[libcxx-commits] [PATCH] D111283: [clang] template / auto deduction deduces common sugar

Mon Sep 12 12:34:00 PDT 2022

ychen added a comment.

In D111283#3784648 <https://reviews.llvm.org/D111283#3784648>, @mizvekov wrote:

> In D111283#3784615 <https://reviews.llvm.org/D111283#3784615>, @ychen wrote:
>
>> `A<T>` is currently modeled as ElaboratedType. It was `TemplateSpecializationType` before. Reading comments for `ElaboratedType`, it looks like sugar type might not be needed here?
>
> Ah you might be seeing an ElaboratedTYpe here, where there was none before, perhaps because of https://reviews.llvm.org/D112374, and not because of this patch.
>
> Yeah ElaboratedType is just sugar that should have no effect on partial ordering. But then I think no sugar should have effect on partial ordering. What is stopping you from just looking at the canonical type instead? On a canonical type, you would never see an ElaboratedType node, or a TemplateSpecializationType which is not dependent.

Thanks for the link. I'm not blocked by any of these patches, instead just trying to have a mental model of when to expect the sugared type :-). For partial ordering, the `TemplateSpecializationType` could be dependent, since the injected template arguments are dependent, I guess that's the reason there is the `ElaboratedType`?

> Is this related to the AutoType canonicalization issues we were discussing in the other patch of yours?

Nope. I found this AST difference while investigating D133683 <https://reviews.llvm.org/D133683>.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D111283/new/

https://reviews.llvm.org/D111283