test: fix TEST_PREVIOUS_RELEASES check #18903

Pythons os.getenv() returns None by default when a key isn't found. So if this test is run locally and you haven't set TEST_PREVIOUS_RELEASES or previous release binaries aren't available, it will always fail.

Remaining jobs: [wallet_upgradewallet.py]
1/1 - wallet_upgradewallet.py failed, Duration: 0 s

stdout:
2020-05-07T03:24:36.506000Z TestFramework (INFO): Initializing test directory /var/folders/z2/cn877pxd3czdfh47mfkmbwgm0000gn/T/test_runner_₿_🏃_20200507_112436/wallet_upgradewallet_0
2020-05-07T03:24:36.514000Z TestFramework (ERROR): Assertion failed
Traceback (most recent call last):
  File "bitcoin/test/functional/test_framework/test_framework.py", line 471, in start_nodes
    node.start(extra_args[i], *args, **kwargs)
  File "bitcoin/test/functional/test_framework/test_node.py", line 205, in start
    self.process = subprocess.Popen(self.args + extra_args, env=subp_env, stdout=stdout, stderr=stderr, cwd=cwd, **kwargs)
  File "/Users/x/.pyenv/versions/3.5.6/lib/python3.5/subprocess.py", line 676, in __init__
    restore_signals, start_new_session)
  File "/Users/x/.pyenv/versions/3.5.6/lib/python3.5/subprocess.py", line 1289, in _execute_child
    raise child_exception_type(errno_num, err_msg)
FileNotFoundError: [Errno 2] No such file or directory: 'bitcoin/releases/v0.16.3/bin/bitcoind'

During handling of the above exception, another exception occurred:

Not sure why we are doing bash env var juggling here. This should be an option on the test framework like everything else. E.g. --factor

Not sure why we are doing bash env var juggling here.

🤷‍♂️ I just want to be able to run the full functional test suite locally without pointless failures. I don't understand how we've ended up with a test that will fail by default when run outside the CI.

A workaround could also be to remove the prev releases folder

feature_backwards_compatibility.py and wallet_upgradewallet.py are skipped when the /releases folder is absent. That folder is not in the repo. Once the folder is present, the test assumes you intend to use it and will fail if anything is missing. Unless you explicitly turn off TEST_PREVIOUS_RELEASES.

Having to always set TEST_PREVIOUS_RELEASES=true is annoying.

Not sure why we are doing bash env var juggling here. This should be an option on the test framework like everything else. E.g. --factor

It mimics the behaviour of is_wallet_compiled and is_zmq_compiled (I'm also not a fan of the config.ini juggling involved there). But maybe we can use a --previous_releases option instead of the env var, as long as the default depends on whether /releases is present.

Closing in favour of #19014.

Just to make sure, #19014 doesn't change any behaviour. You will still run into failures if the prev releases dir exists but is missing the binaries. The only way to fix that is to remove the dir or add the binaries.