c++ - Can boost::lockfree::stack accept string pointers safely? -

i've discovered passing string pointer boost::lockfree::queue cause memory leak because string pointer cannot released.

is situation same boost::lockfree::stack?

the requirement boost::lockfree::stack is:

• t must have copy constructor

if regular string pointer cannot used, there other way string can put boost::lockfree::stack?

regular string

when try this

boost::lockfree::stack<string> my_stack(128); 

i these errors

 boost_static_assert(boost::has_trivial_assign<t>::value);  boost_static_assert(boost::has_trivial_destructor<t>::value); 

which seems strange me, because of inexperience, because requirements boost::lockfree::queue strangely has no documentation

• t must have copy constructor
• t must have trivial assignment operator
• t must have trivial destructor

1. the missing documentation doxygen bug, documentation page gets name macro :(, , here: http://www.boost.org/doc/libs/1_55_0/doc/html/boost/lockfree/boost_no_cxx1_idp100289128.html

2. indeed stack appears have same element type requirements queue, despite fact documentation doesn't mention it. what's worse, seems boost::string_ref hasn't been kept pod either.

so suggestion store pointers "pool" of immutable string objects , free pool @ time when queue freed. way, have manage lifetime of queue in such way memory consumption doesn't grow out of control.

in ways resembles implementing poor-man's gc, , yes, will force regular delays in throughput re-initialize queue; in fairness common field of lock-less memory management, , boost lockfree documentation states, "viable" memory reclamation schemes in existence patented.

you might have @ libcds: lock-free concurrent datastructures in case want toy of these algorithms. think author of library has been in active conversation boost developers proposing libcds boost, , did not take place because of patent issues.