Rollup merge of #58609 - gabi-250:mutable-refs, r=oli-obk
Allow Self::Module to be mutated. This only changes `&Self::Module` to `&mut Self::Module` in a couple of places. `codegen_allocator` and `write_metadata` from `ExtraBackendMethods` mutate the underlying LLVM module. As such, it makes sense for these two functions to receive a mutable reference to the module (as opposed to an immutable one). I am trying to implement `codegen_allocator` for my backend, and I need to be able to mutate `Self::Module`: https://github.com/rust-lang/rust/blob/f66e4697ae286985ddefc53c3a047614568458bb/src/librustc_codegen_ssa/traits/backend.rs#L41 Modifying the module in `codegen_allocator`/`write_metadata` is not a problem for the LLVM backend, because [ModuleLlvm](https://github.com/rust-lang/rust/blob/master/src/librustc_codegen_llvm/lib.rs#L357) contains a raw pointer to the underlying LLVM module, so it can easily be mutated through FFI calls. I am trying to avoid interior mutability and `unsafe` as much as I can. What do you think? Does this change make sense, or is there a reason why this should stay the way it is?
Showing
想要评论请 注册 或 登录