Since CT can mean both Cipher Text as well as Clear Text, the variable
name scriptCT lead to confusion about its meaning. We rename it to make
it more clear how it's used.
In this commit, we fix the peer selection for the pruned block
dispatcher. Turns out that the bits are actually mutually exclusive: a
node can set the network bit _and_ the limited (pruned bit).
The prior logic would only seek out "pure" nodes, meaning older nodes
that wouldn't set the limited bit when they're a full-full node. We fix
this by ignoring the limited bit, and we'll only look at the network and
segwit bits.
A unit test has also been added to show the issue in the prior filter
function.
Fixeslightningnetwork/lnd#5864.
The loadAccountInfo does load an account as its name suggests. But after
loading it, the account is also added to a map to cache it. That map
write should be seen as a writing operation and therefore the _write_
lock must be held, not just the read lock.
If SendOutputs is called on a watch-only wallet then a transaction is
created but without any signatures. Publishing that transaction will
always fail. But the attempt to publish will already update the internal
state of the wallet so we shouldn't try to publish this unsigned TX.
Instead we return a new error along with the unsigned transaction to
give the caller the chance to sign and publish it through other means.
If we're running in watch-only mode, there is no unlock possible.
Therefore, we also don't need to prevent any unlocks from happening when
doing coin selection in that mode.
In this commit, we add a new method `DeriveFromKeyPathCache` that gives
callers a way to more quickly obtain a private key they know they'll be
using frequently. This method lets a caller avoid the write database
transaction as well as the EC operations to derive the key itself (BIP
32).
To allow a wallet to be created directly from an extended master root
key (xprv), we move the derivation from seed to extended key to the
loader instead of the address manager itself.
We can now get rid of our incorrect dust calculation which did not
give exact values for segwit outputs as it was based on spending a
P2PKH output instead.
We want to make sure that every database backend has their own
implementation of the View() and Update() methods to make sure any
custom retry/commit/rollback strategies are used even when invoked
through the package-provided View() and Update() functions.
To make sure we don't create any manual DB transactions, we refactor the
txToOutputs method to use walletdb.Update and the new
walletdb.ErrDryRunRollBack error for making sure a rollback is issued.
To make sure we don't create any manual DB transactions, we refactor the
ImportAccountDryRun method to use walletdb.Update and the new
walletdb.ErrDryRunRollBack error for making sure a rollback is issued.
We need a way to signal that a DB transaction should be rolled back
because its changes were a dry-run only. We add a new specific error for
that case so the automatic rollback on error in the walletdb.View and
walletdb.Update comes to play. But we can still return the result of a
dry-run where we need to by inspecting for this specific error.
Let the method `addRelevantTx` use `InsertTxCheckIfExists` to insert a relevant
transaction. If the transaction has already been recorded, the method call
`addRelevantTx` will return early and not proceed with duplicating work of
checking every output and duplicating the tx notification to the notification
server.
Expired leases are kept in the database until the next cleaning round.
This commit makes sure that expired leases look like they do not exist
to outside callers.
The submodules wallet/txauthor and wallet/txsizes were updated without a
new tag being created. This works fine when compiling btcwallet from
source but fails when using "go get" as that ignores the replace
directives.
The wallet loader has a dependency to the internal/prompt package for
prompting the user for certain inputs (e.g. wallet password or new
seed).
This makes it impossible for projects that use the wallet as a
dependency and always provide those inputs as parameters to compile for
JavaScript/WebAssembly targets because the prompt code uses some
terminal functionality that is not available in JS syscalls.
By providing a JS specific implementation that just returns an error we
can compile the dependent projects.
Adding acutal support for prompting the user in the browser is currently
not planned as that can easily be circumvented by providing all inputs
as parameters.
This method returns the first N external and internal addresses, which
can be presented to users to confirm whether the account has been
imported correctly.
