(&v != &v) for function parameters passed by value

[slonik-az]

Zig Version

0.12.0-dev.1606+569182dbb

Steps to Reproduce and Observed Behavior

Related to #16343. A new stack variable is created each time the address of the function parameter is taken resulting in &v == &v being false.

Here is the repro

const std = @import("std");

fn pro_cmp_address(v: i32) bool {
    return (&v == &v);
}

test "pro compare address" {
    try std.testing.expectEqual(true, pro_cmp_address(123)); // Test fails
}

Expected Behavior

It is expected that &v == &v is true. Taking address of a variable should have no side effects.

[slonik-az]

One possible solution would be to follow Carbon language approach and ban taking address of the function parameter altogether (make it compile error). The rational for this drastic act being that compiler is free to put a parameter into a register, in which case no address is available anyway. If for some reason one needs the address of the object passed into the function one can change the function's signature and explicitly pass it by (const) pointer.

[xdBronch]

this can be closed now

[andrewrk]

Fix landed in 8245d7f.