GitLab

Mlang currently only has a Python backend. The next backend that we would want to add is an OCaml backend, as we would want to perform fuzzing on the code generated from Mlang : https://caml.inria.fr/pub/docs/manual-ocaml/afl-fuzz.html.

The OCaml backend should follow a structure similar to the Python backend in src/mlang/backend_compilers/bir_to_python.ml, and use a type like this one for the values of the generated program:

type m_value =
 | Undefined 
 | Float of float 

Also, the generated OCaml should include preprocessor directives that link back to the original M and M++ source files for better error reporting during the fuzzing. This can be done using the following syntax:

(* file foo.ml *)
let bar = assert false in
# 5 "foo_1.ml"
let baz = assert false in
# 85 "foo_2.ml
let fizz = assert false in 

In this example, the assert false will point to:

• bar -> line 2 of foo.ml
• baz-> line 5 of foo_1.ml
• fizz-> line 85 of foo_2.ml