All Node classes in individual test cases subclass from NodeConnCB. Many
have duplicate definitions for methods that are defined in the base
class. This commit removes those duplicate definitions.
This commit removes ~290 lines of duplicate code.
This commit adds some helper functions to NodeConnCB which are useful
for many tests:
- NodeConnCB now keeps track of the number of each message type that
it's received and the most recent message of each type. Many tests
assert on the most recent block, tx or reject message.
- NodeConnCB now keeps track of its connection state by setting a
connected boolean in on_open() and on_close()
- NodeConnCB now has wait_for_block, wait_for_getdata,
wait_for_getheaders, wait_for_inv and wait_for_verack methods
I have updated the individual test cases to make sure that there are no
namespace problems that cause them to fail with these new definitions.
Future commits will remove the duplicate code.
maxblocksinflight tested that a node would not send get_data messages
for more than 16 new blocks at the same time. bitcoin core no longer
responds to block invs with get_data, since it does headers-first
sync'ing. This test was therefore testing nothing and can be removed.
the sendheaders test script tests that bitcoin will not send get_headers
for more than 16 blocks simultaneously.
This commit reduces spammy logging by the test framework. It truncates
logging send/receive message in mininode to 500 characters. mininode
was previously logging the entire message sent received, which can be up
to 1MB for a full block.
sync_with_ping currently returns false if the timeout expires, and it is
the caller's responsibility to fail the test. However, none of the tests
currently assert on sync_with_ping()'s return code. This commit adds an
assert to sync_with_ping so the test will fail if the timeout expires.
This commit also removes all the duplicate implementations of
sync_with_ping() from the individual tests.
p2p-compactblocks was incorrectly using sync_with_ping() when sending in
invalid block. The node would disconnect us and never respond to the
ping, so the sync_with_ping would just time out after 30 seconds and
continue with the test.
This commit adds a send_await_disconnect() method that sends the
message, and then waits for the node to disconnect us. In this commit
I've added the method to p2p-compactblocks.py, but a future commit could
move it to mininode since it could be useful more generally.
This commit reduces the p2p-compactblock runtime by 30 seconds.
forknotify would intermittently fail because the alert file was not
being written fast enough. This commit adds a timeout so the test does
not fail immediately.
bip9-sofforks.py stop-starts the bitcoind node twice during the test
run, but it doesn't wait for the connection from mininode to open before
continuing with the test. This leads to race conditions where the test
can fail getblocktemplate() because it has no p2p connections.
This commit merges the NodeConnCB and SingleNodeConnCB into a single
class (called NodeConnCB). The original intent for the NodeConnCB was to
be able to have a python 'mininode' connect to multiple running
bitcoinds. This has never been used and can be achieved more easily by
having multiple NodeConns backed by a common datastore if it is ever
needed.
The changes in mininode.py are just code moves (and merging the two
classes into a single class). The code changes in the individual test
cases are changing the subclasses to subclass from NodeConnCB instead of
SingleNodeConnCB. There is a lot of duplicate code in the subclasses
that can be removed in future commits.
rpt-tests.py outputs progress information as it runs tests. This commit
adds a --quiet option that suppresses that progress output and only
prints a summary of results (and logs from failed tests).
assumevalid was merged as part of PR 9484, but was not added to the
test_runner, so is not run even as part of the extended tests.
This commit adds assumevalid to the list of tests in test_runner. It
also clarifies the code in assumevalid considerably.
Currently, functional test cases can either pass or fail. There are
occasions when it is helpful to skip tests, for example if the
system they are running on does not meet the requirements for the test.
The rest of the test suite can run without being marked as a failure.
This commit adds framework for tests to skip if their requirements
aren't met.
John Newbery