Closed mkcp closed 6 years ago
Oh yeah, that nil index thing is definitely worrying. We assign indices to ALL ops in the history, right? How did it get one without... an index?
Might be worth sprinkling some (assert (:index op)) through that checker to see if you can trace it back to the source?
I expected this to be done, but for some reason all of the knossos.report-tests are passing while it's failing to generate any linear.svg against etcd. Need a bit more time to figure this out
Both
knossos.linear
andknossos.wgl
render debugging output with the same ops' indices as they are passed in. I also updated linear to return plain maps, same as WGL, and added some cases from WGL to linear-test.One change that I'm suspect of is that I needed to add a check for
:index nil
inknossos.linear/step
so that report-test would stopNullPointerException
ing. As far as I can tell this doesn't change the behavior for linear outside of report, so I'm puzzled as to why this is happening. https://github.com/jepsen-io/knossos/compare/master...mkcp:support-jepsen-indices?expand=1#diff-8f4c4f892cd240e9882c1d2b666c1a35R211