September 2011 - Boost - lists.boost.org

Re: [boost] [any] new version
by Martin Bidlingmaier 01 Sep '11

01 Sep '11

> Mathias Gaunard wrote: >> You could make next_id a function with weak extern linkage, which should >> be able to make it work. Thanks for the replies, I hope I fixed the linkage problems now, by making next_id() inline. There were other linker related issues, which are hopefully fixed now as well. The modified any.hpp header and a simple linker test are attached -- NEU: FreePhone - 0ct/min Handyspartarif mit Geld-zurück-Garantie! Jetzt informieren: http://www.gmx.net/de/go/freephone

2 2

[shared_ptr] Implicit 2 phase construction with enable_shared_from_this
by Sebastian Karlsson 01 Sep '11

01 Sep '11

There's been some discussion about solving the problem with shared_from_this() not being available in the constructor. This is fairly natural as shared_ptr ownership doesn't happen until after the constructor leaves, and we're not even guaranteed that the object enters shared_ptr ownership at all. What I'm suggesting is the addition of a private void enable_shared_from_this<T>::accepted_shared_ptr() {}. Then inside enable_shared_from_this<T>::_internal_accept_owner we add: const_cast< T* >( static_cast< const T* >( this ) )->accepted_shared_ptr(); Just beneath: weak_this_ = shared_ptr<T>( *ppx, py ); This allows clients of enable_shared_from_this to add their own accepted_shared_ptr() which will correctly be called while I believe not breaking any old code. I just added the const_cast because I'm not familiar with the rationale behind _internal_accept_owner being const. This change nets us: 1 ) Generic code dealing with construction of shared_ptrs ( in any of its forms ) does not need to bother themselves with 2 phase construction. 2 ) The distinction between construction & transfer of ownership to shared_ptr makes it easier to cope with both the case where T is not managed by a shared_ptr and the case where it is. One gotcha I've found is that for example: struct foo : enable_shared_from_this< foo > { foo() { boost::shared_ptr< foo > tmp( this, null_deleter() ); } // We might be allocated on the stack, so use a null_deleter for now void accepted_shared_ptr() { /* here we can patch it up with shared_from_this(), which will use our assigned deletion strategy */ } } Does not behave as one would want to. Basically in this case I'd want to use a shared_ptr<> which makes no assumptions about whether we're managed by a shared_ptr or not, we want a shared_ptr which acts as a plain pointer. The problem however is that the shared_ptr which we construct with the null_deleter will grab the enable_shared_from_this before even construction is complete. This can be solved by the following, as seen in boost::make_shared: ... boost::shared_ptr< foo > empty_with_deleter( static_cast< foo* >( 0 ), null_deleter() ); boost::shared_ptr< foo > tmp( empty_with_deleter, this ); // This one is safe to use ... Which doesn't enter boost::detail::sp_enable_shared_from_this. Would love some input from someone a bit more knowledgeable about shared_ptr. Kind regards, Sebastian Karlsson

1 0

[review] Conversion - About specialization, overloading and ODR
by Vicente J. Botet Escriba 01 Sep '11

01 Sep '11

Hi, during the Boost.Conversion review and pre-review it has been argued that as the user must provide specializations of the conversion function for other UDT's, it could promotr ODR violations as two independent libraries can provide the same specialization making them incompatible for an end user. This issue is present for any library for which the user could provide an overload or could specialize a class. The standard library protects itself from this issue saying that is undefined behavior any specializations of standard classes (except maybe numeric_traits that has a single parameters and it is intendeed to be specialized by the library providing the UDT). My question is: Should Boost (all the Boost libraries) document explicitly that any class specialization or overload of free fuctions could result in undefined behavior? Is this already implicitly acknowledged? Best, Vicente

3 7

[review] Conversion review ends today
by Gordon Woodhull 01 Sep '11

01 Sep '11

Folks, The review of Vicente's Conversion library ends today. Unfortunately there have only been four reviews, which is hardly enough to provide proper feedback. If you've considered writing a review or if you have any thoughts about the topic, please speak up now. Or if you'd like to submit a late review, just let me know. Thanks, Gordon

5 19