Discrepancy in (integer *= float); between LSL Mono and LSL Luau

[canny]

I'm reporting this one because we are looking for all the differences, but I don't think that it has any real use.

// LSL Mono

default { state_entry() {

    integer a = 2;

    (a *= 2.2);

    llOwnerSay((string)a);  // --> 4

}}

// LSL Luau

default { state_entry() {

    integer a = 2;

    (a *= 2.2);  // --> Script run-time error : cannot take result of integer *= float

}}

a *= 2.2; works in both, float f = (a *= 2.2); errors in both.

https://secondlife.canny.io/admin/feedback/slua-alpha/p/discrepancy-in-integer-float-between-lsl-mono-and-lsl-luau

[canny]

This issue has been linked to a Canny post: Discrepancy in (integer *= float); between LSL Mono and LSL Luau 🎉

[HaroldCindy]

Yep, we need to block all places where we take the result of int *= float, but wrapping it in parens isn't really "taking the result".

I suspect we need to fix this in Tailslide, since that's where the LLExpression::getResultNeeded() that's used by SLua's LSL compiler lives. I imagine it's doing that calculation before stripping out parens, and doesn't account for them.

[Suzanna-Linn]

This is a different error around the same place.

Now without parentheses (that it would work) but with a global variable:

// LSL Luau
integer a = 2;
default { state_entry() {;
    a *= 2.2;  // --> Script run-time error: lsl_script:0: attempt to perform arithmetic (mul) on nil and number
}}

is it storing global variables with a "_g" prefix?

slua/Compiler/src/LSLCompiler.cpp

Lines 2009 to 2012 in 1025126

	case SYM_GLOBAL:
	// prevent shadowing of builtin globals, since they're addressed by name
	name = "_g" + name;
	break;

but accessing it without the prefix?

slua/Compiler/src/LSLCompiler.cpp

Lines 1239 to 1243 in 1025126

	if (lval_sym->getSubType() == SYM_GLOBAL)
	{
	source_reg = allocReg(bin_expr);
	pushGlobal(source_reg, lval_sym->getName());
	}