Packages failing on Julia 0.6 vs 0.5

package

#1

Now that the first release candidate for 0.6 is out, it’s worth taking a quick look at the state of the package ecosystem. The below gist is pulled from the data at http://pkg.julialang.org/pulse.html, running on PackageEvaluator each night. Just under 50% of registered packages pass their tests on Julia 0.6.0-rc1 on PackageEvaluator (which runs in an Ubuntu 14.04, x86_64 VM), compared to almost 70% for Julia 0.5.2.

Gist of results diff and script used to generate it

The results are sorted by number of github stars, as a proxy for popularity. If anyone wants to help out with any of these, it would be worth first checking out master of the package to see if there’s been a fix that hasn’t yet been released, or checking if there are any open pull requests otherwise. There are a small number of cases that fail on 0.5 but pass on 0.6, for one reason or another. Replace “Example” with the package name in http://pkg.julialang.org/detail/Example.html to get to a detail page where you can see full test logs. Some packages may only be listed for 0.5 or only for 0.6, which could happen for a few reasons - either the package timed out while running its tests on one of the Julia versions, or the package might be deprecated and not installable on 0.6, or it might be new enough that it’s 0.6-only, etc. If you’re curious about a specific package and the logs aren’t enough to explain what’s happening, feel free to ask about it and I can take a look at the raw data sources.


#2

Fixing this error may be beyond me… :slight_smile:

ERROR: LoadError: LoadError: error compiling inv: error compiling trtrs!: 
could not load library "libopenblas64_"
dlopen(libopenblas64_.dylib, 1): Library not loaded: /usr/local/opt/gcc@6/lib/gcc/6/libgfortran.3.dylib
 Referenced from: /Users/travis/julia/lib/julia//libopenblas64_.dylib
 Reason: image not found

#3

Is that a nightly-specific Travis issue? It’s related to https://github.com/JuliaLang/julia/pull/21717, should hopefully not influence release candidates at all if we’re careful.


#4

Yes, sorry, that’s an error on nightly (0.7).