I’m brand new to Julia, just recently having installed it on two separate mac osx 10.14.1 machines with fresh installs of Julia 1.0.3 (one via mac app bundle and the other via homebrew), and going through basic tutorials everything seems fine. In both cases, however, packages StatsBase (v0.27.0) and Clustering (v0.12.1) report some broken tests when running pkg> test XXX (replace XXX with StatsBase or Clustering). I tried removing and re-adding the packages and also updating them, but no change. I’ve raised these issues on the respective package GitHub pages, which are part of JuliaStats, but was advised that there is no problem with these packages on their end and that I should bring the discussion here as these issues may reflect a problem in a more fundamental component of Julia, whatever that might mean. Although a few broken tests may not seem like huge deal, as a new user I’m highly concerned about migrating to Julia when what I consider fundamental stats packages work for the package maintainer but throw errors on my brand new install. Any help or advice would be greatly appreciated. The errors are:
JuliaStats: 1 broken test reported for Ambiguities
Clustering: 1 broken test reported for each of silhouettes, MCL and V-measure
Oh, I see. The testset is passing (that’s why it says “Testing Clustering tests passed” at the end).
You’re seeing the “Broken” tests. Those are specifically tests which have been marked as @test_broken. That is, they are known to not pass, and are just included in the testset by the developers as a reminder for future work. All software has issues, missing features, or edge cases, so the fact that there are some issues which the authors are aware of really doesn’t indicate anything being wrong with the package.
If you’re really interested, you could look for instances of @test_broken in the codebase and learn about what the known issues are. But the presence of known issues shouldn’t need to affect your confidence in the package.
I see. That’s good to hear. As a new user I’d recommend modifying the test reports to help avoid confusion. I did notice the “tests passed” summary at the bottom, but that didn’t make sense to me given the broken tests reported prior. I suspect that some sort of clarification, however brief, about what the broken tests mean in the summary would be really helpful for most users.