save-analysis: fix ICE on partially resolved path #37908
Conversation
|
r? @arielb1 (rust_highfive has picked a reviewer for you, use r? to override) |
|
r? @eddyb |
|
Can we have a test for this? There is a large change(#37676) to how we store partial/full resolution in the queue. |
| @@ -497,7 +497,7 @@ impl<'l, 'tcx: 'l> SaveContext<'l, 'tcx> { | |||
| } | |||
|
|
|||
| pub fn get_path_data(&self, id: NodeId, path: &ast::Path) -> Option<Data> { | |||
| let def = self.tcx.expect_def(id); | |||
| let def = option_try!(self.tcx.expect_resolution(id).maybe_full_def()); | |||
There was a problem hiding this comment.
Can you do it here, without adding a method?
There was a problem hiding this comment.
I could, but it would be (IMO) a bit of an abstraction violation - depth and whether a def is 'baked' or not seem like implementation details that we shouldn't be dealing with here. Plus the method seems like a thing we'd might want - there are similar methods all over the ty ctxt, etc.
There was a problem hiding this comment.
The reason I'm saying this is because of #37676 which will catch that case differently here and the post-HIR-lowering compiler in general stops using PathResolution at all (which was a "temporary" hack I introduced ages ago, and it causes serious issues, even if it may seem more convenient in some cases).
Occurs when we produce save-analysis before type checking is complete (due to errors).
|
@eddyb change made |
|
@bors r+ |
|
📌 Commit b1f86fb has been approved by |
|
⌛ Testing commit b1f86fb with merge d5814b0... |
save-analysis: fix ICE on partially resolved path Occurs when we produce save-analysis before type checking is complete (due to errors).
Occurs when we produce save-analysis before type checking is complete (due to errors).