[boost] [utility] A fix for the const issue of value_initialized

Subject: [boost] [utility] A fix for the const issue of value_initialized
From: Niels Dekker - mail address until 2010-10-10 (niels_address_until_2010-10-10_at_[hidden])
Date: 2009-02-17 14:04:10

The boost::value_initialized<T> utility has a major issue regarding 
const-correctness: It has a const-qualified conversion operator, 
returning a non-const reference to the T object it owns. I have already 
created a ticket, <https://svn.boost.org/trac/boost/ticket/2548>, and 
mailed Fernando Cacciola (the author of value_initialized) on this 
issue. He agrees that it should be fixed, and he allows me to do so. The 
patch I've attached to the ticked replaces the current conversion 
operator by two of them, overloaded for const and non-const access:

  operator T const &() const { return this->data(); }
  operator T&() { return this->data(); }

The patch also fixes value_initialized::data() in a similar way.  Please 
let me know if you have any problem with the patch. Otherwise I'd like 
to apply it to the trunk by the end of the week: 
https://svn.boost.org/trac/boost/attachment/ticket/2548/value_initialized_constness.patch

Note that the issue has been discussed here already five years ago: 
http://lists.boost.org/Archives/boost/2004/01/59987.php  At that time, 
some non-conforming compilers couldn't deal with conversion operators 
overloaded for const and non-const access.  But now I've tested the 
value_initialized patch on a few compilers (MSVC 2003, MSVC 2008, 
CodeGear), and they all succeeded.  After the commit, I'll keep a close 
look at the regression page, of course.

Kind regards, Niels

--
Niels Dekker
http://www.xs4all.nl/~nd/dekkerware
Scientific programmer at LKEB, Leiden University Medical Center