future.h

Go to the documentation of this file.

/* Copyright (C) 2010 to 2015 Chris Vine
 
The library comprised in this file or of which this file is part is
distributed by Chris Vine under the GNU Lesser General Public
License as follows:
 
   This library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public License
   as published by the Free Software Foundation; either version 2.1 of
   the License, or (at your option) any later version.
 
   This library is distributed in the hope that it will be useful, but
   WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Lesser General Public License, version 2.1, for more details.
 
   You should have received a copy of the GNU Lesser General Public
   License, version 2.1, along with this library (see the file LGPL.TXT
   which came with this source code package in the src/utils sub-directory);
   if not, write to the Free Software Foundation, Inc.,
   51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
However, it is not intended that the object code of a program whose
source code instantiates a template from this file or uses macros or
inline functions (of any length) should by reason only of that
instantiation or use be subject to the restrictions of use in the GNU
Lesser General Public License.  With that in mind, the words "and
macros, inline functions and instantiations of templates (of any
length)" shall be treated as substituted for the words "and small
macros and small inline functions (ten lines or less in length)" in
the fourth paragraph of section 5 of that licence.  This does not
affect any other reason why object code may be subject to the
restrictions in that licence (nor for the avoidance of doubt does it
affect the application of section 2 of that licence to modifications
of the source code in this file).
 
*/
 
#ifndef CGU_FUTURE_H
#define CGU_FUTURE_H
 
#include <memory>
#include <exception>
 
#include <pthread.h>
#include <glib.h>
 
#include <c++-gtk-utils/thread.h>
#include <c++-gtk-utils/mutex.h>
#include <c++-gtk-utils/callback.h>
#include <c++-gtk-utils/param.h>
#include <c++-gtk-utils/intrusive_ptr.h>
#include <c++-gtk-utils/shared_ptr.h>
#include <c++-gtk-utils/emitter.h>
#include <c++-gtk-utils/timeout.h>
#include <c++-gtk-utils/cgu_config.h>
 
namespace Cgu {
 
namespace Thread {
 
struct FutureThreadError: public std::exception {
  virtual const char* what() const throw() {return "FutureThreadError\n";}
};
 
struct FutureWhenError: public std::exception {
  virtual const char* what() const throw() {return "FutureWhenError\n";}
};
 
