Explicitly reject routes which double back

[TheBlueMatt]

I managed to generate a route which went out over an SCID and came right back over the same SCID, which nodes rejected, but really should have been rejected by LDK prior to sending over it.

We might also consider rejecting such forwards (I think we allow them?) but there's not a ton of reason to.

[TheBlueMatt]

Should be easy to implement - just add another check pre-send.

[tnull]

Wait, how is this even possible? Even with the smallest edge weight, shouldn't we always produce loopless routes?

[TheBlueMatt]

I believe I managed to do it while building a naive rebalancing router which included my own channel as both a first-hop and last-hop hint.

[Sharmalm]

Hey can i try this issue?
I believe we have to add check pre-send in router:: get_route .

[tnull]

Hmm, I think if it was really due to the same hop being included multiple times, this should have been fixed by #2575/c480b0c ?

[tnull]

Closing this as it likely was completed by #2575. @TheBlueMatt feel free to reopen if you think this was caused by something else.

[TheBlueMatt]

I dont think our router will do this, but I did it manually in some broken code I was writing. I was thinking here that we should check for this in outbound_payments.rs as a precondition for sending.