Beyond the crash already made public, CVE-2018-17144 let a miner inflate the Bitcoin supply in versions 0.15.X to 0.16.2 by spending an output twice inside one transaction. Developers held back that part to give systems time to move to 0.16.3 or 0.17.0rc4. Over half the hashrate is believed patched, no exploit attempts are known, and affected users are urged to upgrade.
CVE-2018-17144 makes Bitcoin Core 0.14.0 through 0.16.2 crash on a block containing a transaction that spends one input twice, and users are told to upgrade at once to 0.16.3. Only a miner ready to give up at least 12.5 BTC in allowed income could produce such a block.
The verifytxoutproof RPC in Bitcoin Core 0.16.2 no longer falls to an expensive attack on SPV proofs disclosed in early June, though real SPV clients remain exposed. A separate fix makes abandontransaction drop every descendant of a transaction instead of stopping at its children.
OP_CODESEPARATOR in legacy signature scripts is no longer relayed by Bitcoin Core 0.16.1, a policy change that makes future fixes for slow-to-validate blocks easier to deploy. Miners lose the -blockmaxsize option in favour of -blockmaxweight and get a fix for a bug that could have cost them revenue when they produced two blocks in rapid succession.
Segwit becomes the wallet default in Bitcoin Core 0.16.0: new addresses are p2sh-segwit unless -addresstype says otherwise, and native bc1 addresses under BIP173 can be generated and paid. New wallets are HD only and earlier versions reject them, while the GUI send screen now turns on BIP125 RBF by default.
As a precaution against future network forks, Bitcoin Core 0.15.1 drops peers offering a chain below the minimum-chain-work during initial sync, along with outbound peers serving headers known to be invalid. A tip stuck for 30 minutes now triggers an extra outbound connection. Miners lose direct use of blockmaxsize, which is deprecated and only implies a weight limit.
A startup crash in the GUI prompted Bitcoin Core 0.15.0.1: some clients upgrading to 0.15.0 fell over on a custom fee setting the interface no longer offers. A minimal patch, #11332, handles the invalid value.
Fee estimates in Bitcoin Core 0.15.0 reach 1,008 blocks ahead, up from 25, and GUI users can raise the fee on an unconfirmed transaction through Replace-By-Fee. A per-output chainstate database speeds up initial block download by 30-40%. Converting it happens automatically on first start and takes about 5 to 30 minutes.
Regular Bitcoin Core contributors have no ties to btc1, the Segwit2x implementation, and the project rejects claims that it is a necessary update, calling it a contentious break from network rules. Segwit, the soft fork due within days, is unrelated to that hard fork and asks nothing of most users. Segwit2x also plans no strong replay protection, the statement warns.
Anyone who enabled UPnP is urged to install Bitcoin Core 0.14.2 without delay: an integer signedness error in the bundled miniupnpc, CVE-2017-8798, lets attackers on the local network cause a denial of service or possibly other harm. UPnP stays off by default, so nodes that never switched it on are unaffected.
Non-segwit miners can keep using getblocktemplate after activation, because Bitcoin Core 0.14.1 strips segwit transactions from the template. The -dbcache default moves to 450MiB and the limit now holds through flush peaks. Renamed arguments in createrawtransaction and disconnectnode break callers that use named arguments from 0.14.0.
A new node's wallet could send and receive before sync ends under the hybrid full block SPV mode on the Bitcoin Core roadmap, which fetches blocks by the wallet's oldest key first while full validation continues in the background. Bloom filter privacy leaks go away at the price of more bandwidth, and payments should wait for 6 or more confirmations.
Swapping ECDSA for Schnorr signatures would shrink a 50-of-50 multisig to the size of a single-key spend and hide its policy from observers. Bitcoin Core's roadmap names the main obstacle as cancellation, where a malicious signer subtracts another user's key. Delinearization answers that attack, though its security proof is still in peer review.
Signature caching in Bitcoin-Qt 0.7.0, Ultraprune with LevelDB in 0.8.0 and headers-first sync in Bitcoin Core 0.10 feature in part 1 of a look back at speedups to the reference software. Parallel script verification alone gave 35% to 100% in developers' benchmarks.
New nodes finished initial block download 5.7 times faster on Bitcoin Core 0.14.0 than on 0.13.2 in an Amazon EC2 test with default settings: 6 hours 24 minutes against 1 day 12 hours 40 minutes. Assumed valid blocks and memory shared between the mempool and UTXO cache drive that gain. Wallets also get optional fee bumping, off by default.
Wallet sends that hit the limit on unconfirmed chains no longer return an RPC error in Bitcoin Core 0.13.2. They report success, and mempool insertion is retried at rebroadcast with no restart needed. Fee estimates for a 1-block target are disabled, and handling of invalid compact blocks is fixed.
Wrapping segwit in P2SH carries the biggest byte cost: 19% extra for P2WSH/P2SH and 11% for P2WPKH/P2SH on a 180-byte transaction, by Bitcoin Core's tally of what BIP141 would bring. Validation needs about five more SHA256 hashes per transaction. Bugs, larger blocks and lower fees are weighed as risks, and the authors take no position on activation.
Miners get code to signal for the segwit soft fork in Bitcoin Core 0.13.1, with signalling open from 15 November 2016 at the start of the next retarget period. Lock-in takes 95% of blocks in a 2,016-block period, and activation comes one period later. Wallets stay compatible, so nobody but miners may need to act.
Segwit needs two groups on board: miners signaling bit 1 in 95% of blocks over a retarget period, and enough full nodes running Bitcoin Core 0.13.1 or a compatible implementation. Signaling opens at the first retarget that falls on or after 15 November 2016. Separate steps cover miners, node operators, wallet users and developers, including those who choose not to upgrade.
Segwit code lands in Bitcoin Core 0.13.0 but stays inactive on mainnet; using it there will take a later version, expected to be 0.13.1. Compact block relay (BIP152) trims bandwidth spikes, and new wallets use BIP32 HD keys, so one backup covers every future private key. Linux ARM builds also join the official downloads.