This test was passing because we never fetch blocks if we only receive an inv and not the header (after 037159cebf1eae4445050cec029986514ed4e9e2), and this test wasn't delivering the header.
I noticed the test falsely passing when I intentionally broke this behavior in one of my branches.
announce_block() is only called from one place. I'd be tempted to remove the method entirely and explicitly call send_message(<msg_header>) from the test case so it's obvious to the reader what's happening.
Updated to remove announce_block().
1091 | # and then check to see if that particular getdata has been received. 1092 | - self.old_node.announce_block(block4, use_header=False) 1093 | + # Since 0.14, inv's will only be responded to with a getheaders, so send a header 1094 | + # to announce this block. 1095 | + msg = msg_headers() 1096 | + msg.headers = [ CBlockHeader(block4) ]
nit: spaces aren't required around list elements within brackets: https://www.python.org/dev/peps/pep-0008/#pet-peeves
I prefer the space there, otherwisethetextlookstoosquishedtome, so I'll leave this (until we merge a style guide for the project that calls for removing them).
Tested ACK c8e316bc3dd152a2d36a947195fe4b96ca4d34db
One small style nit. Up to you whether you take it or leave it.
This test was passing because we never fetch blocks if we only receive
an inv and not the header (after 037159cebf1eae4445050cec029986514ed4e9e2),
and this test wasn't delivering the header.
Rebased
utACK f19abd90537fe71e757fdbfc52cd2288e967bb6e