Emacs, C++ code completion for vectors

I am new to Emacs, and I have the following code as a sample. I have installed GNU Emacs 23.1.1 (i386-mingw-nt6.1.7600), installed cedet-1.0pre7.tar.gz. , installed ELPA, and company. You can find my simple Emacs configuration at the bottom.

The problem is, when I type q[0] in main() and press . (dot), I see the 37 members of the vector, not Person although first_name and last_name are expected. The completion works as expected in the function greet() but it has nothing to do with vector.

My question is, how can I accomplish code completion for vector elements too?

#include <iostream>
#include <vector>
using namespace std;

class Person
{
  public:
    string first_name;
    string last_name;
};

void greet(Person a_person)
{
  // a_person.first_name is completed as expected!
  cout << a_person.first_name << "|";
  cout << a_person.last_name << endl;
};

int main()
{
  vector<Person> q(2);

  Person guy1;
  guy1.first_name = "foo";
  guy1.last_name = "bar";

  Person guy2;
  guy2.first_name = "stack";
  guy2.last_name = "overflow";

  q[0] = guy1;
  q[1] = guy2;
  greet(guy1);
  greet(guy2);
  // cout q[0]. I want to see first_name or last_name here!
}

My Emacs configuration:

;;; This was installed by package-install.el.
;;; This provides support for the package system and
;;; interfacing with ELPA, the package archive.
;;; Move this code earlier if you want to reference
;;; packages in your .emacs.
(when
    (load
     (expand-file-name "~/.emacs.d/elpa/package.el"))
  (package-initialize))

(load-file "~/.emacs.d/cedet/common/cedet.el")
(semantic-load-enable-excessive-code-helpers)
(require 'semantic-ia)

(global-srecode-minor-mode 1)
(semantic-add-system-include "/gcc/include/c++/4.4.2" 'c++-mode)
(semantic-add-system-include "/gcc/i386-pc-mingw32/include" 'c++-mode)
(semantic-add-system-include "/gcc/include" 'c++-mode)


(defun my-semantic-hook ()
  (imenu-add-to-menubar "TAGS"))
(add-hook 'semantic-init-hooks 'my-semantic-hook)

This is a known problem with the Semantic analyzer. I currently cannot deal with Template Specialization, which is used in the gcc STL (your problem stems from such a specialization in allocator.h). This has been discussed on the mailing list:

http://thread.gmane.org/gmane.emacs.semantic/2137/focus=2147