2014-05-30 3:08 GMT+02:00 Glen Fernandes :

On Thu, May 29, 2014 at 9:19 AM, Adam Wulkiewicz wrote:

...

Hi,

I remember that there was plans to release an "official" STD-conformant implementation of boost::unique_ptr<>. I found: http://boost.2283326.n4.nabble.com/unique-ptr-for-C-03-td4650231.html

http://boost.2283326.n4.nabble.com/C-03-unique-ptr-emulation-td2658449.html

...

Though both threads are about a C++03 emulation of unique_ptr, I wonder if it is worth providing a boost::unique_ptr for C++ implementations still in use that support enough of C++11 r-value references but ship with a broken, or non-conforming, std::unique_ptr implementation.

[Such as VC10 and VC11 whose std::unique_ptr has a broken reset() or whose std::unique_ptr has a broken conversion to bool when a different deleter type is specified. I believe there are a few bugs reported about VC12's std::unique_ptr but I haven't looked into them.]

Actually I had both in mind. If r-val refs were supported it should be STD-conformant implementation, if not, the emulation e.g. using Boost.Move. Plus boost::make_unique() using r-val refs and variadic templates if they were supported or emulating them e.g. using Move and Preprocessor. Someone could say that we should move forward and use C++11/14 but the reality is that many projects are bound to some old compiler versions. Regards, Adam

[Glen Fernandes]

[Such as VC10 and VC11 whose std::unique_ptr has a broken reset() or whose std::unique_ptr has a broken conversion to bool when a different deleter type is specified. I believe there are a few bugs reported about VC12's std::unique_ptr but I haven't looked into them.]

I am not aware of any bugs in 2013's unique_ptr, except for its empty deleter optimization (I have local changes to make this conformant for the next major version), and specification/implementation issues in shared_ptr(unique_ptr). STL