diff options
author | John Newbery <john@johnnewbery.com> | 2017-03-08 15:56:30 -0500 |
---|---|---|
committer | John Newbery <john@johnnewbery.com> | 2017-03-20 10:40:31 -0400 |
commit | 00902c48cdc5d6a531abfd279fd8e7779cf0a0b8 (patch) | |
tree | 41891484a478a7478937fe4a12b63e6c7f31799a /test/rpc-tests/test_framework/comptool.py | |
parent | d34995a7bac6ed20ce42aa00c6252b900786e649 (diff) |
Rename qa directory to test
Diffstat (limited to 'test/rpc-tests/test_framework/comptool.py')
-rwxr-xr-x | test/rpc-tests/test_framework/comptool.py | 410 |
1 files changed, 410 insertions, 0 deletions
diff --git a/test/rpc-tests/test_framework/comptool.py b/test/rpc-tests/test_framework/comptool.py new file mode 100755 index 0000000000..70d1d700ef --- /dev/null +++ b/test/rpc-tests/test_framework/comptool.py @@ -0,0 +1,410 @@ +#!/usr/bin/env python3 +# Copyright (c) 2015-2016 The Bitcoin Core developers +# Distributed under the MIT software license, see the accompanying +# file COPYING or http://www.opensource.org/licenses/mit-license.php. +"""Compare two or more bitcoinds to each other. + +To use, create a class that implements get_tests(), and pass it in +as the test generator to TestManager. get_tests() should be a python +generator that returns TestInstance objects. See below for definition. + +TestNode behaves as follows: + Configure with a BlockStore and TxStore + on_inv: log the message but don't request + on_headers: log the chain tip + on_pong: update ping response map (for synchronization) + on_getheaders: provide headers via BlockStore + on_getdata: provide blocks via BlockStore +""" + +from .mininode import * +from .blockstore import BlockStore, TxStore +from .util import p2p_port + +import logging + +logger=logging.getLogger("TestFramework.comptool") + +global mininode_lock + +class RejectResult(object): + """Outcome that expects rejection of a transaction or block.""" + def __init__(self, code, reason=b''): + self.code = code + self.reason = reason + def match(self, other): + if self.code != other.code: + return False + return other.reason.startswith(self.reason) + def __repr__(self): + return '%i:%s' % (self.code,self.reason or '*') + +class TestNode(NodeConnCB): + + def __init__(self, block_store, tx_store): + NodeConnCB.__init__(self) + self.conn = None + self.bestblockhash = None + self.block_store = block_store + self.block_request_map = {} + self.tx_store = tx_store + self.tx_request_map = {} + self.block_reject_map = {} + self.tx_reject_map = {} + + # When the pingmap is non-empty we're waiting for + # a response + self.pingMap = {} + self.lastInv = [] + self.closed = False + + def on_close(self, conn): + self.closed = True + + def add_connection(self, conn): + self.conn = conn + + def on_headers(self, conn, message): + if len(message.headers) > 0: + best_header = message.headers[-1] + best_header.calc_sha256() + self.bestblockhash = best_header.sha256 + + def on_getheaders(self, conn, message): + response = self.block_store.headers_for(message.locator, message.hashstop) + if response is not None: + conn.send_message(response) + + def on_getdata(self, conn, message): + [conn.send_message(r) for r in self.block_store.get_blocks(message.inv)] + [conn.send_message(r) for r in self.tx_store.get_transactions(message.inv)] + + for i in message.inv: + if i.type == 1: + self.tx_request_map[i.hash] = True + elif i.type == 2: + self.block_request_map[i.hash] = True + + def on_inv(self, conn, message): + self.lastInv = [x.hash for x in message.inv] + + def on_pong(self, conn, message): + try: + del self.pingMap[message.nonce] + except KeyError: + raise AssertionError("Got pong for unknown ping [%s]" % repr(message)) + + def on_reject(self, conn, message): + if message.message == b'tx': + self.tx_reject_map[message.data] = RejectResult(message.code, message.reason) + if message.message == b'block': + self.block_reject_map[message.data] = RejectResult(message.code, message.reason) + + def send_inv(self, obj): + mtype = 2 if isinstance(obj, CBlock) else 1 + self.conn.send_message(msg_inv([CInv(mtype, obj.sha256)])) + + def send_getheaders(self): + # We ask for headers from their last tip. + m = msg_getheaders() + m.locator = self.block_store.get_locator(self.bestblockhash) + self.conn.send_message(m) + + def send_header(self, header): + m = msg_headers() + m.headers.append(header) + self.conn.send_message(m) + + # This assumes BIP31 + def send_ping(self, nonce): + self.pingMap[nonce] = True + self.conn.send_message(msg_ping(nonce)) + + def received_ping_response(self, nonce): + return nonce not in self.pingMap + + def send_mempool(self): + self.lastInv = [] + self.conn.send_message(msg_mempool()) + +# TestInstance: +# +# Instances of these are generated by the test generator, and fed into the +# comptool. +# +# "blocks_and_transactions" should be an array of +# [obj, True/False/None, hash/None]: +# - obj is either a CBlock, CBlockHeader, or a CTransaction, and +# - the second value indicates whether the object should be accepted +# into the blockchain or mempool (for tests where we expect a certain +# answer), or "None" if we don't expect a certain answer and are just +# comparing the behavior of the nodes being tested. +# - the third value is the hash to test the tip against (if None or omitted, +# use the hash of the block) +# - NOTE: if a block header, no test is performed; instead the header is +# just added to the block_store. This is to facilitate block delivery +# when communicating with headers-first clients (when withholding an +# intermediate block). +# sync_every_block: if True, then each block will be inv'ed, synced, and +# nodes will be tested based on the outcome for the block. If False, +# then inv's accumulate until all blocks are processed (or max inv size +# is reached) and then sent out in one inv message. Then the final block +# will be synced across all connections, and the outcome of the final +# block will be tested. +# sync_every_tx: analogous to behavior for sync_every_block, except if outcome +# on the final tx is None, then contents of entire mempool are compared +# across all connections. (If outcome of final tx is specified as true +# or false, then only the last tx is tested against outcome.) + +class TestInstance(object): + def __init__(self, objects=None, sync_every_block=True, sync_every_tx=False): + self.blocks_and_transactions = objects if objects else [] + self.sync_every_block = sync_every_block + self.sync_every_tx = sync_every_tx + +class TestManager(object): + + def __init__(self, testgen, datadir): + self.test_generator = testgen + self.connections = [] + self.test_nodes = [] + self.block_store = BlockStore(datadir) + self.tx_store = TxStore(datadir) + self.ping_counter = 1 + + def add_all_connections(self, nodes): + for i in range(len(nodes)): + # Create a p2p connection to each node + test_node = TestNode(self.block_store, self.tx_store) + self.test_nodes.append(test_node) + self.connections.append(NodeConn('127.0.0.1', p2p_port(i), nodes[i], test_node)) + # Make sure the TestNode (callback class) has a reference to its + # associated NodeConn + test_node.add_connection(self.connections[-1]) + + def clear_all_connections(self): + self.connections = [] + self.test_nodes = [] + + def wait_for_disconnections(self): + def disconnected(): + return all(node.closed for node in self.test_nodes) + return wait_until(disconnected, timeout=10) + + def wait_for_verack(self): + def veracked(): + return all(node.verack_received for node in self.test_nodes) + return wait_until(veracked, timeout=10) + + def wait_for_pings(self, counter): + def received_pongs(): + return all(node.received_ping_response(counter) for node in self.test_nodes) + return wait_until(received_pongs) + + # sync_blocks: Wait for all connections to request the blockhash given + # then send get_headers to find out the tip of each node, and synchronize + # the response by using a ping (and waiting for pong with same nonce). + def