fix(legacy_swap): check for existing maker/taker payment before timeout

[shamardy]

This PR moves payment existence check in maker_payment/send_taker_payment before timeout validation and skips timeout if payment is already sent, as the taker swap should proceed to waiting for maker to spend the taker payment.

Should fix #2136 , but we are still not aware why the electrum request doesn't timeout.

[laruh]

Thanks for the fix!
Changes are well-structured and code in general is good to me, I have non mandatory suggestion

[dimxy]

So this actually fixes the situation (for taker case):
Taker sent the payment but logged no event for that. Then went offline for a long time. When Taker restarts he finds the timeout has gone and fails the swap in a state in which he does not refund (but he can refund)?

[cipig]

When Taker restarts he finds the timeout has gone and fails the swap in a state in which he does not refund (but he can refund)?

not sure if i understood the question, but taker can only refund if he removes the TakerPaymentTransactionFailed event from his swap JSON

      {
         "event" : {
            "data" : {
               "error" : "taker_swap:1488] Timeout 1717611160 > 1717610301"
            },
            "type" : "TakerPaymentTransactionFailed"
         },
         "timestamp" : 1717611160346
      },

and replaces it with a TakerPaymentSent event and also adds the correct txid and txhash from that particular swap to the event... like this

      {
         "event" : {
            "data" : {
               "tx_hash" : "84fa6b8d50aff9a727c61b8f1bf33564c8958d120669da33fa385d45f90f6da0",
               "tx_hex" : "0400008085202f8901a33d00490a3d9e3cbccbb394013c91d4f80dc1f2a19191632a29747d2f439301010000006b48304502210094e97cb9999aa77357f29f23a8e0bf778dbbf238523a92875992b7ea8af303e502204e1554dd35d6cbdebe3f5780609163473aa9b918b7213bfbe018dd672a9a12cd0121028a38bafdd6c1aa3ec34c480d6ae9383bee54eef7c115e51a5aeab699734ff055ffffffff03008589dc0000000017a914ea8dcc7b4ea177f7a2c999d7a603aab8d6308b2f870000000000000000166a14718d7a526e8d14f45b5795db8b864924b0cd90c6d2bee2f60d0000001976a914d89058fbcce7402fb342dffadf04bd32dba29abe88ac89a16066000000000000000000000000000000"
            },
            "type" : "TakerPaymentSent"
         },
         "timestamp" : 1717611160346
      },

this was the situation from before this change...

[dimxy]

I am trying to work out how taker came to situation that he actually sent taker payment but did not log any event about this (to memorise this, it's not clear from the issue description)

[cipig]

I am trying to work out how taker came to situation that he actually sent taker payment but did not log any event about this (to memorise this, it's not clear from the issue description)

That's the big question, indeed. I first thought the electrums returned a timeout, but sent the tx anyway. But that's not the case, else we would have an error from electrum there in the event and not simply taker_swap:1488] Timeout 1717611160 > 1717610301. Will show the affected user this, maybe he can tell us more about the "circumstances".

[mariocynicys]

That's the big question, indeed. I first thought the electrums returned a timeout, but sent the tx anyway. But that's not the case, else we would have an error from electrum there in the event and not simply taker_swap:1488] Timeout 1717611160 > 1717610301. Will show the affected user this, maybe he can tell us more about the "circumstances".

if the process is killed after this line

komodo-defi-framework/mm2src/mm2_main/src/lp_swap/taker_swap.rs

Line 1581 in f401497

.send_taker_payment(SendPaymentArgs {

but before the swap event is written, this could happen.

[mariocynicys]

tiny notes. LGTM otherwise

[shamardy]

@mariocynicys @laruh please review the refactor next week.

[dimxy]

Should we not allow this test locktime only if enabled by a cfg feature?
(I guess we tend to make all test code cfg enabled)

[shamardy]

Removed USE_TEST_LOCKTIME occurrences in swaps code completely, and from the whole codebase. Took me some time to make the tests work again but this should make the legacy swap code cleaner. We should look into removing all other env vars in swaps code when we have the time.

[dimxy]

I added a note why the swap failed on send taker payment and manual swap file editing was required.
This PR fixes this issue: even if TakerPaymentSent was not saved in the file, the fix checks that the payment tx in fact was sent and the swap continues.

(PS. I guess, there may be similar issues in other swap steps due to the event not saved)

LGTM

[laruh]

LGTM!
I have last non blocking suggestion

[laruh]

I suppose its better to name it setup_taker_watcher_reward, to follow watcher reward setup naming on maker_swap.rs

and then rename process_watcher_logic to process_taker_watcher_logic

Or just rename setup_maker_watcher_reward to setup_watcher_reward, its the private function anyway