without partially backporting a new testframework.
- Adds a condition to NodeConn that when asyncore calls handle_read
without any data, this must be a disconnect and closes the socket
- Adds a little loop in the p2p-acceptblock client that waits for
the socket to be in a closed state
- Makes expected disconnects non-optional in p2p-acceptblock
- Syncs the test descriptions and outputs with reality
Implements, tests and assures wallet operator ease-of-sovereignty and making
sure that user-defined fee logic persists through versions in the future by
fixing and testing -paytxfee
1. There was an override of ANY user-defined value to CWallet::GetMinimumFee
- Former logic: always override any value with either -mintxfee or
-mintxrelayfee, whichever is highest
- Proposed logic in this pull request:
- if the user specifies a value, only override when it is lower than
-mintxfee or -mintxrelayfee - this works because we set any default
-mintxfee to be the same as -paytxfee, unless the user explicitly
sets a -mintxfee.
- if no value has been specified, use the rate from -mintxfee or
-mintxrelayfee, whichever is highest
2. Test that the interaction between the wallet parameters -paytxfee and
-mintxfee function as intended. This has to be done using rpc tests rather
than unit tests because it tests the actual parameters passed to the
executables.
3. Undoing the override exposed a misconfiguration in the bumpfee.py test,
where fees were explicitly set higher, yet ignored in subsequent bumps.
Test that the interaction between the wallet parameters
-paytxfee and -mintxfee function as intended. This has to
be done using rpc tests rather than unit tests because it
tests the actual parameters passed to the executables.
recreated seeds by selecting peers across 20+ nodes that:
1. Keep up with updates, in this case run 1.14.3 or higher
2. Have NODE_NETWORK set
3. Are not known scanning clients
4. Are connectable
5. Are stable - I ordered nodes by time they have been connected to
one of my nodes
resulting set is 120 good nodes for mainnet and 20 good nodes for
testnet, serialized into chainparamsseeds.h using the contrib/
scripts.
This test failed because the sheer number of blocks required to
trigger the max length of a fork we'd keep, exceed the v4 fork
height on regtest.
- Adapted the blocktools.py miner to mine 0x00620004 blocks as done
elsewhere too (be it suboptimal, but at least consistent)
- Adapted the test to work with 1440 blocks (Dogecoin limit)
instead of 288 (Bitcoin limit)
- Made p2p-acceptblock a standard test instead of an extended test
Use CAmount rather than unsigned int for amounts for consistency
with other fee rate amounts.
This does change the type from unsigned int to unsigned int64, and
while it is unlikely anyone would need a dust limit higher than
unsigned int, again this ensures the theoretical maximum is in line
with other rates.
Fix a rare crash bug where no best chain can be activated, and therefore when trying
to find the height of the best chain via the last block triggers a null pointer dereference.
Moves the break in ActivateBestChain() when a node is being shut
down from the beginning of the loop block to the end, allowing
pindexNewTip to be populated with chainActive.Tip() rather than
its initial NULL value. Solves issues when shutting down nodes
while inside the ActivateBestChain loop.
