Strange bug(Segmentation fault: 11) solely existed in julia v1.0

Firstly, the codes run well in v1.1 and nightly version.

When it comes to CI in v1.0, there is a strange phenomenon:


It firstly appeared when a project requires both DataFrames.jl and MLStyle.jl in Julia v1.0, but at that time I didn’t get a break with this.

Now I’m contributing to Query.jl, taking advantage of MLStyle.jl to largely simplify some ast manipulations: Simplify macro rewriters and avoid some prospective bugs by thautwarm · Pull Request #238 · queryverse/Query.jl · GitHub , thus it has become a severe impedement for it makes tests on v1.0 failed.

The code failed at https://github.com/thautwarm/MLStyle.jl/blob/master/src/toolz.jl#L23 , where I implemented a state monad to make MLStyle’s complex state transformations clean.

Also, this is on the traceback stacks:

($)(f, a) = f(a)

I cannot make sure what happened here, but it seems to be a solid bug in Julia v1.0.

You could try a git bisect to locate the exact commit that fixed it, ans ask for it to be backported for the next 1.0.x release.

Okay, I’ll have a try in this week.