Given we build `InFlightHtlcs` per route-fetch call, there's no
reason to pass them out by reference rather than simply giving the
user the full object. This also allows them to tweak the in-flight
set before fetching a route.
We already hold them in a vec, so there's no cost to passing them
by ownership vs making it a slice. Further, this helps bindings as
we can't represent slices to non-pointers in a sensible way.
In bindings we can't practically pass a mutable transaction, and
instead need to pass an owned transaction and have the sign method
return a signed copy. We do this here for all build modes as the
API is roughly equivalent also to Rust users.
We introduce a `UnfundedChannelContext` which contains a counter for the
current age of an unfunded channel in timer ticks. This age is incremented
for every `ChannelManager::timer_tick_ocurred` and the unfunded channel
is removed if it exceeds `UNFUNDED_CHANNEL_AGE_LIMIT_TICKS`.
The value will not be persisted as unfunded channels themselves are not
persisted.
One of a series of follow-up commits to address some issues found
in PR 2077, where we split channels up into different maps and structs
depending on phase in their life.
We had some inconsistencies so far in referring to channels such as
`OutboundV1Channel` and `InboundV1Channel` as pending and unfunded.
From here we refer to these kinds of channels only as "unfunded".
This is a slight conflation with the term "unfunded" in the contexts
of denial of service mitigation. There, "unfunded" actually refers to
non-0conf, inbound channels that have not had their funding transaction
confirmed. This might warrant changing that usage to "unconfirmed inbound".
The C bindings generation currently has issues looking through a
generic associated type. While this should be fixed in the bindings
generator, its easy to fix here for now and we can revisit it
later.
The `RefCell` was necessary in a previous iteration of the code in which
the iterator was not `Clone` so we needed interior mutability in order
to consume the iterator. Now that it is `Clone`, we can drop it, as
we're no longer mutating the original iterator.
As done with inbound feerate updates, we can afford to commit less in
fees, as long as we still may the minimum mempool feerate. This enables
users to spend a bit more of their balance, as less funds are being
committed to transaction fees.
Channels supporting anchors outputs no longer require their feerate
updates to target a prompt confirmation since commitment transactions
can be bumped when broadcasting. Commitment transactions must now at
least meet the minimum mempool feerate, until package relay is deployed,
such that they can propagate across node mempools in the network by
themselves.
The existing higher bound no longer applies to channels supporting
anchor outputs since their HTLC transactions don't have any fees
committed, which directly impact the available balance users can send.
Now that we support channels with anchor outputs, we add a new
ConfirmationTarget variant that, for now, will only apply to such
channels. This new variant should target estimating the minimum feerate
required to be accepted into most node mempools across the network.
With anchors, we've yet to change the frequency or aggressiveness of
feerate updates, so it's likely that commitment transactions have a
good enough feerate to confirm on its own. In any case, when producing a
child anchor transaction, we should already take into account the fees
paid by the commitment transaction itself, allowing the user to save
some satoshis. Unfortunately, in its current form, this will still
result in overpaying by a small margin at the expense of making the coin
selection API more complex.
There's no need to yield such an event when the commitment transaction
already meets the target feerate on its own, so we can simply broadcast
it without an anchor child transaction. This may be a common occurrence
until we are less aggressive about feerate updates.
In an older PR a reviewer had asked why the discarding of a channel
being blocked on another monitor update is okay if the blocked
channel has since closed. At the time, this was not actually okay -
the monitor updates in the channel weren't moved to the
`ChannelManager` on close so the whole pipeline was busted, but
with the changes in 4041f0899f the
handling of channel closes with pending monitor updates is now
correct, and so is the existing code block.
If a `ChannelMonitorUpdate` completes being persisted, but the
`ChannelManager` isn't informed thereof (or isn't persisted) before
shutdown, on startup we may still have it listed as in-flight. When
we compare the available `ChannelMonitor` with the in-flight set,
we'll notice it completed and remove it, however this may leave
some post-update actions dangling which need to complete.
Here we handle this with a new `BackgroundEvent` indicating we need
to handle any post-update action(s) for a given channel.
Historically, we used `vec_type` for all TLV Vec reads/writes, but
it is asymmetric and thus somewhat confusing - on the write side it
always writes a TLV entry, even if there are zero elements. On the
read side, it happily accepts a missing TLV, providing a
zero-length vector.
In 85b573ddad a new `optional_vec`
TLV format was added which was symmetric, but only supports
optional vecs.
Now that we've migrated entirely to the new `required_vec` TLV
type, we can entirely remove the awkward `vec_type`.