/**
 * @class Cgu::Thread::Future future.h c++-gtk-utils/future.h
 * @brief A class representing a pthread thread which will
 * provide a value.
 * @sa Cgu::Thread::Thread Cgu::Thread::JoinableHandle Cgu::AsyncResult Cgu::Thread::TaskManager
 *
 * The Thread::Future class will launch a worker thread, run the
 * function it represents in that thread until it returns, and store
 * the return value so that it can be waited on and/or extracted by
 * another thread.  A new Thread::Future object representing the
 * function to be called is created by calling make().  The worker
 * thread is then started by calling run(), and the value extracted or
 * waited for by calling get().  The run() method can only be called
 * once, but any number of threads can wait for and/or extract the
 * return value by calling the get() method.  The class also provides
 * a SafeEmitter @ref DoneEmitterAnchor "done_emitter" public object
 * which emits when the worker thread has finished, and an associated
 * when() function.
 *
 * The template parameter type is the type of the return value of the
 * function called by the Thread::Future object.  The return value can
 * be any type, including any arbitrarily large tuple or other struct
 * or standard C++ container (but see below about copying).
 *
 * A Thread::Future object cannot represent a function with a void
 * return type - a compilation error will result if that is attempted.
 * If no return value is wanted, then the Thread::Thread class can be
 * used directly.  (However, if in a particular usage this class is
 * thought to be more convenient, the function to be represented by it
 * can be wrapped by another function which provides a dummy return
 * value, such as a dummy int.  One possible case for this is where
 * more than one thread wants to wait for the worker thread to
 * terminate, as pthread_join() and so Thread::Thread::join() only
 * give defined behaviour when called by one thread.)  In addition, a
 * Thread::Future object cannot represent a function with a non-const
 * reference argument - that would not normally be safe, and a compile
 * error will be generated if that is attempted.  However, from
 * version 1.2.13, const reference arguments can be bound to
 * Thread::Future objects (this is safe, as the Thread::Future object
 * will keep its own copy of the argument passed to it).
 * 
 * The make() method takes a plain function, static member function or
 * non-static member function, and the function can take up to two
 * arguments.  In the case of a non-static member function, the
 * referenced object whose member function is to be called must remain
 * in existence until the worker thread has completed.  It works in a
 * similar way to the Callback::make() factory function, except that
 * Thread::Future<>::make() returns the Thread::Future object by
 * Cgu::IntrusivePtr<Cgu::Thread::Future<> >.
 *
 * It is to be noted that the target function to be represented by a
 * Thread::Future object must not allow any exception other than
 * Thread::Exit, an exception deriving from std::exception or a
 * cancellation pseudo-exception to escape from it when it is
 * executed.  This includes ensuring that, for any argument of that
 * function which is of class type and not taken by reference to
 * const, the argument type's copy constructor does not throw anything
 * other than these, and that the assignment operator of the return
 * value (if of class type) of the target function does not throw
 * anything other than these.  (If the target function or the copy
 * constructor of a value argument or the assignment operator of the
 * return value throws Thread::Exit or an exception deriving from
 * std::exception, the exception is safely consumed and the
 * Thread::Future object's error flag is set.  However, if the
 * assignment operator of the return value throws, it should leave the
 * assignee in a state in which it can safely be destroyed and in
 * which, if that assignee is further copied from, the copy either
 * throws an exception or produces an object which can also be
 * destroyed -- but these are minimum requirements for any reasonable
 * assignment operator, and met by any assignment operator offering
 * the basic exception guarantee.)
 *
 * Copies of the objects to be passed to the target function as
 * arguments are taken internally by the implementation to avoid
 * dangling references.  This copying of arguments takes place twice
 * when make() is called and, if the target function does not take an
 * argument by reference to const, once in the worker thread when
 * run() is called.  If such an argument is a pointer or a fundamental
 * type (an integral or floating-point type), optimization by copy
 * elision will reduce the number of times argument copying takes
 * place when make() is called to once, but the standard does not
 * permit that with class types where the constructor or destructor
 * have side effects or it cannot be ascertained whether they have
 * side effects.  Therefore, if class objects are received by the
 * target function as arguments, it is best for them to be constructed
 * on free store and for the target function to receive them by
 * pointer or by Cgu::SharedLockPtr.
 *
 * Copying of the return value of the target function represented by
 * the Thread::Future object also takes place.  The Thread::Future
 * object will store the return value of that function, so that it is
 * available to the get() method and any 'when' callback, and
 * therefore copy it once (unless, that is, the program is compiled
 * under C++11 and the target function's return value type has a move
 * assignment operator, in which case that operator is called).  For
 * safety reasons, the get() method returns by value and so will cause
 * it to be copied once more, so for return values comprising complex
 * class objects which are to be extracted using the get() method, it
 * is often better if the function represented by the Thread::Future
 * object allocates the return value on free store and returns it by
 * pointer or Cgu::SharedLockPtr (although the example below does not
 * do that).
 *
 * It should be noted that where the when() method is used, the return
 * value is passed to the 'when' callback by reference to const and so
 * without the copying carried out by the get() method: therefore, if
 * the program is compiled under C++11, the return value has a move
 * assignment operator and the when() method is to be employed, and
 * the 'when' callback only needs to call const methods of the return
 * value, it may be more efficient not to allocate the return value on
 * free store.
 *
 * This is a usage example:
 *
 * @code 
 *   class Numbers {
 *   public:
 *     std::vector<long> get_primes(int n); // calculates the first n primes
 *                                          // and puts them in a vector
 *    ...
 *   };
 *
 *   Numbers obj;
 *
 *   // get the first 1,000 primes
 *   using namespace Cgu;
 *
 *   IntrusivePtr<Thread::Future<std::vector<long> > > future = 
 *     Thread::Future<std::vector<long> >::make(obj, &Numbers::get_primes, 1000);
 *   [or with C++11 the more pleasing:
 *   auto future = Thread::Future<std::vector<long>>::make(obj, &Numbers::get_primes, 1000);]
 *
 *   future->run();
 *   ... [ do something else ] ...
 *   std::vector<long> result(future->get());
 *   for (std::vector<long>::const_iterator i = result.begin();
 *        i != result.end();
 *        ++i)
 *     std::cout << *i << std::endl;
 * @endcode
 *
 * If get_primes() were a static member function or plain function,
 * the syntax would be:
 *
 * @code 
 *   IntrusivePtr<Thread::Future<std::vector<long> > > future = 
 *     Thread::Future<std::vector<long> >::make(&Numbers::get_primes, 1000);
 *   or with C++11:
 *   auto future = Thread::Future<std::vector<long>>::make(&Numbers::get_primes, 1000);
 * @endcode
 *
 * The Cgu::Thread::Future::when() functions
 * -----------------------------------------
 *
 * From version 1.2.16, the return value of the thread function
 * represented by Cgu::Thread::Future can be obtained asynchronously
 * using Cgu::Thread::Future::when() to execute a function in a glib
 * main loop when the thread function completes.  The above example
 * could be reimplemented as:
 *
 * @code 
 *   class Numbers {
 *   public:
 *     std::vector<long> get_primes(int n); // calculates the first n primes
 *                                          // and puts them in a vector
 *    ...
 *   };
 *
 *   void print_primes(const std::vector<long>& result) {
 *     for (std::vector<long>::const_iterator i = result.begin();
 *          i != result.end();
 *          ++i)
 *       std::cout << *i << std::endl;
 *   }
 *
 *   Numbers obj;
 *
 *   using namespace Cgu;
 *
 *   IntrusivePtr<Thread::Future<std::vector<long> > > future = 
 *     Thread::Future<std::vector<long> >::make(&Numbers::get_primes, 1000);
 *   future->when(Callback::make(&print_primes));
 *   future->run();
 * @endcode
 *
 * In this example, the callback which prints the primes to the
 * console would execute in the default program main loop once the
 * thread function providing those primes returns.
 *
 * The Cgu::Thread::Future::fail() functions
 * -----------------------------------------
 * 
 * The Thread::Future::when() functions have an associated optional
 * Thread::Future::fail() function which causes a 'fail' callback to
 * execute in a glib main loop in the event of certain exceptions
 * arising in executing the thread function or a thread being
 * cancelled (the documentation on Thread::Future::fail() gives
 * further details).  The 'fail' callback must be fully bound.  Whilst
 * a worker thread can pass error status to the 'fail' callback via
 * shared data bound to both the thread function and the 'fail'
 * callback (held by, say, a SharedLockPtr object), or a global error
 * stack, 'fail' callbacks are generally best reserved either for use
 * with entirely unexpected exceptions, where the most reasonable
 * course is to perform some orderly logging and shutdown, or to
 * report thread cancellation.  For handlable exceptions, in an
 * asynchronous environment the best course is often to catch them and
 * deal with them in the thread function itself and return a value of
 * the return type for the 'when' callback indicating no result.
 *
 * Overloaded functions
 * --------------------
 *
 * Where a member function or ordinary function represented by a
 * Thread::Future object is overloaded, this may cause difficulties in
 * template type deduction when Thread::Future<>::make() is called.
 * Functions may be overloaded on numbers of argument without
 * difficulty.  For example:
 * @code
 *   class Numbers {
 *   public:
 *     int calc(int i);
 *     int calc(int i, int j);
 *    ...
 *   };
 *
 *   Numbers obj;
 *
 *   using namespace Cgu;
 *
 *   int i = 1, j = 2;
 *
 *   IntrusivePtr<Thread::Future<int> > f1 = 
 *     Thread::Future<int>::make(obj, &Numbers::calc, i);     // OK
 *   IntrusivePtr<Thread::Future<int> > f2 = 
 *     Thread::Future<int>::make(obj, &Numbers::calc, i, j);  // OK
 * @endcode
 * However, they cannot be overloaded on types without explicit
 * disambiguation when Thread::Future<>::make() is called.  For
 * example the following will fail to compile unless explicitly
 * disambiguated:
 * @code
 *   class Numbers {
 *   public:
 *     int calc(int i);
 *     int calc(double d);
 *    ...
 *   };
 *
 *   Numbers obj;
 *
 *   using namespace Cgu;
 *
 *   int i = 1;
 *   double d = 2.0;
 *
 *   IntrusivePtr<Thread::Future<int> > f1 = 
 *     Thread::Future<int>::make(obj, &Numbers::calc, i);                   // won't compile
 *   IntrusivePtr<Thread::Future<int> > f2 = 
 *     Thread::Future<int>::make(obj, &Numbers::calc, d);;                  // won't compile
 *   IntrusivePtr<Thread::Future<int> > f3 = 
 *     Thread::Future<int>::make(obj, static_cast<int (Numbers::*)(int)>(&Numbers::calc), i);    // OK
 *   IntrusivePtr<Thread::Future<int> > f4 = 
 *     Thread::Future<int>::make(obj, static_cast<int (Numbers::*)(double)>(&Numbers::calc), d); // OK
 * @endcode
 */
 
