@gsl_function macros

[lxvm]

Hi,

I have been writing an extension for GSL.jl in Integrals.jl and I encountered #110. Following this post on discourse I rewrote @gsl_function to work with runtime closures, however I am not sure I am handling the memory correctly. For example, do I have to use GC.@preserve as written here? Or can the function_ field of the gsl_function struct become a Base.CFunction?

The relevant documentation also makes it clear that a cfunction closure is not supported on all platforms, which makes this pr less desirable.

The above being said, if/when a solution is clear the following would need to be done

• rewrite the 5 other gsl_function macros with the same solution

[simonbyrne]

Since GSL lets you pass an arbitrary thunk parameter (see https://julialang.org/blog/2013/05/callback/), I would suggest doing something similar to what we do in HDF5.jl:
https://github.com/JuliaIO/HDF5.jl/blob/91ef2841dcc5ddd52ef3094120b8921e0c1c5749/src/api/helpers.jl#L76-L86
Define a single "helper" cfunction which unwraps the thunk and evaluates the function, and pass the function itself via the thunk. This would avoid the runtime closure issue completely.

[lxvm]

Thank you for the guidance and links @simonbyrne! Due to the advantages you describe, the ease of mirroring a C idiom, and since my PhD adviser is the author of the blog post, I am in favor of the solution in your pr.