A Brooks

@kennyjwilli What about option 4, memoization? The downside is that if it does fail, memoization will keep trying the expensive operation. But if the expensive operation tends to fail early...

Looks like there's duplicates of 254, too. Here's what I found with a quick-and-dirty `ripgrep` search: ``` src/kaocha/api.clj:84: (throw+ {:kaocha/early-exit 254})))) src/kaocha/plugin.clj:48: (output/error-and-throw {:kaocha/early-exit 254} nil src/kaocha/plugin.clj:56: {:kaocha/early-exit 254} nil...

@NoahTheDuke I think I'll close this PR so you can open a separate one with the do change. Or are you planning to repurpose this PR?

Thanks for the report! I believe this the intended behavior of `#meta-merge` but applying `#kaocha/v1` to both is going to result in duplicates like you discovered. I'm not able to...

Oh okay. I think we can clarify that the `tests.user.edn` file should not itself include `#kaocha/v1`. We also have a meta-merge issue #184, so that's already covered.

Any thoughts on this? @plexus

For a real-world example, I think #190 describes a test with a similar problem, but I don't think the suggested fix would catch it, because we're building up the value...

That's true. I suppose we could catch this mistake with some spec'd functions because we could check if we're calling a function that is spec'd to only return truthy values....

We now have the `--diff-style :none` option, so you can turn it off. We could do more tweaking, probably, but I think this is a good solution. What do you...

If more people encounter this, we can reopen, but I think we're safe to close it.