template <class Val>
class Future: public IntrusiveLockCounter {
 
  std::auto_ptr<Cgu::Thread::Thread> thread_a;
  std::auto_ptr<Cgu::Callback::Callback> cb_a;
 
  mutable Mutex mutex;
  Cond cond;
  Val val;
  bool done;
  bool running;
  bool error;
  bool emitter_error;
 
#ifndef DOXYGEN_PARSING
 
  class Empty {};
 
  template<class Obj, class Func>
  struct arg_pair {
    Obj obj;
    Func func;
    arg_pair(Obj obj_, Func func_):
      obj(obj_), func(func_) {}
  };
 
  template<class Obj, class Func, class Arg1>
  struct arg_triple {
    Obj obj;
    Func func;
    typename Cgu::RemoveRefCond<Arg1, true>::Type arg1;
    arg_triple(Obj obj_, Func func_, typename Cgu::Param<Arg1>::ParamType arg1_):
      obj(obj_), func(func_), arg1(arg1_) {}
  };
 
  template<class Obj, class Func, class Arg1, class Arg2>
  struct arg_quadruple {
    Obj obj;
    Func func;
    typename Cgu::RemoveRefCond<Arg1, true>::Type arg1;
    typename Cgu::RemoveRefCond<Arg2, true>::Type arg2;
    arg_quadruple(Obj obj_, Func func_, typename Cgu::Param<Arg1>::ParamType arg1_,
          typename Cgu::Param<Arg2>::ParamType arg2_):
      obj(obj_), func(func_), arg1(arg1_), arg2(arg2_) {}
  };
 
  struct when_mainloop_args {
    gint priority;
    GMainContext* context;
    when_mainloop_args(gint p, GMainContext* c): priority(p), context(c) {}
  };
 
#endif
 
 /* All the run_wrapper() methods are static.  This is because gcc-4.2
  * has problems with instantiating a non-static nested template
  * member function from a static nested template member function
  * (Future<>::make())
  */
  template <class Ret, class T>
  static void run_wrapper(const arg_pair<T&, Ret (T::*)()>& args,
              Future<Val>* instance);
 
  template <class Ret, class Arg1, class T>
  static void run_wrapper(const arg_triple<T&, Ret (T::*)(Arg1), Arg1>& args,
              Future<Val>* instance);
 
  template <class Ret, class Arg1, class Arg2, class T>
  static void run_wrapper(const arg_quadruple<T&, Ret (T::*)(Arg1, Arg2), Arg1, Arg2>& args,
              Future<Val>* instance);
 
  template <class Ret, class T>
  static void run_wrapper_const(const arg_pair<const T&, Ret (T::*)() const>& args,
                Future<Val>* instance);
 
  template <class Ret, class Arg1, class T>
  static void run_wrapper_const(const arg_triple<const T&, Ret (T::*)(Arg1) const, Arg1>& args,
                Future<Val>* instance);
 
  template <class Ret, class Arg1, class Arg2, class T>
  static void run_wrapper_const(const arg_quadruple<const T&, Ret (T::*)(Arg1, Arg2) const, Arg1, Arg2>& args,
                Future<Val>* instance);
 
  template <class Ret>
  static void run_wrapper_static(Ret (*func)(),
                 Future<Val>* instance);
 
  template <class Ret, class Arg1>
  static void run_wrapper_static(const arg_triple<Empty, Ret (*)(Arg1), Arg1>& args,
                 Future<Val>* instance);
 
  template <class Ret, class Arg1, class Arg2>
  static void run_wrapper_static(const arg_quadruple<Empty, Ret (*)(Arg1, Arg2), Arg1, Arg2>& args,
                 Future<Val>* instance);
 
  void cancel_cleanup();
 
  void execute_done(const Cgu::Callback::SafeFunctorArg<const Val&>&);
  void post_done(const Cgu::Callback::SafeFunctorArg<const Val&>&,
         const when_mainloop_args&);
  void execute_done_rel(const Cgu::SharedLockPtr<Cgu::SafeEmitterArg<const Val&> >&);
  void post_done_rel(const Cgu::SharedLockPtr<Cgu::SafeEmitterArg<const Val&> >&,
             const when_mainloop_args&);
 
