The ratio of the median of the fees and probability cost is overall not
a bad factor to combine these two features. This is what the
test_real_data shows.
Changelog-None
Signed-off-by: Lagrang3 <lagrang3@protonmail.com>
The fee_fallback test would fail after fixing the computation of the
median. Now by we can restore it by making the probability cost factor
1000x higher than the ratio of the median. This shows how hard it is to
combine fee and probability costs and why is the current approach so
fragile.
Changelog-None
Signed-off-by: Lagrang3 <lagrang3@protonmail.com>
Rusty: "We don't generally use NDEBUG in our code"
Instead use a compile time flag ASKRENE_UNITTEST to make checks on unit
tests that we don't normally need on release code.
Changelog-none
Signed-off-by: Lagrang3 <lagrang3@protonmail.com>
- use graph_max_num_arcs/nodes instead of tal_count in bound checks,
- don't use ccan/lqueue, use instead a minimalistic queue
implementation with an array,
- add missing const qualifiers to temporary tal allocators,
- check preconditions with assert,
- remove inline specifier for static functions,
Changelog-None
Signed-off-by: Lagrang3 <lagrang3@protonmail.com>
The calculation of the median values of probability and fee cost in the
linear approximation had a bug by counting on non-existing arcs.
Changelog-none: askrene: fix the median
Signed-off-by: Lagrang3 <lagrang3@protonmail.com>
We use an arc "array" in the graph structure, but not all arc indexes
correspond to real topological arcs. We must be careful when iterating
through all arcs, and check if they are enabled before making operations
on them.
Changelog-None: askrene: fix bug, not all arcs exists
Signed-off-by: Lagrang3 <lagrang3@protonmail.com>
Add a new function to compute a MCF using a more general description of
the problem. I call it mcf_refinement because it can start with a
feasible flow (though this is not necessary) and adapt it to achieve
optimality.
Changelog-None: askrene: add a MCF refinement
Signed-off-by: Lagrang3 <lagrang3@protonmail.com>
It is just a copy-paste of "dijkstra" but the name
implies what it actually is. Not an implementation of minimum cost path
Dijkstra algorithm, but a helper data structure.
I keep the old "dijkstra.h/c" files for the moment to avoid breaking the
current code.
Changelog-EXPERIMENTAL: askrene: add priorityqueue
Signed-off-by: Lagrang3 <lagrang3@protonmail.com>
As we can see from the previous test, l3 tells us why it rejected the payment:
```
lightningd-3 2024-11-19T03:56:27.151Z DEBUG 022d223620a359a47ff7f7ac447c85c46c923da53389221a0054c11c1e3ca31d59-chan#1: Failing HTLC because of an invalid payload (TLV 10 pos 104): cltv_expiry 136 > payment_constraint 130
```
It turns out, we were not updating the block height in the plugin!
Without this, when we create a (non-dummy) blinded path we set a
too-low CLTV restriction, and it doesn't work after a few blocks!
Note we were actually triggering this error in the xpay tests!
Reported-by: Vincenzo Palazzo
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Changelog-Fixed: Offers: Receiving bolt12 payments where we have no public channels would fail a few blocks after startup.
Payer metadata is a field that controls the payer ID
provided during the fetchinvoice process.
There are use cases where this is highly useful, such as
proving that the payer has paid for the correct item.
Imagine visiting a merchant's website to pay for multiple offers, where
one of these offers is a default offer (with no description and no set amount).
In this scenario, the merchant could claim not to have received
payment for a specific item. Since the same offer may be used to
fetch invoices for different products, there needs to be a way to
identify which product the invoice corresponds to.
With this commit, it will be possible to inject payer metadata,
which helps solve the issue described above.
For example, possible payer metadata could be `to_hex(b"{payer_node_id}.{product_id}.{created_at}")`.
Changelog-Added: JSON-RPC: `fetchinvoice` allows setting invreq_metadata via `payer_metadata` parameter.
Signed-off-by: Vincenzo Palazzo <vincenzopalazzodev@gmail.com>
This is required for VLS which wants to know (and potentially decline) invoices
we're trying to pay.
As a nice side effect, our "check" command for xpay now does much more thorough
checking of arguments.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
As the first user of a persistent layer, this tripped tests which
assumed the datastore would be empty!
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Note: won't work with grpc (or probably other tools), since the output
is different. But good for testing.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Changelog-Added: Config: option `xpay-handle-pay` can be used to call xpay when pay is used in many cases (but output is different from pay!)
Because we initalized plugin->io_rpc_conn *after* calling plugin->init,
send_outreq would do a (harmless, in our case) wakeup on an uninitialized address:
```
==1164079== Conditional jump or move depends on uninitialised value(s)
==1164079== at 0x1628FC: backend_wake (poll.c:227)
==1164079== by 0x160B98: io_wake (io.c:384)
==1164079== by 0x1160A8: ld_rpc_send (libplugin.c:255)
==1164079== by 0x1187E0: send_outreq (libplugin.c:1099)
==1164079== by 0x115041: init (xpay.c:1620)
```
Solution is simple: set plugin->io_rpc_conn to NULL, and don't wake it in this case.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
The `send_outreq` function is a good place to suspend and resume
traces, since these are usually the places where we hand off control
back to the `io_loop`. This assumes that we do not continue doing
heavy liftin after we have queued an `outreq` call, but that is most
likely the case anyway. This frees us from having to track suspensions
whenever we call the RPC from a plugin.
And use it for `exposesecret-passphrase`. This is probably overly
cautious, but it makes me feel a little better that we won't leak it
to someone with read-only access.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Being able to back up the hsm_secret is critical, but you cannot do
this through a UI, because of course we do not allow such access.
People have lost funds because they didn't back up.
This allows access to the hsm_secret if you use a password set in the
config file. (If it's not set, the command does not work). This is a
compromise, of course.
Changelog-Added: `exposesecret` command for encouraging hsm_secret backups.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This means we don't consume *all* the CPU.
Changelog-Fixed: Plugins: `autoclean` is now gentler on the node when doing giant cleans.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
On a large node, especially with postgres, this causes every other command
to take 30 seconds plus. The first, obvious, step is to reduce how many
commands we will do at once.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Given an {outpoint}, sets the description on the matching outpoint (if exists).
Note that if no outpoint exists in bookkeeper, will return an empty list
Changleog-Added: PLUGINS: bookkeeper has a new RPC `bkrp-editdescriptionbyoutpoint` which will set/update a description for an outpoint creation event.
This takes an {payment_id} and {description}.
It looks for all chain + channel events that match
that {payment_id} and updates the description for those events.
We return all the updated events. If no events are updated, an empty
list is returned.
Changelog-Added: PLUGINS: bookkeeper has a new RPC `bkpr-editdescriptionbypaymentid` which will update the description for any event with matching payment_id
