p2p functional test failure

[selsta]

https://github.com/monero-project/monero/actions/runs/12969219043/job/36173269466?pr=9723#step:11:1380

[TEST STARTED] p2p
Resetting blockchain
Creating wallet
Generating 80 blocks
Testing P2P reorg
Testing P2P tx propagation
Traceback (most recent call last):
  File "/home/runner/work/monero/monero/tests/functional_tests/p2p.py", line 187, in <module>
    P2PTest().run_test()
  File "/home/runner/work/monero/monero/tests/functional_tests/p2p.py", line 46, in run_test
    self.test_p2p_tx_propagation()
  File "/home/runner/work/monero/monero/tests/functional_tests/p2p.py", line 182, in test_p2p_tx_propagation
    assert len(res.tx_hashes) == 1
  File "/home/runner/work/monero/monero/utils/python-rpc/framework/rpc.py", line 50, in __getattr__
    return self[key]
KeyError: 'tx_hashes'
[TEST FAILED] p2p

Before we can put out v0.18.4.0 we have investigate this to make sure this is not a regression with recent merges. The test failure does not show up consistently.

[selsta]

@vtnerd could this be related to #9459?

[vtnerd]

A quick question before I dig deeper: have you seen this on the release-v0.18 branch or just the master branch? The linked CI build was for master so I'm testing against that for now.

[selsta]

It did not trigger on release-v0.18 yet. It generally only triggered once so far, seems to be a rare bug.

[vtnerd]

I haven't gotten this to trigger locally, so I have no daemon logs to investigate. And I don't see a way to download the daemon logs from Github.

My best guess is that this is a timing bug in the test itself. The sleep timeout is 5 seconds, but the fluff (incoming) timer for transactions is typically going to be in the 3-7.25 second range. The counter argument is that it seems like the test should fail more frequently if this were the culprit.

I don't have any other theories as to the cause at the moment. #9459 definitely seems unrelated, unless there was a p2p connection loss + reconnect just before test_p2p_tx_propagation ran. Otherwise, if #9459 were responsible, there should've been an error earlier in the test functions.

[selsta]

Thank you for looking into it. If we can't reproduce it locally and if doesn't show up again on CI then not sure if we can do anything else.

[iamamyth]

I think the test can be improved quite a bit to avoid such spurious failures.

[iamamyth]

See #9762