  // this is a static function taking the future object by IntrusivePtr to
  // ensure that the future object remains in existence whilst this
  // function might execute
  static void fail_cb(const Cgu::IntrusivePtr<Future<Val> >& future,
              const Cgu::Callback::SafeFunctor& func,
              bool& ret);
 
  // private constructor - this class can only be created with Thread::Future::make()
  Future(): val(), done(false), running(false), error(false), emitter_error(false) {}
  // and it cannot be copied except by smart pointer
  Future(const Future&);
  Future& operator=(const Future&);
 
public:
 
/**
 * Constructs a new Cgu::Thread::Future object (returned by
 * Cgu::IntrusivePtr<Cgu::Thread::Future<Val> >).  The type parameter
 * Val represents the return value of the function to be represented
 * by the new object.
 * @exception std::bad_alloc It might throw std::bad_alloc if memory
 * is exhausted and the system throws in that case.  (This exception
 * will not be thrown if the library has been installed using the
 * \--with-glib-memory-slices-no-compat configuration option: instead
 * glib will terminate the program if it is unable to obtain memory
 * from the operating system.)
 * @exception Cgu::Thread::MutexError It might throw
 * Cgu::Thread::MutexError if initialisation of the contained mutex
 * fails.  (It is often not worth checking for this, as it means
 * either memory is exhausted or pthread has run out of other
 * resources to create new mutexes.)
 * @exception Cgu::Thread::CondError It might throw
 * Cgu::Thread::CondError if initialisation of the contained condition
 * variable fails.  (It is often not worth checking for this, as it
 * means either memory is exhausted or pthread has run out of other
 * resources to create new condition variables.)
 * @note This method will also throw if the default constructor of the
 * return value type throws.
 *
 * Since 1.0.2
 */
  template <class Ret, class T>
  static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > make(T& t,
                               Ret (T::*func)());
 
/**
 * Constructs a new Cgu::Thread::Future object (returned by
 * Cgu::IntrusivePtr<Cgu::Thread::Future<Val> >).  The type parameter
 * Val represents the return value of the function to be represented
 * by the new object.
 * @exception std::bad_alloc It might throw std::bad_alloc if memory
 * is exhausted and the system throws in that case.  (This exception
 * will not be thrown if the library has been installed using the
 * \--with-glib-memory-slices-no-compat configuration option: instead
 * glib will terminate the program if it is unable to obtain memory
 * from the operating system.)
 * @exception Cgu::Thread::MutexError It might throw
 * Cgu::Thread::MutexError if initialisation of the contained mutex
 * fails.  (It is often not worth checking for this, as it means
 * either memory is exhausted or pthread has run out of other
 * resources to create new mutexes.)
 * @exception Cgu::Thread::CondError It might throw
 * Cgu::Thread::CondError if initialisation of the contained condition
 * variable fails.  (It is often not worth checking for this, as it
 * means either memory is exhausted or pthread has run out of other
 * resources to create new condition variables.)
 * @note This method will also throw if the copy constructor of the
 * bound argument throws, or the default constructor of the return
 * value type throws.
 *
 * Since 1.0.2
 */
  template <class Ret, class Arg1, class T>
  static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > make(T& t,
                               Ret (T::*func)(Arg1),
                               typename Cgu::Param<Arg1>::ParamType arg1);
 
/**
 * Constructs a new Cgu::Thread::Future object (returned by
 * Cgu::IntrusivePtr<Cgu::Thread::Future<Val> >).  The type parameter
 * Val represents the return value of the function to be represented
 * by the new object.
 * @exception std::bad_alloc It might throw std::bad_alloc if memory
 * is exhausted and the system throws in that case.  (This exception
 * will not be thrown if the library has been installed using the
 * \--with-glib-memory-slices-no-compat configuration option: instead
 * glib will terminate the program if it is unable to obtain memory
 * from the operating system.)
 * @exception Cgu::Thread::MutexError It might throw
 * Cgu::Thread::MutexError if initialisation of the contained mutex
 * fails.  (It is often not worth checking for this, as it means
 * either memory is exhausted or pthread has run out of other
 * resources to create new mutexes.)
 * @exception Cgu::Thread::CondError It might throw
 * Cgu::Thread::CondError if initialisation of the contained condition
 * variable fails.  (It is often not worth checking for this, as it
 * means either memory is exhausted or pthread has run out of other
 * resources to create new condition variables.)
 * @note This method will also throw if the copy constructor of a
 * bound argument throws, or the default constructor of the return
 * value type throws.
 *
 * Since 1.0.2
 */
  template <class Ret, class Arg1, class Arg2, class T>
  static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > make(T& t,
                               Ret (T::*func)(Arg1, Arg2),
                               typename Cgu::Param<Arg1>::ParamType arg1,
                               typename Cgu::Param<Arg2>::ParamType arg2);
 
