Closed thestick613 closed 4 years ago
What is your version that you compiled from source? can you share the version?
Latest from git:
"version": "v0.9.0-44-g01a82d3",
@thestick613 In my opinion, this looks like how an event out the normal flow, because you stop the node with the Ctrl + C during operation of payment splitting, so when you force the stop of your node, it can be inside an assignment operation that in the normal operation flow this assignment is running every time and inside the code don't need to check if this variable is null, but is interesting for me to understand what happened inside your node.
From your issue here I see that your listsendpay
work fine, you can share the output of this command? (If you can)
I want to understand what is the splitting status of payment
I didn't stop the node with Ctrl+C, only the lightning-cli pay
command, as it was running for more then 10 minutes, and the lightningd log stopped producing any output.
@thestick613 ops, my fault. Anyway, if you share your output or partial output I and someone here can try to reproduce what the lightningd daemon did and try to understand if there is some propriety with value undefined.
From your message here I understand that your split is very big.
Most of them reverted. I had over 100 pending htlcs, now there are only 7.
Maybe there is some condition that makes pay delay normal, I think that @cdecker know very well the command implementation and he can say better than me if there is some condition that made the pay command slow.
In addition, if you can share an example of your listsendpays
I can try to understand what happened inside the listpays
. Because the command listpays
It makes the compression of the mpp payments before return the response.
In addition, from the @ZmnSCPxj message inside the issue #3925
Note that ^C on lightning-cli will not cancel the closure
In my understanding, if you stop the lightning-cli interface, you did not stop the lightnind during the operation (in this case during the operation of pay). It mean that you could run your listpay message during the pay operation and it is wrong for the lighningd
normal flow because you can have the some payment incomplete.
In my opinion, this could be a possible answer at the your stack trace
For reference this is the code that is crashing (first line in the loop):
From this it seems that either root
or b11
is NULL
(strcmp
may crash if we pass it a NULL
). root
seems strange since we'd have to have added a NULL
to the list which would crash when adding afaik.
@thestick613 are you calling listpays
with a bolt11
to filter by?
Attempted an MPP payment, which stopped responding after 5 minutes, so i Ctrl+C the
pay
command. I then attempted to close a channel, which outputedcannot co-op close channel w/ active htlcs
in the log. Then i ranlistpays
to see the status of the MPP payments, which resulted in this crash: