Cannot run multiple instances of TestWrapper

[jachiang]

Running setup() on multiple instances of TestWrapper results in a AssertionError:

>>> test = util.TestWrapper()
>>> test2 = util.TestWrapper()
>>> test.setup()
2019-09-13T18:08:04.389000Z TestFramework./var/folders/x6/74wk_20d3t90825bykj23y780000gn/T/bitcoin_func_test_0hnz1ze_ (INFO): Initializing test directory /var/folders/x6/74wk_20d3t90825bykj23y780000gn/T/bitcoin_func_test_0hnz1ze_
>>> test2.setup()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/jamesc/Dropbox/repos/taproot/util.py", line 83, in setup
    super(TestWrapper,self).setup()
  File "/Users/jamesc/Dropbox/repos/test/bitcoin/test/functional/test_framework/test_framework.py", line 196, in setup
    self.network_thread = NetworkThread()
  File "/Users/jamesc/Dropbox/repos/test/bitcoin/test/functional/test_framework/mininode.py", line 469, in __init__
    assert not self.network_event_loop
AssertionError

This occurs because we are initialising a network thread twice. Initialising with a different thread for each TestWrapper object requires refactor of NetworkThread class in mininode.py.

This has been added as a todo to the TestFramework main method refactor PR.

[jnewbery]

I think we don't want users to be able to create multiple TestWrapper instances (and it shouldn't be possible since #17).

@jachiang - is there a reason you want to have more than one instance of TestWrapper?

[jachiang]

@jachiang - is there a reason you want to have more than one instance of TestWrapper?

@jnewbery - No reason for the workshop. It's more that I don't see any reason why it shouldn't be possible to run multiple instances of TestWrapper - If a required resource (e.g. port) is not available due to user misconfiguration, setup() should raise an exception. I think the initial design intention was to limit a single network thread per test run. It would seem cleaner to extend that to a single network thread per test instance.

I definitely think #17 is the best solution for the upcoming workshop. Perhaps for the master PR, it feels like cleaner code to allow for multiple instances. There may be some future use-case in running multiple test instances for benchmarking node cluster configurations?

[jnewbery]

Closing this since it's not needed for the workshop. You can follow up with a PR in the bitcoind repo if you think it's important.