/**
 * Constructs a new Cgu::Thread::Future object (returned by
 * Cgu::IntrusivePtr<Cgu::Thread::Future<Val> >).  The type parameter
 * Val represents the return value of the function to be represented
 * by the new object.
 * @exception std::bad_alloc It might throw std::bad_alloc if memory
 * is exhausted and the system throws in that case.  (This exception
 * will not be thrown if the library has been installed using the
 * \--with-glib-memory-slices-no-compat configuration option: instead
 * glib will terminate the program if it is unable to obtain memory
 * from the operating system.)
 * @exception Cgu::Thread::MutexError It might throw
 * Cgu::Thread::MutexError if initialisation of the contained mutex
 * fails.  (It is often not worth checking for this, as it means
 * either memory is exhausted or pthread has run out of other
 * resources to create new mutexes.)
 * @exception Cgu::Thread::CondError It might throw
 * Cgu::Thread::CondError if initialisation of the contained condition
 * variable fails.  (It is often not worth checking for this, as it
 * means either memory is exhausted or pthread has run out of other
 * resources to create new condition variables.)
 * @note This method will also throw if the default constructor of the
 * return value type throws.
 *
 * Since 1.0.2
 */
  template <class Ret, class T>
  static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > make(const T& t,
                               Ret (T::*func)() const);
 
/**
 * Constructs a new Cgu::Thread::Future object (returned by
 * Cgu::IntrusivePtr<Cgu::Thread::Future<Val> >).  The type parameter
 * Val represents the return value of the function to be represented
 * by the new object.
 * @exception std::bad_alloc It might throw std::bad_alloc if memory
 * is exhausted and the system throws in that case.  (This exception
 * will not be thrown if the library has been installed using the
 * \--with-glib-memory-slices-no-compat configuration option: instead
 * glib will terminate the program if it is unable to obtain memory
 * from the operating system.)
 * @exception Cgu::Thread::MutexError It might throw
 * Cgu::Thread::MutexError if initialisation of the contained mutex
 * fails.  (It is often not worth checking for this, as it means
 * either memory is exhausted or pthread has run out of other
 * resources to create new mutexes.)
 * @exception Cgu::Thread::CondError It might throw
 * Cgu::Thread::CondError if initialisation of the contained condition
 * variable fails.  (It is often not worth checking for this, as it
 * means either memory is exhausted or pthread has run out of other
 * resources to create new condition variables.)
 * @note This method will also throw if the copy constructor of the
 * bound argument throws, or the default constructor of the return
 * value type throws.
 *
 * Since 1.0.2
 */
  template <class Ret, class Arg1, class T>
  static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > make(const T& t,
                               Ret (T::*func)(Arg1) const,
                               typename Cgu::Param<Arg1>::ParamType arg1);
 
/**
 * Constructs a new Cgu::Thread::Future object (returned by
 * Cgu::IntrusivePtr<Cgu::Thread::Future<Val> >).  The type parameter
 * Val represents the return value of the function to be represented
 * by the new object.
 * @exception std::bad_alloc It might throw std::bad_alloc if memory
 * is exhausted and the system throws in that case.  (This exception
 * will not be thrown if the library has been installed using the
 * \--with-glib-memory-slices-no-compat configuration option: instead
 * glib will terminate the program if it is unable to obtain memory
 * from the operating system.)
 * @exception Cgu::Thread::MutexError It might throw
 * Cgu::Thread::MutexError if initialisation of the contained mutex
 * fails.  (It is often not worth checking for this, as it means
 * either memory is exhausted or pthread has run out of other
 * resources to create new mutexes.)
 * @exception Cgu::Thread::CondError It might throw
 * Cgu::Thread::CondError if initialisation of the contained condition
 * variable fails.  (It is often not worth checking for this, as it
 * means either memory is exhausted or pthread has run out of other
 * resources to create new condition variables.)
 * @note This method will also throw if the copy constructor of a
 * bound argument throws, or the default constructor of the return
 * value type throws.
 *
 * Since 1.0.2
 */
  template <class Ret, class Arg1, class Arg2, class T>
  static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > make(const T& t,
                               Ret (T::*func)(Arg1, Arg2) const,
                               typename Cgu::Param<Arg1>::ParamType arg1,
                               typename Cgu::Param<Arg2>::ParamType arg2);
 
/**
 * Constructs a new Cgu::Thread::Future object (returned by
 * Cgu::IntrusivePtr<Cgu::Thread::Future<Val> >).  The type parameter
 * Val represents the return value of the function to be represented
 * by the new object.
 * @exception std::bad_alloc It might throw std::bad_alloc if memory
 * is exhausted and the system throws in that case.  (This exception
 * will not be thrown if the library has been installed using the
 * \--with-glib-memory-slices-no-compat configuration option: instead
 * glib will terminate the program if it is unable to obtain memory
 * from the operating system.)
 * @exception Cgu::Thread::MutexError It might throw
 * Cgu::Thread::MutexError if initialisation of the contained mutex
 * fails.  (It is often not worth checking for this, as it means
 * either memory is exhausted or pthread has run out of other
 * resources to create new mutexes.)
 * @exception Cgu::Thread::CondError It might throw
 * Cgu::Thread::CondError if initialisation of the contained condition
 * variable fails.  (It is often not worth checking for this, as it
 * means either memory is exhausted or pthread has run out of other
 * resources to create new condition variables.)
 * @note This method will also throw if the default constructor of the
 * return value type throws.
 *
 * Since 1.0.2
 */
  template <class Ret>
  static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > make(Ret (*func)());
 
