The meaning of providing a constant for a mutable global import

[lars-t-hansen]

Suppose a module is this:

(module (import "m" "g" (global (mut i32))))

and we try to instantiate it with the import object {m: {g: 42}}. Is this allowed or not?

Firefox currently allows it and there seems to be no difficulty in allowing it generally, but I think the spec is vague here. The current spec text to cover this seems to be (document/js-api/index.bs line 335):

        1. If |externtype| is of the form [=𝗀𝗅𝗈𝖻𝖺𝗅=] |globaltype|,
            1. If |globaltype| [=Type=](|v|) is [=Number=],
                1. If |globaltype| is [=𝗂𝟨𝟦=], throw a {{LinkError}} exception.
                1. Let |value| be [=ToWebAssemblyValue=](|v|, |globaltype|.<em>[=global type|valtype=]</em>)
                1. Let |store| be the current agent's [=associated store=].
                1. Let (|store|, |globaladdr|) be [=alloc_global=](|store|, |globaltype|, |value|).
                1. Set the current agent's [=associated store=] to |store|.

A globaltype can't be just a JS type I think, it needs to carry a mutability specifier. One could argue that since this is a number value the implicit mutability specifier is const, but need it be so?

[lars-t-hansen]

FWIW the original proposal text did allow a number to be imported into a mutable global: https://github.com/WebAssembly/mutable-global/blob/master/proposals/mutable-global/Overview.md#webassemblyinstance-constructor

[binji]

One could argue that since this is a number value the implicit mutability specifier is const, but need it be so?

It makes the most sense to me for it to be immutable. What is the benefit of allowing an imported Number to be mutable?

[rossberg]

Allowing an implicit coercion from Number to mutable Global does not seem desirable.

I think the occurrence of |globaltype| in the second line of the snippet is just a typo?

[lars-t-hansen]

I agree that it makes most sense for it to be immutable (and we can always open it up later), I just wanted to make sure the decision was made explicitly. The spec text needs to synthesize the const part of the globaltype here explicitly.

@rossberg, yes I was confused by that |globaltype|, couldn't tell if it was magic spec syntax or what.