It is possible to implement it but I've concerns because the coroutine-fn has another signature/return type as the signature given to coroutine as template argument. It might confuse users.
typedef coroutine< int( int) > coro_t;
void fn( coro_t::self_t&, int) {...}
no sure - probably not, but it's not that hard to add an BOOST_STATIC_ASSERT :)
Does the current implementation check at compile time that the return type of the coroutine function is the one of the signature?
Sorry. I believed you had reached to mix them. Maybe you could reach to make a StackAllocator model a standard allocator just by renaming the stack allocator functions.
standard-allocator requires functions like deallocate/destory allocate/create.
stack-allocators do not construct objects the allocate/deallocate only memory chunks. I believe those are two different concepts (beside the special requirements of stack-allocators on the return address of the allocated chunk).
Yes. It is a bad idea to mix them. I don't remember if you plan to add standard allocator or not. Could you confirm?
yes - the version in my git-repo uses allocators (std::allocator as default) in coroutines ctor (beside of the stack-allocator). regards, Oliver