Re: [boost] [Review.Coroutine] Some comments

Subject: Re: [boost] [Review.Coroutine] Some comments
From: Vicente J. Botet Escriba (vicente.botet_at_[hidden])
Date: 2012-09-08 10:25:15

Le 08/09/12 14:30, Oliver Kowalke a écrit :
> Am 08.09.2012 10:48, schrieb Vicente J. Botet Escriba:
>> Le 07/09/12 21:07, Oliver Kowalke a écrit :
>>> Hi,
>>>
>>> most of Vicente's suggestions are done ... 
>>> (git://gitorious.org/boost-dev/boost-dev.git, branch coroutine)
>>>
>>> Am 04.09.2012 23:06, schrieb Vicente J. Botet Escriba:
>>>> * Could you add an standard Allocator parameter as e.g. the one of 
>>>> packaged_task?
>>>>
>>>>    template <class F, class Allocator>
>>>>    explicit packaged_task(allocator_arg_t, const Allocator& a, F&& f);
>>> boost::coroutine uses internally intrusive_ptr instead of shared_ptr 
>>> - the trick of storing the rebounded allocator as deleter in 
>>> shared_ptr is not possible (because intrusive_ptr has no such 
>>> 'deleter' concept).
>>> I would not pay for shared_ptr because it is heavy weight compared 
>>> to intrusiv_ptr (no need for atomic increment etc.). intrusive_ptr 
>>> has only the size of a pointer.
>> I'm not suggesting that you use shared_ptr. I guess there are other 
>> means to get this work, e.g doing it explicitly in the destructor and 
>> move operations, isn't it?
> I can't imagine that it will work.
> Anyway, I could store an delete-functor (calls allocator::destroy() + 
> deallocate()) in context_base as shared_ptr does and call this delete 
> functor from intrusive_ptr_release. Let's see if it will work.
I'm sure you will find a good solution.
>
>>>
>>> On solution could be to make coroutine non-copyable,
>>
>> Is coroutine copyable?
> moveable-only - I mean to derive from boost::noncopyable
I'm lost. How this can help here?
>
>>
>>>
>>>> * Parameters and move semantics. Could the signature parameters be 
>>>> rvalue references?
>>>>
>>>>    R  operator()(A0  a0,  ...,  A9  a9);
>>>>
>>>> I'm asking this because I think that Boost.Tuple has not 
>>>> implemented yet move semantics.
>>> boost::tuple does not handle moveable-only objects - so it will not 
>>> work. we've to wiat until the supprot is added to boost::tuple.
>>>
>> Have you take a look at the boost::fusion::tuple move C++11 evolution?
> I'll take a look into the docu
I don't know when these should be released, but I know they are working 
on it. In any case I expect that boost::fusion::tuple will manage with 
move semantics one day, but to my knowledge no body is working on 
boost::tuple. Maybe someone has fresh infos.

Best,
Vicente