bitcoin#25454 introduces a 10 point penalty for remitting more than
MAX_BLOCKS_TO_ANNOUNCE unconnected block headers. Whether they are
connected or not is determined by taking the first entry and running
its hashPrevBlock through LookupBlockIndex. This new behaviour causes a
test failure in p2p_dos_header_tree.py in Dash.
Bitcoin doesn't face a test failure with this new behaviour as the first
non-fork block in its test data is the dump for block 1 (00000000b873e7
9784647a6c82962c70d228557d24a747ea4d1b8bbe878e1206) but Dash uses block
0 (00000bafbc94add76cb75e2ec92894837288a481e5c005f6563d91623bf8bc2c),
the genesis block.
By definition of a genesis block, it has a hashPrevBlock of 0, which
cannot be looked up. This trips the penalty. This doesn't cause any
problems in the field as nobody is expected to ever broadcast the
genesis block but it does cause a test failure for us.
We need to correct that by getting rid of the genesis block from the
test data.
Backport notice:
- data have real blocks from testnet
- due to big gap in blocks before checkpoints (half-year) for sack of this test is added one more checkpoint, otherwise blocks are not accepted due to non-mockable time for other chains except regtest
- data for 2 blocks in split chain are generated locally for testnet
--------------
333317ce6b67aa92f7363d48cd750712190b4b6b test: Test that low difficulty chain fork is rejected (MarcoFalke)
fa31dc1bf4ee471c4641eef8de02702ba0619ae7 test: Pass down correct chain name in tests (MarcoFalke)
Pull request description:
To prevent OOM, Bitcoin Core will reject chain forks at low difficulty by default. This is the only use-case of checkpoints, so add a test for it to make sure the feature works as expected. If it didn't work, checkpoints would have no use-case and we might as well remove them
ACKs for top commit:
Sjors:
Thanks for adding the node 1 example. Code review ACK 333317c
Tree-SHA512: 90dffa540d0904f3cffb61d2382b1a26f84fe9560b7013e4461546383add31a8757b350616a6d43217c59ef7b8b2a1b62bb3bab582c679cbb2c660a782ce7be1
