Closed gavin-norman-sociomantic closed 6 years ago
Interesting... it seems that reading and checking the started / error code in the handleSupportedCode
method (instead of in the handle
method) fixes the flaky tests. I'm not entirely sure why.
Green now.
Interesting... it seems that reading and checking the started / error code in the handleSupportedCode method (instead of in the handle method) fixes the flaky tests. I'm not entirely sure why.
@nemanja-boric-sociomantic interested whether you have any thoughts on that. The most obvious thing I can think of is what will happen in AllNodesRequest.run
. I don't see what difference that would make, though.
I'm sceptical towards the fact that this caused/fixed flaky tests :thinking: .
Ditto.
It was very repeatable, though. With the previous change: ~2/4 builds would fail. With the current change: no builds fail.
How ironic that as soon as I type that, I notice that the current build has failed with the same problem :D
Moved flaky part to https://github.com/sociomantic-tsunami/dhtproto/pull/93.
:+1: Looks very noncontroversal.
Gah, it's too early in the year for flaky tests :(