/**
 * Constructs a new Cgu::Thread::Future object (returned by
 * Cgu::IntrusivePtr<Cgu::Thread::Future<Val> >).  The type parameter
 * Val represents the return value of the function to be represented
 * by the new object.
 * @exception std::bad_alloc It might throw std::bad_alloc if memory
 * is exhausted and the system throws in that case.  (This exception
 * will not be thrown if the library has been installed using the
 * \--with-glib-memory-slices-no-compat configuration option: instead
 * glib will terminate the program if it is unable to obtain memory
 * from the operating system.)
 * @exception Cgu::Thread::MutexError It might throw
 * Cgu::Thread::MutexError if initialisation of the contained mutex
 * fails.  (It is often not worth checking for this, as it means
 * either memory is exhausted or pthread has run out of other
 * resources to create new mutexes.)
 * @exception Cgu::Thread::CondError It might throw
 * Cgu::Thread::CondError if initialisation of the contained condition
 * variable fails.  (It is often not worth checking for this, as it
 * means either memory is exhausted or pthread has run out of other
 * resources to create new condition variables.)
 * @note This method will also throw if the copy constructor of the
 * bound argument throws, or the default constructor of the return
 * value type throws.
 *
 * Since 1.0.2
 */
  template <class Ret, class Arg1>
  static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > make(Ret (*func)(Arg1),
                               typename Cgu::Param<Arg1>::ParamType arg1);
/**
 * Constructs a new Cgu::Thread::Future object (returned by
 * Cgu::IntrusivePtr<Cgu::Thread::Future<Val> >).  The type parameter
 * Val represents the return value of the function to be represented
 * by the new object.
 * @exception std::bad_alloc It might throw std::bad_alloc if memory
 * is exhausted and the system throws in that case.  (This exception
 * will not be thrown if the library has been installed using the
 * \--with-glib-memory-slices-no-compat configuration option: instead
 * glib will terminate the program if it is unable to obtain memory
 * from the operating system.)
 * @exception Cgu::Thread::MutexError It might throw
 * Cgu::Thread::MutexError if initialisation of the contained mutex
 * fails.  (It is often not worth checking for this, as it means
 * either memory is exhausted or pthread has run out of other
 * resources to create new mutexes.)
 * @exception Cgu::Thread::CondError It might throw
 * Cgu::Thread::CondError if initialisation of the contained condition
 * variable fails.  (It is often not worth checking for this, as it
 * means either memory is exhausted or pthread has run out of other
 * resources to create new condition variables.)
 * @note This method will also throw if the copy constructor of a
 * bound argument throws, or the default constructor of the return
 * value type throws.
 *
 * Since 1.0.2
 */
  template <class Ret, class Arg1, class Arg2>
  static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > make(Ret (*func)(Arg1, Arg2),
                               typename Cgu::Param<Arg1>::ParamType arg1,
                               typename Cgu::Param<Arg2>::ParamType arg2);
/**
 * Runs the function represented by this Cgu::Thread::Future object in
 * a new worker thread.  That function will only be run once.  If this
 * is the first time this method has been called, it will start the
 * worker thread and return true, and if it has previously been
 * called, this method will do nothing and return false.  This method
 * will not wait for the worker thread to complete before returning.
 * This method is thread safe and may be called by a different thread
 * from the one which called make().
 * @return true if this is the first time this method has been called,
 * or false if this method has previously been called.
 * @exception Cgu::Thread::FutureThreadError This method might throw
 * Cgu::Thread::FutureThreadError if it has not previously been called
 * and the thread did not start properly.  If it does throw, this
 * Cgu::Thread::Future object is defunct and further attempts to call
 * this method will return immediately with a false value.  (It is
 * often not worth checking for this exception, as it means either
 * memory is exhausted, the pthread thread limit has been reached or
 * pthread has run out of other resources to start new threads.)
 * @exception std::bad_alloc This method might throw std::bad_alloc if
 * it has not previously been called, and memory is exhausted and the
 * system throws in that case.  If it does throw, this
 * Cgu::Thread::Future object is defunct and further attempts to call
 * this method will return immediately with a false value.  (This
 * exception will not be thrown if the library has been installed
 * using the \--with-glib-memory-slices-no-compat configuration
 * option: instead glib will terminate the program if it is unable to
 * obtain memory from the operating system.)
 * @note 1. Any Cgu::Thread::Exit exception, or any uncaught exception
 * derived from std::exception, which is thrown from the worker thread
 * will be caught and consumed and the error flag will be set.  The
 * worker thread will safely terminate and unwind the stack in so
 * doing.
 * @note 2. As this wrapper class can provide error reporting in a way
 * that Cgu::Thread::Thread of itself cannot, it would be desirable to
 * consume any other uncaught exceptions.  However, this cannot be
 * done: annoyingly, NPTL's forced stack unwinding does not allow this
 * if thread cancellation is to be made available.  Neither POSIX nor
 * the C++98/03 standard specifies what is to happen if an uncaught
 * exception propagates out of a thread when the thread exits.  In
 * practice, most implementations will call std::terminate() and so
 * terminate the entire program (and this is required by the C++11
 * standard).  Accordingly, a user must make sure that no exceptions,
 * other than Cgu::Thread::Exit or those derived from std::exception
 * or any cancellation pseudo-exception, can propagate from the
 * function which this Cgu::Thread::Future object represents, nor from
 * the copy constructor of any argument type that that function takes
 * by value nor from the assignment operator of the return value of
 * that function.
 * @note 3. If the worker thread is cancelled by a call to cancel()
 * while in the middle of executing the function which this
 * Cgu::Thread::Future object represents, the error flag will be set.
 *
 * Since 1.0.2
 */
  bool run();
 
