89e7c4274c
5f36e0ff1erpc: fix getblockstats UTXO overhead accounting (Lőrinc)76190489e6coins: pack `Coin` height/coinbase consistently (Lőrinc)1f309d1aa2coins: make `Coin::fCoinBase` a bool (Lőrinc) Pull request description: The [`getblockstats` RPC](https://github.com/bitcoin/bitcoin/pull/10757) currently overestimates UTXO overhead by treating the `fCoinBase` bitfield as an additional `bool` in `PER_UTXO_OVERHEAD`. However, `fCoinBase` and `nHeight` are stored as bitfields and effectively packed into a single 32-bit value; counting an extra bool in the overhead calculation is unnecessary. This PR introduces the following changes across three commits: * Store `fCoinBase` as a `bool` bitfield to reduce implicit conversions at call sites. * Use a consistent height/coinbase packing style across `Coin` serialization, undo serialization, and coinstats hashing (casting `nHeight` to `uint32_t` before shifting to avoid signed-promotion UB). * Adjust UTXO overhead estimation to match the actual `Coin` layout and update the related tests accordingly. ACKs for top commit: achow101: ACK5f36e0ff1esedited: ACK5f36e0ff1evasild: ACK5f36e0ff1eoptout21: crACK5f36e0ff1eTree-SHA512: f4a44debed358e9e130da9d6fae5f89289daa34f0bdb7155edc3e9b691c219451f4c80b1e16aca5b011f0fa2fa975484ef1af4ca4563b7c6ba4ca9dd133f30be
Various test vectors
mainnet_alt.json
For easier testing the difficulty is maximally increased in the first (and only) retarget period, by producing blocks approximately 2 minutes apart.
The alternate mainnet chain was generated as follows:
- use faketime to set node clock to 2 minutes after genesis block
- mine a block using a CPU miner such as https://github.com/pooler/cpuminer
- restart node with a faketime 2 minutes later
for i in {1..2016}
do
t=$(( 1231006505 + $i * 120 ))
faketime "`date -d @$t +'%Y-%m-%d %H:%M:%S'`" \
bitcoind -connect=0 -nocheckpoints -stopatheight=$i
done
The CPU miner is kept running as follows:
./minerd -u ... -p ... -o http://127.0.0.1:8332 --no-stratum \
--coinbase-addr 1NQpH6Nf8QtR2HphLRcvuVqfhXBXsiWn8r \
--algo sha256d --no-longpoll --scantime 3 --retry-pause 1
The payout address is derived from first BIP32 test vector master key:
pkh(xprv9s21ZrQH143K3QTDL4LXw2F7HEK3wJUD2nW2nRk4stbPy6cq3jPPqjiChkVvvNKmPGJxWUtg6LnF5kejMRNNU3TGtRBeJgk33yuGBxrMPHi/44h/0h/0h/<0;1>/*)#fkjtr0yn
It uses pkh() because tr() outputs at low heights are not spendable (unexpected-witness).
This makes each block deterministic except for its timestamp and nonce, which
are stored in mainnet_alt.json and used to reconstruct the chain without
having to redo the proof-of-work.
The timestamp was not kept constant because at difficulty 1 it's not sufficient to only grind the nonce. Grinding the extra_nonce or version field instead would have required additional (stratum) software. It would also make it more complicated to reconstruct the blocks in this test.
The getblocktemplate RPC code needs to be patched to ignore not being connected
to any peers, and to ignore the IBD status check.
On macOS use faketime "@$t" instead.