DST raw pointers - *-pointers are fat pointers

[nrc]

r? @pnkfelix or @nikomatsakis

[nikomatsakis]

I will review, I'd like to get more familiar with this code.

[nrc]

@nikomatsakis r? updated to fix coercion fn calls

[huonw]

maybe this new line could have a trailing , to make it work better in diffs?

[pnkfelix]

I don't think it can because our macro_rules does not allow for optional trailing commas. Or at least, you do not get them automatically with $(...),*, and I do not know if you can express them otherwise. The last time I tried to emulate the effect in macro_rules! i was not successful.

[huonw]

Ah, this felt so much like an enum definition that didn't even cross my mind.

I guess we could have a compulsory trailing comma via $(... , )*. Also, [vec!`](http://doc.rust-lang.org/master/std/macro.vec!.html) somehow manages to satisfy the compiler to make optional trailing commas work.

Anyway, not important.

[nrc]

It occurs to me that we should not in fact allow implicit coercions where don't allow the corresponding implicit derefs for unsafe pointers. E.g.,

    let x: *const S = &S;
    let y: *const T = x;

The second line here should be an error. I believe combining the two statements is OK:

    let y: *const T = &S;

Since we derefing a borrowed pointer, unsizing, then making an unsafe reference.

I'll remove the ability to do the bogus adjustments on Monday.

[nikomatsakis]

On Fri, Aug 29, 2014 at 01:39:48AM -0700, Nick Cameron wrote:

It occurs to me that we should not in fact allow implicit coercions where don't allow the corresponding implicit derefs for unsafe pointers. E.g.,

    let x: *const S = &S;
    let y: *const T = x;

The second line here should be an error. I believe combining the two statements is OK:

    let y: *const T = &S;

Since we derefing a borrowed pointer, unsizing, then making an unsafe reference.

I agree that we should be consistent, although I think we should
probably allow the adjustments in the case of *T. I think the bottom
line is that we need to agree upon the rules and write up a
"descriptive" RFC so that we are all on the same page. I'll take a
stab at that. Per our conversation at work week, we can eschew deep
conversions (for now, at least).

I'll remove the ability to do the bogus adjustments on Monday.

In a separate PR?

[nikomatsakis]

r+ modulo nits

[nrc]

Addressed reviewer comments