sync_blocks(self, blockhash, num_blocks): + def blocks_requested(): + return all( + blockhash in node.block_request_map and node.block_request_map[blockhash] + for node in self.test_nodes + ) + + # --> error if not requested + if not wait_until(blocks_requested, attempts=20*num_blocks): + raise AssertionError("Not all nodes requested block") + + # Send getheaders message + [ c.cb.send_getheaders() for c in self.connections ] + + # Send ping and wait for response -- synchronization hack + [ c.cb.send_ping(self.ping_counter) for c in self.connections ] + self.wait_for_pings(self.ping_counter) + self.ping_counter += 1 + + # Analogous to sync_block (see above) + def sync_transaction(self, txhash, num_events): + # Wait for nodes to request transaction (50ms sleep * 20 tries * num_events) + def transaction_requested(): + return all( + txhash in node.tx_request_map and node.tx_request_map[txhash] + for node in self.test_nodes + ) + + # --> error if not requested + if not wait_until(transaction_requested, attempts=20*num_events): + raise AssertionError("Not all nodes requested transaction") + + # Get the mempool + [ c.cb.send_mempool() for c in self.connections ] + + # Send ping and wait for response -- synchronization hack + [ c.cb.send_ping(self.ping_counter) for c in self.connections ] + self.wait_for_pings(self.ping_counter) + self.ping_counter += 1 + + # Sort inv responses from each node + with mininode_lock: + [ c.cb.lastInv.sort() for c in self.connections ] + + # Verify that the tip of each connection all agree with each other, and + # with the expected outcome (if given) + def check_results(self, blockhash, outcome): + with mininode_lock: + for c in self.connections: + if outcome is None: + if c.cb.bestblockhash != self.connections[0].cb.bestblockhash: + return False + elif isinstance(outcome, RejectResult): # Check that block was rejected w/ code + if c.cb.bestblockhash == blockhash: + return False + if blockhash not in c.cb.block_reject_map: + logger.error('Block not in reject map: %064x' % (blockhash)) + return False + if not outcome.match(c.cb.block_reject_map[blockhash]): + logger.error('Block rejected with %s instead of expected %s: %064x' % (c.cb.block_reject_map[blockhash], outcome, blockhash)) + return False + elif ((c.cb.bestblockhash == blockhash) != outcome): + return False + return True + + # Either check that the mempools all agree with each other, or that + # txhash's presence in the mempool matches the outcome specified. + # This is somewhat of a strange comparison, in that we're either comparing + # a particular tx to an outcome, or the entire mempools altogether; + # perhaps it would be useful to add the ability to check explicitly that + # a particular tx's existence in the mempool is the same across all nodes. + def check_mempool(self, txhash, outcome): + with mininode_lock: + for c in self.connections: + if outcome is None: + # Make sure the mempools agree with each other + if c.cb.lastInv != self.connections[0].cb.lastInv: + return False + elif isinstance(outcome, RejectResult): # Check that tx was rejected w/ code + if txhash in c.cb.lastInv: + return False + if txhash not in c.cb.tx_reject_map: + logger.error('Tx not in reject map: %064x' % (txhash)) + return False + if not outcome.match(c.cb.tx_reject_map[txhash]): + logger.error('Tx rejected with %s instead of expected %s: %064x' % (c.cb.tx_reject_map[txhash], outcome, txhash)) + return False + elif ((txhash in c.cb.lastInv) != outcome): + return False + return True + + def run(self): + # Wait until verack is received + self.wait_for_verack() + + test_number = 1 + for test_instance in self.test_generator.get_tests(): + # We use these variables to keep track of the last block + # and last transaction in the tests, which are used + # if we're not syncing on every block or every tx. + [ block, block_outcome, tip ] = [ None, None, None ] + [ tx, tx_outcome ] = [ None, None ] + invqueue = [] + + for test_obj in test_instance.blocks_and_transactions: + b_or_t = test_obj[0] + outcome = test_obj[1] + # Determine if we're dealing with a block or tx + if isinstance(b_or_t, CBlock): # Block test runner + block = b_or_t + block_outcome = outcome + tip = block.sha256 + # each test_obj can have an optional third argument + # to specify the tip we should compare with + # (default is to use the block being tested) + if len(test_obj) >= 3: + tip = test_obj[2] + + # Add to shared block_store, set as current block + # If there was an open getdata request for the block + # previously, and we didn't have an entry in the + # block_store, then immediately deliver, because the + # node wouldn't send another getdata request while + # the earlier one is outstanding. + first_block_with_hash = True + if self.block_store.get(block.sha256) is not None: + first_block_with_hash = False + with mininode_lock: + self.block_store.add_block(block) + for c in self.connections: + if first_block_with_hash and block.sha256 in c.cb.block_request_map and c.cb.block_request_map[block.sha256] == True: + # There was a previous request for this block hash + # Most likely, we delivered a header for this block + # but never had the block to respond to the getdata + c.send_message(msg_block(block)) + else: + c.cb.block_request_map[block.sha256] = False + # Either send inv's to each node and sync, or add + # to invqueue for later inv'ing. + if (test_instance.sync_every_block): + # if we expect success, send inv and sync every block + # if we expect failure, just push the block and see what happens. + if outcome == True: + [ c.cb.send_inv(block) for c in self.connections ] + self.sync_blocks(block.sha256, 1) + else: + [ c.send_message(msg_block(block)) for c in self.connections ] + [ c.cb.send_ping(self.ping_counter) for c in self.connections ] + self.wait_for_pings(self.ping_counter) + self.ping_counter += 1 + if (not self.check_results(tip, outcome)): + raise AssertionError("Test failed at test %d" % test_number) + else: + invqueue.append(CInv(2, block.sha256)) + elif isinstance(b_or_t, CBlockHeader): + block_header = b_or_t + self.block_store.add_header(block_header) + [ c.cb.send_header(block_header) for c in self.connections ] + + else: # Tx test runner + assert(isinstance(b_or_t, CTransaction)) + tx = b_or_t + tx_outcome = outcome + # Add to shared tx store and clear map entry + with mininode_lock: + self.tx_store.add_transaction(tx) + for c in self.connections: + c.cb.tx_request_map[tx.sha256] = False + # Again, either inv to all nodes or save for later + if (test_instance.sync_every_tx): + [ c.cb.send_inv(tx) for c in self.connections ] + self.sync_transaction(tx.sha256, 1) + if (not self.check_mempool(tx.sha256, outcome)): + raise AssertionError("Test failed at test %d" % test_number) + else: + invqueue.append(CInv(1, tx.sha256)) + # Ensure we're not overflowing the inv queue + if len(invqueue) == MAX_INV_SZ: + [ c.send_message(msg_inv(invqueue)) for c in self.connections ] + invqueue = [] + + # Do final sync if we weren't syncing on every block or every tx. + if (not test_instance.sync_every_block and block is not None): + if len(invqueue) > 0: + [ c.send_message(msg_inv(invqueue)) for c in self.connections ] + invqueue = [] + self.sync_blocks(block.sha256, len(test_instance.blocks_and_transactions)) + if (not self.check_results(tip, block_outcome)): + raise AssertionError("Block test failed at test %d" % test_number) + if (not test_instance.sync_every_tx and tx is not None): + if len(invqueue) > 0: + [ c.send_message(msg_inv(invqueue)) for c in self.connections ] + invqueue = [] + self.sync_transaction(tx.sha256, len(test_instance.blocks_and_transactions)) + if (not self.check_mempool(tx.sha256, tx_outcome)): + raise AssertionError("Mempool test failed at test %d" % test_number) + + logger.info("Test %d: PASS" % test_number) + test_number += 1 + + [ c.disconnect_node() for c in self.connections ] + self.wait_for_disconnections() + self.block_store.close() + self.tx_store.close() |