Hi Guys,
Steven is correct. After adding the following copy contructor to group_helper class:
group_helper(const group_helper& gp) { std::cout << "group_helper::copy ctor called" << std::endl; }
The number of ctors matches that of the dtors. So the threaded code is correct thanks to boost::shared_ptr.
Cheers,
Robert
On Sun, Aug 23, 2009 at 10:05 PM, Mathias Gaunard
<mathias.gaunard@ens-lyon.org> wrote:
Boost lzw wrote:
class group_helper { private:
boost::shared_ptr<FileThread> m_object; public: group_helper() : m_object(boost::shared_ptr<FileThread>(new FileThread))
{ std::cout << "group_helper::ctor called" << std::endl;}
void operator()() { m_object->find(); m_object->print(); }
~group_helper() { std::cout << "group_helper::***dtor called" << std::endl; }
};
int main()
{
group_helper gp;
boost::thread_group threads;
for (int i = 0; i < 1; ++i)
threads.create_thread(gp); threads.join_all();
return 0;
}
===========
Output: 1 call to "group_helper::ctor called"
6 calls to "group_helper::***dtor called"
===========
If the code is correct, does this mean that with thread_group we can no longer expect the # of ctors matches that of the dtors in threaded code?