Missing result locations for builtin types

[vesim987]

Zig Version

0.11.0-dev.3826+7a197f124

Steps to Reproduce and Observed Behavior

While porting microzig to latest zig version I have found couple of missing result locations for builtins.

const Foo = struct {
    bar: u32,
};

const MyStruct = struct {
    foo: *Foo,
};

export fn issue_1() void {
    _ = MyStruct{
        .foo = &.{
            .bar = @intCast(20),
        },
    };
}

fn do_something(_: *Foo) void {}

export fn issue_2() void {
    do_something(&.{
        .bar = @intCast(20),
    });
}


export fn isse_3 () void {
  _ = [1]u16{
      @intCast(20),
  };
}

It also applies to other builtins.

Expected Behavior

Every example to compile.

[nektro]

issue_1 and issue_2 create a *const Foo in their bodies, so i would not expect those to compile

[mlugg]

It makes sense that issue_1 and issue_2 don't work, but perhaps we want them to. issue_3 should certainly work however - #16226 is technically a duplicate but we can track that issue there for separation.

[perillo]

@mlugg, Issue #16226 has been closed.

I suspect that the problem is no longer "missing result locations", but about "how to cast away constness".

[mlugg]

Yes, #16226 has been closed because the issue_3 snippet now works. The other issues detailed here were resolved by #16512, so this issue can now be closed. The constness cast failure you noticed is just a minor bug in the given test case, unrelated to the issue at hand.

[perillo]

The constness cast failure you noticed is just a minor bug in the given test case, unrelated to the issue at hand.

What I assumed from your comment:

It makes sense that issue_1 and issue_2 don't work, but perhaps we want them to

is that since the destination is not a const pointer, then the compiler should automatically cast away the constness, since it is safe.