Re: [PATCH Coroutines]Insert the default return_void call at correct position

[Iain Sandoe <idsandoe@googlemail.com>]

Hello Bin, Nathan,

Iain Sandoe <iain@sandoe.co.uk> wrote:
>
>
> We are seeking to clarify the standard wording around this (and the cases  
> where
> unhandled_exception() returns - hopefully during the WG21 meeting this  
> week,
>
> FWIW, I think your interpretation makes sense here.

It’s not clear if the committee will have time to process wording changes  
needed for this during the current session, however, it is agreed that  
injecting the return_void at the closing brace of the user’s original  
function (your solution) is the correct approach.

So, from my point of view this patch DTRT and should be applied,

thanks
Iain


>> Thanks,
>> bin
>>
>> On Mon, Feb 3, 2020 at 1:55 PM bin.cheng <bin.cheng@linux.alibaba.com>  
>> wrote:
>>> Hi,
>>>
>>> Exception in coroutine is not correctly handled because the default
>>> return_void call is now inserted before the finish suspend point,
>>> rather than at the end of the original coroutine body.  This patch
>>> fixes the issue by generating following code:
>>> co_await promise.initial_suspend();
>>> try {
>>>  // The original coroutine body
>>>
>>>  promise.return_void(); // The default return_void call.
>>> } catch (...) {
>>>  promise.unhandled_exception();
>>> }
>>> final_suspend:
>>> // ...
>>>
>>> Bootstrap and test on x86_64.  Is it OK?
>>>
>>> Thanks,
>>> bin
>>>
>>> gcc/cp
>>> 2020-02-03  Bin Cheng  <bin.cheng@linux.alibaba.com>
>>>
>>>      * coroutines.cc (build_actor_fn): Factor out code inserting the
>>>      default return_void call to...
>>>      (morph_fn_to_coro): ...here, also hoist local var declarations.
>>>
>>> gcc/testsuite
>>> 2020-02-03  Bin Cheng  <bin.cheng@linux.alibaba.com>
>>>
>>>      * g++.dg/coroutines/torture/co-ret-15-default-return_void.C: New.