Re: [boost] ScopeExit Review Begins Today

From: Ilya Sokolov (fal_delivery_at_[hidden])
Date: 2007-08-24 07:10:57

Andrey Semashev wrote:

 > Consider the following examples:
 >
 > Example 1:
 >
 > void foo(std::set< int >& my_set)
 > {
 >   bool commit = false;
 >   std::set< int >::iterator it = my_set.insert(10);
 >   BOOST_SCOPE_EXIT((my_set)(it)(commit))
 >   {
 >     // Which one will be erased here?
 >     // Not necessarily the one we intended to.

Sorry for such late response.
Can you explain what is intended value of it?

 >     if (!commit)
 >       my_set.erase(it);
 >   }
 >   BOOST_SCOPE_EXIT_END
 >
 >   // Do smth. with "it"
 >
 >   it = my_set.find(5); // and reuse it
 >
 >   // Do smth. with "it" again
 >
 >   commit = true;
 > }

 > Example 2:
 >
 > class A
 > {
 >   int m_i;
 >
 >   void foo()
 >   {
 >     m_i = 1;
 >     BOOST_SCOPE_EXIT((m_i))
 >     {
 >       // What's the value of m_i here?
 >       std::cout << m_i << std::endl;
 >     }
 >     BOOST_SCOPE_EXIT_END
 >
 >     bar();
 >   }

I believe BOOST_SCOPE_EXIT should be moved to the end of foo(). I am
thinking about BOOST_SCOPE_EXIT as of try/finally with different layout
of braces:

void baz()
{ // 0
   try
   { // 1
       foo();
   } // 2
   finally
   { // 3
       bar();
   } // 4
} // 5

void baz()
{ // 0 and 1
   foo();

   BOOST_SCOPE_EXIT()
   { // 3
       bar();
   } // 4
   BOOST_SCOPE_EXIT_END
} // 5 and 2

 >   // Assume that bar's implementation is somewhere far away
 >   void bar();
 > };
 >
 > These are just illustrations, I understand that I could have created
 > copies of the needed variables on the stack. But I think this should
 > be the default behavior of the library, and passing by reference
 > should be an available option.

[snip]