Previously, each NodeConnCB had its own lock to synchronize data structures
used by the testing thread and the networking thread, and NodeConn provided a
separate additional lock for synchronizing access to each send buffer. This
commit replaces those locks with a single global lock (mininode_lock) that we
use to synchronize access to all data structures shared by the two threads.
Updates comptool and maxblocksinflight to use the new synchronization
semantics, eliminating previous race conditions within comptool, and re-enables
invalidblockrequest.py in travis.
f46a680 Better mruset unit test (Pieter Wuille)
d4d5022 Use ring buffer of set iterators instead of deque of copies in mruset (Pieter Wuille)
d81cff3 Replace mruset setAddrKnown with CRollingBloomFilter addrKnown (Gavin Andresen)
69a5f8b Rolling bloom filter class (Gavin Andresen)
Use a probabilistic bloom filter to keep track of which addresses
we think we have given our peers, instead of a list.
This uses much less memory, at the cost of sometimes failing to
relay an address to a peer-- worst case if the bloom filter happens
to be as full as it gets, 1-in-1,000.
Measured memory usage of a full mruset setAddrKnown: 650Kbytes
Constant memory usage of CRollingBloomFilter addrKnown: 37Kbytes.
This will also help heap fragmentation, because the 37K of storage
is allocated when a CNode is created (when a connection to a peer
is established) and then there is no per-item-remembered memory
allocation.
I plan on testing by restarting a full node with an empty peers.dat,
running a while with -debug=addrman and -debug=net, and making sure
that the 'addr' message traffic out is reasonable.
(suggestions for better tests welcome)
For when you need to keep track of the last N items
you've seen, and can tolerate some false-positives.
Rebased-by: Pieter Wuille <pieter.wuille@gmail.com>
2703412 Fix default binary in p2p tests to use environment variable (Suhas Daftuar)
29bff0e Add some travis debugging for python scripts (Suhas Daftuar)
d76412b Add script manipulation tools for use in mininode testing framework (Suhas Daftuar)
b93974c Add comparison tool test runner, built on mininode (Suhas Daftuar)
6c1d1ba Python p2p testing framework (Suhas Daftuar)
script.py is modified from the code in python-bitcoinlib, and provides tools
for manipulating and creating CScript objects.
bignum.py is a dependency for script.py
script_test.py is an example test that uses the script tools for running a test
that compares the behavior of two nodes, in a comptool- style test, for each of
the test cases in the bitcoin unit test script files, script_valid.json and
script_invalid.json. (This test is very slow to run, but is a proof of concept
for how we can write tests to compare consensus-critical behavior between
different versions of bitcoind.)
bipdersig-p2p.py is another example test in the comptool framework, which tests
deployment of BIP DERSIG for a single node. It uses the script.py tools for
manipulating signatures to be non-DER compliant.
comptool.py creates a tool for running a test suite on top of the mininode p2p
framework. It supports two types of tests: those for which we expect certain
behavior (acceptance or rejection of a block or transaction) and those for
which we are just comparing that the behavior of 2 or more nodes is the same.
blockstore.py defines BlockStore and TxStore, which provide db-backed maps
between block/tx hashes and the corresponding block or tx.
blocktools.py defines utility functions for creating and manipulating blocks
and transactions.
invalidblockrequest.py is an example test in the comptool framework, which
tests the behavior of a single node when sent two different types of invalid
blocks (a block with a duplicated transaction and a block with a bad coinbase
value).
mininode.py provides a framework for connecting to a bitcoin node over the p2p
network. NodeConn is the main object that manages connectivity to a node and
provides callbacks; the interface for those callbacks is defined by NodeConnCB.
Defined also are all data structures from bitcoin core that pass on the network
(CBlock, CTransaction, etc), along with de-/serialization functions.
maxblocksinflight.py is an example test using this framework that tests whether
a node is limiting the maximum number of in-flight block requests.
This also adds support to util.py for specifying the binary to use when
starting nodes (for tests that compare the behavior of different bitcoind
versions), and adds maxblocksinflight.py to the pull tester.
1ec900a Remove broken+useless lock/unlock log prints (Matt Corallo)
352ed22 Add merkle blocks test (Matt Corallo)
59ed61b Add RPC call to generate and verify merkle blocks (Matt Corallo)
30da90d Add CMerkleBlock constructor for tx set + block and an empty one (Matt Corallo)
This commit adds several tests to the script_invalid.json data which
exercise some edge conditions that are not currently being tested.
These are mainly being added to cover several cases a branch coverage
analysis of btcd showed are not already being covered, but given more
tests of edge conditions are always a good thing, I'm contributing
them upstream.
The test which is intended to prove that the script engine is properly
rejecting non-minimally encoded PUSHDATA4 data is using the wrong
opcode and value. The test is using 0x4f, which is OP_1NEGATE instead
of the desired 0x4e, which is OP_PUSHDATA4. Further, the push of data
is intended to be 256 bytes, but the value the test is using is
0x00100000 (4096), instead of the desired 0x00010000 (256).
This commit fixes both issues.
This was found while examining the branch coverage in btcd against only
these tests to help find missing branch coverage.
Suhas Daftuar