problem with conversion from GAP to a prescribed Julia type?

[ThomasBreuer]

In order to create a Julia array of strings from a GAP list of strings, one can do the following.

gap> l:= [ "a" ];;  arr1:= GAPToJulia( l );
<Julia: Any["a"]>
gap> arr2:= Julia.Base.convert( JuliaEvalString( "Array{String,1}" ), arr1 );
<Julia: ["a"]>
gap> JuliaTypeInfo( arr1 );
"Array{Any,1}"
gap> JuliaTypeInfo( arr2 );
"Array{String,1}"

When I try to skip the intermediate step, by giving the intended type information to GAPToJulia,
I get the following.

gap> GAPToJulia( JuliaEvalString( "Array{String,1}" ), l );
Error, StackOverflowError:
not in any function at *stdin*:13
type 'quit;' to quit to outer loop

[fingolfin]

Heh, and the Julia version leads to a nice segfault:

julia> l = GAP.EvalString(""" l:= [ "a" ];; """)[1][2]
GAP: [ "a" ]

julia> GAP.gap_to_julia( Array{String,1}, l )
Bus error: 10

[fingolfin]

OK, so part of the problem is that we actually don't have a conversion to String, only to AbstractString. Both examples work if one replaces String by AbstractString.

But if one specifies an unsupported conversion, the system currently runs into an infinite recursion (see also issue #165), and then sometimes Julia manages to catch it, and sometimes not (because GAP also deepens the call stack)

[sebasguts]

So, a quick fix is replacing the AbstractString in gap_to_julia with String.

The current return type of the conversion to AbstractString is String anyway, so I would guess having a fallback there is also no problem.