ir: Track the codegen-reachable items, and use it instead of whitelisted_items() for code generation. #836
Conversation
|
r? @fitzgen |
|
(Only the |
|
maybe I should finish reading my inbox before leaving code reviews suggesting exactly what the next item in my inbox does... |
src/ir/context.rs
Outdated
| }); | ||
| // The reversal preserves the expected ordering of traversal, resulting | ||
| // in more stable-ish bindgen-generated names for anonymous types (like | ||
| // unions). |
There was a problem hiding this comment.
Why did this comment move here? It should be with the reverse call.
| self, | ||
| roots.clone(), | ||
| traversal::codegen_edges, | ||
| ).collect::<ItemSet>() |
There was a problem hiding this comment.
Mildly surprised the type inference couldn't figure this out.
| @@ -3299,7 +3298,7 @@ pub fn codegen(context: &mut BindgenContext) -> Vec<P<ast::Item>> { | |||
| } | |||
|
|
|||
| context.resolve_item(context.root_module()) | |||
| .codegen(context, &mut result, whitelisted_items, &()); | |||
| .codegen(context, &mut result, codegen_items, &()); | |||
There was a problem hiding this comment.
File a follow up E-Easy issue to remove this argument, and remove the whitelisted_items parameter from all the codegen functions using ctx.codegen_items() instead?
…ted_items() for code generation. This standardizes the behavior change at rust-lang#834, but without regressions. I've added a few more tests for rust-lang#833 here.
|
@bors-servo r=fitzgen |
|
📌 Commit 4fee077 has been approved by |
|
☀️ Test successful - status-travis |
This standardizes the behavior change at #834, but without regressions.
I've added a few more tests for #833 here.
Closes #834.