/**
 * Gets the value obtained from the function which is represented by
 * this object.  If the worker thread launched by the call to run()
 * has not completed, then this method will block until it has
 * completed.  If run() has not been called, then run() will be called
 * (and this method will block until the launched worker thread
 * completes).  If the function which is represented by this
 * Cgu::Thread::Future object throws Cgu::Thread::Exit or an uncaught
 * exception derived from std::exception, or if any of those
 * exceptions are thrown either by the copy constructor of an argument
 * taken by value by that function, or by the assignment operator of
 * the return value of that function, then the exception will have
 * been consumed by this Cgu::Thread::Future object and the error flag
 * will have been set.  The error flag will also have been set if the
 * worker thread is cancelled or the thread wrapper in this
 * Cgu::Thread::Future object threw std::bad_alloc.  On the error flag
 * being set, this method will unblock and return a default
 * constructed object of the return type.  This method is thread safe
 * and may be called by any thread (and by more than one thread).  It
 * is a cancellation point if it blocks, and from version 1.2.24 is
 * cancellation safe if the stack unwinds on cancellation.  It is also
 * strongly exception safe: no data will be lost if extracting the
 * value fails.
 * @return The value obtained from the function which is represented
 * by this object, or a default constructed object of the return type
 * if the error flag has been set.
 * @exception Cgu::Thread::FutureThreadError This method might throw
 * Cgu::Thread::FutureThreadError if run() has not previously been
 * called and the thread did not start properly when this function
 * called run().
 * @exception std::bad_alloc This method might throw std::bad_alloc if
 * run() has not previously been called, memory is exhausted and the
 * system throws in that case.  (This exception will not be thrown if
 * the library has been installed using the
 * \--with-glib-memory-slices-no-compat configuration option: instead
 * glib will terminate the program if it is unable to obtain memory
 * from the operating system.)
 * @note 1. This method might also throw if the copy constructor of
 * the returned value type throws.
 * @note 2. Question: Couldn't this method return the stored value by
 * reference to const?  Answer: It could.  However, because of return
 * value optimization, which will be implemented by any compiler
 * capable of compiling this library, no advantage would be gained by
 * doing so when initializing a local variable with the return value
 * of this method (the copy constructor will only be called once
 * whether returning by value or const reference).  The advantage of
 * returning by value is that the call to the copy constructor is
 * forced to be within this Thread::Future object's mutex, so
 * different threads' calls to the copy constructor are serialized,
 * and also with blocked cancellation, so this method is cancellation
 * safe.  All calls to this method by different threads are therefore
 * isolated and we do not have to worry about the thread safety of
 * direct access to the stored value via its const methods outside the
 * mutex (which would not be thread safe if the stored value has data
 * members declared mutable) nor about the cancellation safety of the
 * copy constructor.  Of course, for objects which do not have mutable
 * data, a hit arises by returning by value in cases where it is not
 * intended to initialize a local variable at all nor to cancel a
 * thread: where, say, only const methods are to be called on the
 * return value (which could be done directly if this method returned
 * by const reference).  However, that is a design trade-off.
 *
 * Since 1.0.2
 */
  Val get();
 
/**
 * Cancels the worker thread in which the function represented by this
 * object runs, if that function has not yet finished.  If this method
 * is called and the worker thread is still running and is cancelled
 * in response to a call to this method, then the error flag will be
 * set so that a method calling get() can examine whether the result
 * is valid.  If run() has not yet been called or the worker thread
 * has already finished executing the function represented by this
 * object then this function does nothing and returns false.  This
 * method is thread safe and may be called by any thread.  It will not
 * throw.
 * @return true if run() has previously been called and the worker
 * thread has not yet finished executing the function represented by
 * this object, otherwise false (in which case this method does
 * nothing).
 * @note 1. Use this method with care.  When cancelling a thread not
 * all thread implementations will unwind the stack, and so run the
 * destructors of local objects.  This is discussed further in the
 * documentation on Cgu::Thread::Thread::cancel().
 * @note 2. This method might return true because the worker thread
 * has not yet finished, but the error flag might still not be set.
 * This is because the worker thread may not meet a cancellation point
 * before it ends naturally.  It is the error flag which indicates
 * definitively whether the worker thread terminated prematurely in
 * response to a call to this method.
 *
 * Since 1.0.2
 */
  bool cancel();
 
