librustc: Don't overwrite vtables when coercing to trait object. #14830
Conversation
| @@ -152,21 +152,32 @@ pub struct MethodCallee { | |||
| #[deriving(Clone, PartialEq, Eq, Hash, Show)] | |||
| pub struct MethodCall { | |||
| pub expr_id: ast::NodeId, | |||
| pub autoderef: u32 | |||
| pub autoderef: u32, | |||
| pub adjustment: bool | |||
There was a problem hiding this comment.
The meaning of adjustment is pretty opaque. I'm not a big fan of boolean variables in general, but at least we must find a better name here to indicate what sort of adjustment you are talking about.
There was a problem hiding this comment.
Yes, I couldn't think of anything better. I went with adjustment cause it's true for when we add an adjustment (AutoDeref or AutoObject).
There was a problem hiding this comment.
On Thu, Jun 12, 2014 at 06:06:19PM -0700, Luqman Aden wrote:
Yes, I couldn't think of anything better. I went with adjustment
cause it's true for when we add an adjustment (AutoDeref or
AutoObject).
I think I don't quite understand what it means yet. I feel like we
should replace the simple autoderefs: uint either with some kind of
enum, or maybe just rename it to "adjustments: uint" and just count
AutoObject or AutoDeref uniformly. Either change will require a bit of
adjustments elsewhere.
Certainly we need a comment on MethodCall explaining how it works, and
giving a few examples (e.g., an expression with just autoderef, just
autoobject, or both, and what the different things look like). To some
extent that's pre-existing, because there should have been a comment
already: in the past, though, I know I wrote some extensive comments
along these lines for the rooting table, but we ripped that out and I
think the comments with it.
Fixes #14399.