* Announce blocks as HEADERS instead of INV in comptool
Announcing as INV results in one GETHEADERS message per INV item with
every GETHEADERS containing the same block locator info. This in turn
results in many (1088 in p2p-fullblocktests) replies with each having i+1
headers. As every message is sent in one go, the message processing queue
gets overloaded, leading to GETDATA timeouts for blocks and thus failing
the tests.
Announcing the blocks directly with HEADERS avoids all this and results in
one GETDATA per announced header.
This will later conflict with backported improvements from Bitcoin. When we
get to this point, we can simply ignore changes on Dash's side and take
the improvements from Bitcoin (which also switch to header announcement,
but after some refactorings)
* Give wait_for_pings more time
Large reorgs as in the p2p-fullblocktest can cause very long long delays
for the ping that we're waiting for. This was not a problem before #2593,
as before this wait_until was waiting forever.
037159c Remove block-request logic from INV message processing (Matt Corallo)
3451203 [qa] Respond to getheaders and do not assume a getdata on inv (Matt Corallo)
d768f15 [qa] Make comptool push blocks instead of relying on inv-fetch (mrbandrews)
Ever since we #5913 have been sending invalid reject messages
for transactions and blocks.
test: Add basic test for `reject` code
Extend P2P test framework to make it possible to expect reject
codes for transactions and blocks.
Github-Pull: #7179
Rebased-From: 9fc6ed600320411903d7
Remove necessity to call create_callback_map (as well as the function
itself) from the Python P2P test framework. Invoke the appropriate
methods directly.
- Easy to forget to call it and wonder why it doesn't work
- Simplifies the code
- This makes it easier to handle new messages in subclasses
Github-Pull: #7171
Rebased-From: 2f601d215d
