Kyle Kingsbury

This might just be a bad history which needs to be re-written; looks like there's no invoke op for that particular ok.

Yeah, I think you'd need to build a different kind of model that represents the sorts of operations you'd like to perform on your filesystem. I'm not sure what that...

This sounds like you might have changed the version of Jepsen this test uses; there have been a few API changes in the last three years. Changing checker arities was...

This miiight be caused by error recovery code in Jepsen which is trying to help you escape a deadlock caused by some exceptional condition in the generator itself, but it...

This might fiiiiinally be fixed in 0.1.19 with the introduction of pure generators. It'll take some changes to the test suite though; see jepsen.generator.pure's namespace docs for details.

Ah, yeah, I think you're right. Since you have a failing case handy, would you like to submit a PR?

I think they ought to work--the generator APIs are almost all compatible. Did you have problems with any of them?

Huh! So far nobody's done this, I think because most folks aren't running distributed databases on OS X, but it shouldn't be hard to add. See https://github.com/jepsen-io/jepsen/blob/master/jepsen/src/jepsen/net.clj for examples of...

Hi itamar, and thank you. It sounds like your history might be malformed; a process should never complete an operation without having invoked it first. I think knossos probably should...

I'm really not sure! This could be a bug in complete-fold-op.... Might be worth dropping some logging statements in there and watching the state machine evolve.Also note that this comes...