[PATCH] D145639: [Coroutines] Fix premature conversion of return object

[Chuanqi Xu via Phabricator via cfe-commits]

ChuanqiXu added a comment.

In D145639#4198837 <https://reviews.llvm.org/D145639#4198837>, @ilya-biryukov wrote:

> In D145639#4198794 <https://reviews.llvm.org/D145639#4198794>, @ChuanqiXu wrote:
>
>>> ! In D145639#4198815 <https://reviews.llvm.org/D145639#4198815>, @ChuanqiXu wrote:
>>>  The RVO seems to be mandated by the standard.
>>
>> Yeah, I thought so... But WG21 decided to not make it in a recent meeting... So we can only make it as a non-mandated compiler optimization.
>
> That's unexpected. Are there any wording change proposals published?
> Our understanding comes from the reading of dcl.fct.def.coroutine/7 <https://timsong-cpp.github.io/cppwp/n4868/dcl.fct.def.coroutine#7>
>
>> The expression promise.get_return_object() is used to initialize the glvalue result or prvalue result object of a call to a coroutine.
>
> And we **think** RVO is mandated when initializing the **result objects** mentioned in general.
> Also, from https://github.com/llvm/llvm-project/issues/56532:
>
>> Q: if the initialization does occur later, by what mechanism the prvalue result of get_return_object is forwarded to that initialization.
>> A: (see below)
>> Case 1/2. Same type of get_return_object and coroutine return type.
>> **Constructed in the return slot.**
>> Case 3. Different types:
>> (1) Constructed temporary object prior to initial suspend initialized with a call to get_return_object()
>> (2) when coroutine needs to to return to the caller and needs to construct return value for the coroutine it is initialized with expiring value of the temporary obtained in step 1.

The formal file shouldn't be released yet. But we can find the issue report and the proposed solution in https://cplusplus.github.io/CWG/issues/2563.html.

The words you quoted is the recording of the wg21's discussion during the meeting. And that's the **intention** of the designers but it is not reflected to the wording.  What issue 2563 discussed is that if we should delay the conversion. And the RVO is a conclusion of that. But according to the result, we should say the compiler is allowed to do RVO if the types matched but it is not required.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D145639/new/

https://reviews.llvm.org/D145639