Repository navigation
CWG2934 [dcl.fct.def.coroutine] Unclear semantics of exception escaping from unhandled_exception #574
Description
Full name of submitter: Brian Bi
Reference (section label): [dcl.fct.def.coroutine]
Issue description: [dcl.fct.def.coroutine]/14 says:
If the evaluation of the expression
promise.unhandled_exception()exits via an exception, the coroutine is considered suspended at the final suspend point and the exception propagates to the caller or resumer.
All implementations first destroy the promise object followed by the parameter copies, then deallocate the coroutine state, and finally return control to the caller: https://godbolt.org/z/e7edh8d9o
The behaviour of the implementations is expected based on the code in [dcl.fct.def.coroutine]/5, but the wording in p14 suggests that suspension occurs immediately when unhandled_exception() exits, prior to any stack unwinding in the coroutine. The intent should be clarified.
Activity
t3nsor commentedon Jul 13, 2024
Prior discussion can be found here and here.
t3nsor commentedon Sep 9, 2024
Bump
jensmaurer commentedon Sep 9, 2024
Any wording suggestion?
jensmaurer commentedon Sep 9, 2024
CWG2934
[-][dcl.fct.def.coroutine] Unclear semantics of exception escaping from `unhandled_exception`[/-][+]CWG2934 [dcl.fct.def.coroutine] Unclear semantics of exception escaping from `unhandled_exception`[/+]t3nsor commentedon Sep 10, 2024
I think the reason why I didn't write a wording suggestion for this one is that it depends on what we decide we want for [CWG2935]. The wording is cleaner if we assume that an exception propagating out of the coroutine always destroys the coroutine state.