It is not possible to switch to py3 without fixing those bugs or doing the refactor. (E.g. python2 will silently cast floats to ints where needed, but I'd rather explicitly use ints)
To aid review, each commit in this pull will either fix bugs or do some refactor.
Extended tests pass and enabling python deprecation warnings only give the following "false positive" warnings:
$ qa/pull-tester/rpc-tests.py -extended | grep -C 1 DeprecationWarning
/home/marco/workspace/bitcoin/qa/rpc-tests/test_framework/blocktools.py:49: DeprecationWarning: classic int division
halvings = int(height/150) # regtest
/home/marco/workspace/bitcoin/qa/rpc-tests/replace-by-fee.py:33: DeprecationWarning: classic long division
while node.getbalance() < satoshi_round((amount + fee)/COIN):
/home/marco/workspace/bitcoin/qa/rpc-tests/replace-by-fee.py:39: DeprecationWarning: classic long division
txid = node.sendtoaddress(new_addr, satoshi_round((amount+fee)/COIN))
/home/marco/workspace/bitcoin/qa/rpc-tests/replace-by-fee.py:396: DeprecationWarning: classic long division
split_value = int((initial_nValue-fee)/(MAX_REPLACEMENT_LIMIT+1))
201 | elif len(l) < 0x100000000L: 202 | - r = chr(254) + struct.pack("<I", len(l)) 203 | + r = struct.pack("B", 254) + struct.pack("<I", len(l)) 204 | else: 205 | - r = chr(255) + struct.pack("<Q", len(l)) 206 | + r = struct.pack("B", 255) + struct.pack("<Q", len(l))
What about:
r = struct.pack("<BQ", 255, len(l))
etc?
Sure, this makes sense. I've tried to make the diff easier to read but I will address you nit, as it makes the code smaller and possibly easier to read.
Thanks, yes I agree with regard to keeping the minimum diff, but indeed it's better to do this small cleanup (which someone is bound to do later) at the same time.
Tested ACK 4444806