kondziu / fml Goto Github PK

begin
    let x = 20;
    object begin
        let Z = x;
        let Y = object begin 
                        let a = 4;
                        let b = Z;
                    end;
    end;
end;

prints:

thread 'main' panicked at 'Get global error: no such global: Z', src/bytecode/interpreter.rs:618:18
note: run with RUST_BACKTRACE=1 environment variable to display a backtrace

object(a, b) cannot see variable Z in object(Z, Y), but Z can see variable x, Is that desired?

Example .fml file:

function f() -> print("x\n");

Error (using the run command):

thread 'main' panicked at 'Interpreter error: Code address 0002 out of bounds (0..2)', src/main.rs:141:14

It works fine if the implicit global function contains anything else - e.g. just null.

Version 1 doesn't crash.

My compiler+interpreter has pretty much the same bug - there are 2 causes:

• With an empty input file, the parser returns a Top which contains a Null node so there's 1 instruction in the generated code. With a file that contains a function but nothing else, the parser returns a Top which contains the function but no other nodes so the implicit global function ends up having 0 instructions.
• My interpreter first gets the instruction from the code vector (this is the index out of bounds), executes it, then breaks the loop if the instruction pointer has jumped out of the code vector. Moving this check to the start of the loop fixes the crash for me, maybe your interpreter does something similar.

After disecting a fml program

function a() -> 5
print("~\n", a()) 

(using v2 of fml compiler) the bytecode contains the instruction label indexcp01 (00 01 00) twice.
To be more precise, the body of function a is "copied" inside the body of "main" function (λ:)

Included in the zip file is the fml source (with possible extra semicolons, woops), the compiled bytecode and two disected .txt files -- one is pure bytecode, only "formatted" a bit, the second one is human readable version of said bytecode.

fncinsideprint.zip

The spec states that globals refer to String program objects, see https://courses.fit.cvut.cz/NI-RUN/specs/bc.html#globals.
However, the reference compiler implementation seems to make it so that globals refer to Slot objects instead.

What is the intended behavior? Or am I missing something?

Code:

function f() -> begin
    let local = 123;
    let o = object begin
        function m() -> local;
    end;
    o.m();
end;

f();

Error:

thread 'main' panicked at 'Interpreter error: Error evaluating get global #3:

Caused by:
    No such global `local`.', src/main.rs:141:14

I guess FML doesn't support closures but it should probably error when compiling, not interpreting.

Code:

function x(a) -> a;

Error:

thread 'main' panicked at 'Interpreter error: Code address 0002 out of bounds (0..2)', src/main.rs:149:14

On this line it excecutes it if no ENV variable is specified, but then it continues parsing the commandline anyway and thus running it using the default parser, compiler, interpreter again

Possibly forgotten exit $?

Context:

Comparison (== and !=) operators for custom object and null break equality symmetry.

For example:

let obj = object begin end;

print("~\n", obj == null); // true
print("~\n", null == obj); // false

The expected behavior in this case is for both to be false.

This program prints 5 - is that intended? If i uncomment the let, then it crashes as expected.

begin
    //let x = 4;
    x <- 5;
end;
print("~\n", x);

Code:

let a1 = array(123, 1+1);
let a2 = array(456, 2+2);

Error:

thread 'main' panicked at 'Cannot register new global ::size: Cannot register global `#3`, index is already registered as a global.', src/bytecode/compiler.rs:273:30

print("\~\n", 5); prints 5\n (without a newline).
print("\~\~\n", 5, 5); prints 5\5 (with a newline).

The spec says "~ converts to the character ~" - is that supposed to mean it prints a ~ or that it behaves the same as ~? If it behaves the same (it almost does now), there's no way to print a literal ~.

There's also no way to print a \ according to the spec.