If we know the final witness/scriptsig for an input we should add it
right away to the PSBT. Before, if we couldn't finalize any of them we
finalized none of them.
This allows specifying different "policy paths" for the internal and external
descriptors, and adds additional checks to make sure they are compatibile (i.e.
the timelocks are expressed in the same unit).
It's still suboptimal, since the `n_sequence`s are per-input and not per-transaction,
so it should be possibile to spend different inputs with different, otherwise
incompatible, `CSV` timelocks, but that requires a larger refactor that
can be done in a future patch.
This commit also tries to clarify how the "policy path" should be used by adding
a fairly detailed example to the docs.
Previously `send_all` was particularly confusing, because when used on a
`create_tx` it implied two things:
- spend everything that's in the wallet (if no utxos are specified)
- don't create a change output
But when used on a `bump_fee` it only meant to not add a change output
and instead reduce the only existing output to increase the fee.
This has now been split into two separate options that should hopefully
make it more clear to use, as described in #142.
Additionally, `TxBuilder` now has a "context", that basically allows to
make some flags available only when they are actually meaningful, either
for `create_tx` or `bump_fee`.
Closes#142.
This makes it possible to choose a UTXO manually without having to
choose them *all* manually. I introduced the `manually_selected_only`
option to enforce that only manually selected utxos can be used.
To stop the cli semantics changing I made the `utxos` keep the old
behaviour by calling `manually_selected_only`.
This commit upgrades `rust-miniscript` with a fix to only return the prefix that
matches a `hd_keypath` instead of the full derivation path, and then adapts the
signer code accordingly.
This commit closes#108 and #109.
Thanks to the `ToWalletDescriptor` trait we can also very easily validate the checksum
for descriptors that are loaded from strings, if they contain one. Fixes#20.
Instead of having two traits, `Blockchain` and `OnlineBlockchain` that need
to be implemented by the user, only the relevant one (`OnlineBlockchain`, here
renamed to `Blockchain`) will need to be implemented, since we provide a
blanket implementation for the "marker" trait (previously `Blockchain`, here
renamed to `BlockchainMarker`).
Users of the library will probably never need to implement `BlockchainMarker`
by itself, since we expose the `OfflineBlockchain` type that already does
that and should be good for any "offline" wallet. Still, it's exposed since
they might need to import it to define types with generics.
LLFourn
Justin Moon
Alekos Filini
Riccardo Casatta
Daniela Brozzoni
Steve Myers
Murch
Richard Ulrich