< GitHub19>
[bitcoin] rebroad opened pull request #7651: Only need to subtract 1 (originally time was in seconds). (master...AskforTime) https://github.com/bitcoin/bitcoin/pull/7651
< GitHub63>
[bitcoin] Lewuathe opened pull request #7653: Suppress unused variable warning in automated test (master...suppress-unused-variable-warn) https://github.com/bitcoin/bitcoin/pull/7653
< wumpus>
this logic in CFeeRate::GetFee is giving me a headache: https://github.com/bitcoin/bitcoin/blob/master/src/amount.cpp#L24 so it computes the fee from the fee/kb and the size of the transaction, so far so good. But why is there a check to set the fee to nSatoshisPerK if the resulting fee is 0?
< wumpus>
is this a kind of 'rounding'?
< sdaftuar>
wumpus: looks like it showed up in commit d88af560111863c3e9c1ae855dcc287f04dffb02
< sdaftuar>
PR 4465
< wumpus>
"fee fixes". Ominous.
< wumpus>
rationale seems to be "CFeeRate::GetFee(..) currently returns zero for very small nSatoshisPerK. For example, if you set the voluntary fee to 1 satoshi, and the tx has less than 1000 bytes, payTxFee.GetFee(nTxBytes) returns zero, this results in CWallet::GetMinimumFee(..) to return pool.estimateFee(..) instead of the voluntary fee. So its not possible to set the voluntary fee as low as 1 satoshi."
< wumpus>
interesting. Thanks for looking it up. Bah, such things should really be commented in the source code.
< wumpus>
seems like a kind of layering violation to do it there, too, as CFeeRate is used not just in the wallet
< wumpus>
in any case it isn't a big issue, just makes me a bit afraid that the fee code is full of such snags, I don't blame users for getting confused :)
< sdaftuar>
agreed, confusing and scary :)
< kanzure>
re: my question about dry run configuration with libbtcnet, the answer to my question was "yes"-- i noticed later that there was an explicit start() method.
< kanzure>
so the connection configuration happens separately from connection initiation, and could be reviewed or analyzed or inspected prior to establishing active connections (yay)
< morcos>
and then i think there will need to be some special casing allowing block version <= 4 (indefinitely? or do we want to maybe soft fork out the ability to have block version = 4 blocks once the first version bits soft fork activates?)