noinline attribute

[jackhexed]

This RFC adds the @[inline("never")] attribute to replace the local func; function func() pattern to prevent inlining.

Rendered

[Bottersnike]

Given #49 strikes down the idea of a @inline is the motivation against @noinline valid? Personally I think it would make far more sense over a mandatory argument that's only ever a single value (especially given the syntax for arguments to attributes isn't the prettiest too).

[Dekkonot]

I'm inclined to boost @noinline over @inline(never) for the reasons Bottersnike mentioned above. If we ever get an @inline attribute, we can add never as an option without too much trouble.

[jackhexed]

I am happy with either.

[jackhexed]

This is a simple RFC. There has been no discussion beyond @noinline vs @[inline("never")], and I am happy with both options, whichever would get this merged.

Is this is a feature that the Luau maintainers want?

[ishtar112]

Not mentioned as a motivation is the fact that inlined functions are [as expected] discarded from tracebacks, which can be especially problematic for debugging. Today, the only ways to prevent this are to annotate the relevant file as O0 or O1, or as the RFC mentions, to declare a function as function foo rather than local function foo -- neither of these are particularly desirable solutions, especially in comparison to a dedicated attribute. I am of the opinion that this feature is a must.

And to repeat what's already been said, I prefer @noinline over @[inline("never")].