test failure for pull request #300

[ThomasBreuer]

The (first) failure at https://github.com/oscar-system/GAP.jl/pull/300/checks?check_run_id=322748382 looks strange.
My interpretation is that first GAP's GAPToJulia creates a nested array of the following structure.

julia> l = Any[ Any[ 1, 2 ], Any[ 3, 4 ] ]
2-element Array{Any,1}:
 Any[1, 2]
 Any[3, 4]

Then Julia's hcat turns the nested array into a 2-dimensional array.

julia> hcat( l... )
2×2 Array{Int64,2}:
 1  3
 2  4

julia> Base.repr( hcat( l... ) )
"[1 3; 2 4]"

In the seven (or six) jobs where the tests pass, the type of the array says that the entries are Int64, as in the above output, whereas the entries are Any in the (first) failed job.
I would understand if the type would be Array{Any,2} since Julia 1.3, but I cannot reproduce this Any situation in Julia 1.3 or Julia 1.4.
And how can it happen that the test result depends on the used GAP?

[fingolfin]

These failures are flukes, caused by some Travis VMs still using Julia 1.3.0-rc5. Once all are upgraded to 1.3.0 this issue will resolve itself

[fingolfin]

This went away for #300 now, as predicted