There was a bug where the spending address index could have the same key
as the receiving address index if the input and output indexes matched. This lead
to the output always overwriting the input index leading to incorrect balances
with missing spent amounts. This patch separates the two so that they have unique
keys so balances will be correctly calculated.
generalize mininode.py protocol versioning
Fixes#2982 -- replace `overwintered` boolean argument to NodeConn constructor with a protocol version integer, so that adding Sapling (and beyond) will be easier.
Fix absurd fee bug reported in #3281, with tests
This was ported from the Hush `absurd_fee_bug` PR: https://github.com/MyHush/hush/pull/136
Tests executed with `./qa/pull-tester/rpc-tests.sh wallet` and pass on Ubuntu 16.04. Some tests were added for previously untested behavior as well.
Closes#3281.
Add change indicator
This adds to the json returned when calling z_listreceivedbyaddress and z_listuspent an additional field entitiled 'change' which will be either true or false depending on whether or not the note was change from another transaction.
(rpc-test) accurately account for fee without rounding error
Fix for #2807, this test compares balances after doing key exports and imports, and expects these balances to be equal. But they are not exactly equal due to transaction fees, so the test makes them "equal" by rounding a value that has had fees taken out up to the nearest hundredth of a unit (which is much more than the default fee). This obviously is somewhat sloppy. It also converts a balance to float, which really should never be used due to loss of precision (use Decimal instead).
This change makes the test accurately account for the fee using precise comparisons, and removes the use of float. This test doesn't depend on the default fee (0.0001) but instead sets the fee (to that value). This way, if the default fee changes in the future, this test will continue to run. While testing these changes, I set the fee to various values (up to the max, 0.0190), and the test still passes.