Compute sizedness with a fixed-point analysis #1102
Conversation
src/ir/analysis/sizedness.rs
Outdated
| } | ||
|
|
||
| TypeKind::ObjCInterface(..) => { | ||
| trace!(" I guess obj-c interfaces are not zero-sized?"); |
There was a problem hiding this comment.
FWIW, as far as I know, every obj-c "interface" have an isa pointer.
There was a problem hiding this comment.
Cool, I'll note this. FWIW, I was just copying existing behavior.
| trace!(" comp considers its own fields and bases"); | ||
|
|
||
| if !info.fields().is_empty() { | ||
| return self.insert(id, SizednessResult::NonZeroSized); |
There was a problem hiding this comment.
But what if all fields are zero sized 🤔 ?
There was a problem hiding this comment.
They can't be, because of the every-object-must-be-addressable rules. Only base members can truly be zero sized, because its the "same" object.
There was a problem hiding this comment.
That is, with
struct Empty {};
struct Inherits : public Empty {
int x;
};
struct Contains {
Empty empty;
int x;
};sizeof(Contains) is one greater than sizeof(Inherits)
There was a problem hiding this comment.
yeah, makes sense. Thank you!
|
Thanks for the review @pepyakin ! |
This fixes a couple bugs where we weren't properly adding an `_address` byte. It also helps pave the way for computing implicit fields (such as padding, `_address`, vtable pointers, etc) in its own pass, before codegen. Fixes rust-lang#768
|
I added that example, and another for arrays of zero-sized types, as tests. Thanks again for review @pepyakin ! @bors-servo r=pepyakin |
|
📌 Commit d018f42 has been approved by |
|
Added one more test for arrays of zero elements and flexible arrays also. @bors-servo r=pepyakin |
|
📌 Commit fd7afb6 has been approved by |
|
☀️ Test successful - status-travis |
This fixes a couple bugs where we weren't properly adding an
_addressbyte. It also helps pave the way for computing implicit fields (such as padding,_address, vtable pointers, etc) in its own pass, before codegen.Fixes #768
r? @pepyakin