If a node is announced, prefer using a one-hop blinded path with it as
the introduction node to using a two-hop blinded path with a Tor-only
introduction node. The one-hop blinded path is more reliable, thus only
use Tor-only nodes if the recipient is unannounced. And then, prefer
non-Tor-only nodes.
When forming blinded paths, using a reliable node as the introduction
node is important to ensure onion message reliability. Order blinded
paths by how well-connected the introduction node is as a proxy for
reliability.
For short paths (e.g., when the sender and recipient share an LSP), this
may also result in a scenario where initiating a direct connection isn't
necessary. That is helpful when using RGS since it currently doesn't
include node announcements and thus cannot initiate a direct connection.
Tor nodes can have high latency which can have a detrimental effect on
onion message reliability. Prefer using nodes that aren't Tor-only when
creating blinded paths both in offers and in onion message reply paths.
Add a method to NodeInfo to determine if the node has only announced Tor
addresses. Useful for preferring blinded paths that don't use Tor for
better reliability and improved latency.
OfferBuilder is not exported to bindings because it has methods that
take `self` by value and are only implemented for certain type
parameterizations. Define these methods using macros such that different
builders and related methods can be defined for c_bindings.
.. returning `PeerDetails` rather than tuples of peer-associated values.
Previously, we wouldn't offer any way to retrieve the features a
peer provided in their `Init` message.
Here, we allow to retrieve them via a new `PeerDetails` struct,
side-by-side with `SocketAddress`es and a bool indicating the direction
of the peer connection.
This advisory is only relevant for a downstream dependency of
`criterion`, which we currently don't want to bump in order to continue
benchmarking with our MSRV 1.63.0.
We therefore just add it to our ignore list for now.
https://github.com/tkaitchuck/aHash/pull/196 bumped the MSRV of
`ahash` in a patch release, which makes it rather difficult for us
to have it as a dependency.
Further, it seems that `ahash` hasn't been particularly robust in
the past, notably
https://github.com/tkaitchuck/aHash/issues/163 and
https://github.com/tkaitchuck/aHash/issues/166.
Luckily, `core` provides `SipHasher` even on no-std (sadly its
SipHash-2-4 unlike the SipHash-1-3 used by the `DefaultHasher` in
`std`). Thus, we drop the `ahash` dependency entirely here and
simply wrap `SipHasher` for our `no-std` HashMaps.
In the next commit we'll drop the `ahash` dependency in favor of
directly calling `getrandom` to seed our hash tables. However,
we'd like to depend on `getrandom` only on certain platforms *and*
only when certain features (no-std) are set.
This introduces an indirection crate to do so, allowing us to
depend on it only when `no-std` is set but only depending on
`getrandom` on platforms which it supports.
When the `max_total_routing_fee_msat` parameter was added to
`RouteParameters`, the serialization used `map` to get the max fee,
accidentally writing an `Option<Option<u64>>`, but then read it as
an `Option<u64>`. Thus, any `Route`s with a `route_params` written
will fail to be read back.
Luckily, this is an incredibly rarely-used bit of code, so only one
user managed to hit it.
`Route`'s blinded_path serialization logic writes a blinded path
`Option` per path hop, however on read we (correctly) only read one
blinded path `Option` per path. This causes serialization of
`Route`s with blinded paths to fail to round-trip.
Here we fix this by writing blinded paths per path.
`fails_paying_for_bolt12_invoice` tests that we fail to send a
payment if the router returns `Ok` but includes a bogus route (one
with 0-length paths). While this marginally increases our test
coverage, in the next commit we'll be testing that all routes
round-trip serialization, which fails here as bogus routes are not
supported in deserialization.
Because this isn't particularly critical test coverage, we simply
opt to drop the test entirely here.
When an `std::future::Future` is `poll()`ed, we're only supposed to
use the latest `Waker` provided. However, we currently push an
`StdWaker` onto our callback list every time `poll` is called,
waking every `Waker` but also using more and more memory until the
`Future` itself is woken.
Here we fix this by removing any `StdWaker`s stored for a given
`Future` when it is `drop`ped or prior to pushing a new `StdWaker`
onto the list when `poll`ed.
Sadly, the introduction of a `Drop` impl for `Future` means we
can't trivially destructure the struct any longer, causing a few
methods to need to take `Future`s by reference rather than
ownership and `clone` a few `Arc`s.
Fixes#2874
When an `std::future::Future` is `poll()`ed, we're only supposed to
use the latest `Waker` provided. However, we currently push an
`StdWaker` onto our callback list every time `poll` is called,
waking every `Waker` but also using more and more memory until the
`Future` itself is woken.
Here we take a step towards fixing this by giving each `Future` a
unique index and storing which `Future` an `StdWaker` came from in
the callback list. This sets us up to deduplicate `StdWaker`s by
`Future`s in the next commit.
In the next commit we'll fix a memory leak due to keeping too many
`std::task::Waker` callbacks in `FutureState` from redundant `poll`
calls, but first we need to split handling of `StdWaker`-based
future wake callbacks from normal ones, which we do here.
We add the previously discussed `rustfmt.toml` and enforce it in CI for
any files that are not contained in an exclusion list.
To start, we add all current Rust files to this exclusion list. This
means that formatter rules will be enforced for any newly introduced
files, and we'll then start going through the codebase file-by-file,
removing them from the list as we go.
