bumpfee behavior with "Subtract fee from amount" #11122

When using the bumpfee command on a tx that was sent using "Subtract fee from amount", the current response is this error. "Transaction does not have a change output (code -1)". Since "Subtract fee from amount" was specified, the expected behavior would be to subtract the fee from the sent amount which would not require a change output.

bumpfee should be safe to use. Taking money from one of the recipients is not safe.

@MarcoFalke: I believe you have closed this issue without due consideration. When the user has sent a payment using subtractfeefromamount, they have already given explicit permission to deliver less than the requested amount to the requested recipient. Thus, it would not be unsafe for bumpfee to reduce the amount even further in this case. Besides, quite often when using subtractfeefromamount, the recipient is the same entity as the sender.

The wallet doesn't currently track if subtractfeefromamount was set on a tx. But I'll reopen the issue to see if someone is interested to work on this.

I would definitely like an option to subtract fee from the amount, e.g. to get stingy 1 sat/byte self-transfers unstuck. RPC would be the easiest place to start. GUI would require a warning that lowering the amount sent may cause the recipient (e.g. order not processed because they expected an exact amount).

cc @instagibbs @achow101

I think explicitly marking in the wallet a subtractfeefromaount, then allowing reduction of this later, is best UX? Less familiar with adding records to the wallet though.

That's certainly better UX. But the easiest to implement is additional option (and then in a followup make that default to true for such translations).

Well depends on the politics...

Concept ACK either way

On Sat, Jan 2, 2021, 4:28 PM Sjors Provoost notifications@github.com wrote:

That's certainly better UX. But the easiest to implement is additional option (and then in a followup make that default to true for such translations).

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/bitcoin/bitcoin/issues/11122#issuecomment-753447228, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABMAFUYUWVH3RWUDJNKPVM3SX3KLBANCNFSM4DYCWIVA .

@achow101 is this fixed by #22007 ?

Partially. The the new original_change_index can be used here, except when the original transaction didn't have change.