Inlined functions still have protos present in bytecode without references as values

[EpixScripts]

I was toying around with bytecode and seeing how the inlining mechanism works, but then I noticed when I have a function that is inlined and is only used by calling it, the proto is still present in the bytecode and a NEWCLOSURE/DUPCLOSURE instruction is created. However, the closure is not used. For example, if I input the script:

local function test()
    print("e")
end
test()

When compiled with optimization level 2 and debug level 1, disassembly reads:

Function 0 (test):
    2:     print("e")
GETIMPORT R0 1
LOADK R1 K2
CALL R0 1 0
    3: end
RETURN R0 0

Function 1 (??):
    1: local function test()
DUPCLOSURE R0 K0
REMARK inlining succeeded (cost 5, profit 1.60x, depth 0)
    2:     print("e")
GETIMPORT R1 2
LOADK R2 K3
CALL R1 1 0
    4: test()
RETURN R0 0

A closure is created and loaded into register 0, but never used.
Is there a specific reason for this, or would this be considered a bug?

[zeux]

Yeah this is a known limitation of the bytecode compiler. We plan to lift it eventually, but it's not trivial right now due to how the compiler is structured.

[EpixScripts]

Yeah this is a known limitation of the bytecode compiler. We plan to lift it eventually, but it's not trivial right now due to how the compiler is structured.

Ok, that makes sense. For errors that occur within an inline function call, how will line numbers and function names be handled for the error message?

[zeux]

The inlined function disappears from the callstack but the line numbers should correspond to the line inside the inlined function, not the call site (which is why you can see print("e") in the bytecode output above.