling off this attack.
//
// A simple way to think about this is from the wallet's point of view we
// always want the blockchain to move forward. By setting nLockTime this
// way we're basically making the statement that we only want this
// transaction to appear in the next block; we don't want to potentially
// encourage reorgs by allowing transactions to appear at lower heights
// than the next block in forks of the best chain.
//
// Of course, the subsidy is high enough, and transaction volume low
// enough, that fee sniping isn't a problem yet, but by implementing a fix
// now we ensure code won't be written that makes assumptions about
// nLockTime that preclude a fix later.
txNew.nLockTime = chainActive.Height();
// Secondly occasionally randomly pick a nLockTime even further back, so
// that transactions that are delayed after signing for whatever reason,