dmpolukhin wrote: > I have a mixed feeling about whether we should always exclude all coroutines > by default or not. When a coroutine has only one `co-operator` and nothing > else should be safe to use a reference parameter? Am I wrong?
No, it may not be safe even in case of single co_return because of initial coroutine suspend point *before* entering function body. So if the parameter passed by reference is used inside coroutine it is not safe [see more here](https://en.cppreference.com/w/cpp/language/coroutines.html#:~:text=thus,%20may%20become%20dangling), except for the cases when the call is under co_await i.e. it is synchronous call. > So I think to be "on a safe side" I'd suggest a new option "IngoreCoroutines" > that will be `true` by default. If A user 100% knows that all references will > outlive the coroutine he could set it to `false`. In the option description > we could describe why it is not good to disable it as per cppcore-guidlines. > Generally speaking, I think it's not _very_ good to implement > _recommendations_ of cppcore-guidlines as a must in non-cppcore checks, they > should be under a flag. > > I'd be happy to accept your changes with this option, without - I'd want a > second opinion from a maintainer. Let's wait a bit more for maintainer, if no response, I'll add an option and this this case we will have a question what should be the default value for the option. https://github.com/llvm/llvm-project/pull/140912 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