/**
 * A utility enabling the value returned by the thread function
 * represented by this Cgu::Thread::Future object to be dealt with
 * asynchronously rather than by (or in addition to) a call to the
 * get() method.  It causes the callback passed as an argument to this
 * method (referred to below as the 'when' callback) to be executed by
 * a thread's main loop if and when the thread function represented by
 * this Cgu::Thread::Future object finishes correctly - the 'when'
 * callback is passed that thread function's return value when it is
 * invoked.  This method is thread safe, and may be called by any
 * thread.
 *
 * This functionality is implemented by connecting an internal
 * dispatching callback to the done_emitter object.
 *
 * The 'when' callback should take a single unbound argument
 * comprising a const reference to the return type of the thread
 * function represented by this Cgu::Thread::Future object.  (So, in
 * the case of a Future<int> object, the callback function should take
 * a const int& argument as the unbound argument.)  The 'when'
 * callback can have any number of bound arguments, except that a
 * bound argument may not include a copy of this Cgu::Thread::Future
 * object held by intrusive pointer as returned by the make() methods
 * (that would result in this Cgu::Thread::Future object owning, via
 * done_emitter, a reference to itself and so become incapable of
 * being freed).  The 'when' callback may, however, take a pointer to
 * this Cgu::Thread::Future object, as obtained by the
 * Cgu::IntrusivePtr::get() method, because this Cgu::Thread::Future
 * object is guaranteed to remain in existence until the callback has
 * completed executing.
 *
 * This method cannot be called after the thread function represented
 * by this Cgu::Thread::Future object has completed (either
 * successfully or unsuccessfully) so that is_done() would return
 * true, and if this is attempted a Cgu::Thread::FutureWhenError
 * exception will be thrown.  Therefore, generally this method should
 * be called before the run() method has been called.
 *
 * Once the run() method has been called, this Cgu::Thread::Future
 * object will always stay in existence until the thread function
 * represented by it has completed (whether correctly, by cancellation
 * or by a thrown exception), and any 'when' callback (and any other
 * callbacks connected to the done_emitter object) and any 'fail'
 * callback have completed.  Accordingly it is safe to use this method
 * even if the intrusive pointer object returned by the make() methods
 * will go out of scope before the 'when' callback has executed: the
 * callback will execute correctly irrespective of that.
 *
 * Summary: use of this method is safe and has been implemented in a
 * way which does not give rise to timing issues.
 *
 * If memory is exhausted and std::bad_alloc is thrown by the thread
 * wrapper of Cgu::Thread::Future after run() is called or by
 * done_emitter when emitting, or if the thread function represented
 * by this Cgu::Thread::Future object throws Cgu::Thread::Exit, is
 * cancelled, exits with an uncaught exception deriving from
 * std::exception, takes an argument by value whose copy constructor
 * throws such an exception or has a return value whose assignment
 * operator throws such an exception, or if the 'when' callback
 * represents a function taking a non-reference argument whose copy
 * constructor throws an exception, or any other callback has been
 * connected to done_emitter before this method is called which exits
 * with an uncaught exception, then the 'when' callback will not
 * execute (instead the exception concerned will be consumed and an
 * error indicated).  With many systems, swap memory combined with
 * memory over-commit makes it pointless to check for std::bad_alloc
 * (and even more so in programs using glib, as glib aborts a program
 * where it cannot obtain memory from the operating system).  So
 * subject to that, if the user program is designed so that the thread
 * function represented by this Cgu::Thread::Future object does not
 * exit with uncaught exceptions, does not take an argument by value
 * which throws, does not have a return value whose assignment
 * operator throws, does not throw Cgu::Thread::Exit and is not
 * cancelled, and so that the 'when' callback does not exit with an
 * uncaught exception (and the function represented by that callback
 * either takes no arguments of class type by value or the copy
 * constructors of any of its value arguments do not throw), and if
 * this method is called before any other callbacks are connected to
 * done_emitter, the possibility of failure can be disregarded.
 *
 * In cases where that is not true and detecting whether a failure has
 * occurred is required, a fail() method is provided.  It should be
 * noted that a callback handed to the fail() method will not execute
 * in a case of error if the error comprises the 'when' callback
 * exiting with an uncaught exception when it is executed by the main
 * loop, or the copy constructor of any value argument of the function
 * represented by the 'when' callback throwing (such exceptions would
 * be consumed internally in order to protect the main loop and a
 * g_critical message issued).  If the 'when' callback might exit with
 * an uncaught exception when executing or have the copy constructor
 * of a value argument throw, and doing something other than consuming
 * the exception and issuing a g_critical message is required, then a
 * different approach is to start a new thread to wait on the get()
 * method which can act on the result of is_error() directly.
 *
 * If glib < 2.32 is used, the glib main loop must have been made
 * thread-safe by a call to g_thread_init() before this function is
 * called.  glib >= 2.32 does not require g_thread_init() to be called
 * in order to be thread safe.
 *
 * @param cb The 'when' callback (the callback to be executed when the
 * function represented by this Cgu::Thread::Future object has
 * successfully completed).  Ownership is taken of this object, and it
 * will be deleted when it has been finished with.
 * @param priority The priority to be given to the 'when' callback in
 * the main loop after the thread function represented by this
 * Cgu::Thread::Future object has successfully completed.  In
 * ascending order of priorities, priorities are G_PRIORITY_LOW,
 * G_PRIORITY_DEFAULT_IDLE, G_PRIORITY_HIGH_IDLE, G_PRIORITY_DEFAULT
 * and G_PRIORITY_HIGH. The default is G_PRIORITY_DEFAULT.  This
 * determines the order in which the callback will appear in the event
 * list in the main loop, not the priority which the OS will adopt.
 * @param context The glib main context of the thread in whose main
 * loop the 'when' callback is to be executed (the default of NULL
 * will cause the callback to be executed in the main program loop).
 * @return The internal dispatching callback created by this method
 * and connected to done_emitter.  It is made available as a return
 * value so that if wanted it can be disconnected programmatically
 * from done_emitter, or block()/unblock() can be called on it (but if
 * that is to be done, it must be done before the thread function
 * represented by this Cgu::Thread::Future object has completed in
 * order for it to be effective).
 * @exception Cgu::Thread::FutureWhenError This method will throw
 * Cgu::Thread::FutureWhenError if it is called after the thread
 * function represented by this Cgu::Thread::Future object has
 * completed.  If it does so, the 'when' callback will be disposed of.
 * @exception std::bad_alloc This method might throw std::bad_alloc if
 * memory is exhausted and the system throws in that case.  If it does
 * so, the 'when' callback will be disposed of.
 * @note The return value of the function represented by this
 * Cgu::Thread::Future object is stored and passed as an argument to
 * the 'when' callback by const reference.  If other threads might
 * concurrently call this object's get() method, which copies the
 * stored value, the stored type's copy constructor must be thread
 * safe with respect to the stored type's const methods.  This would
 * be relevant if the stored type has data members declared mutable
 * which would be copied by its copy constructor.
 *
 * Since 1.2.16
 */
  Cgu::Callback::SafeFunctor when(const Cgu::Callback::CallbackArg<const Val&>* cb,
                  gint priority = G_PRIORITY_DEFAULT,
                  GMainContext* context = 0);
 
/**
 * This is a version of the utility enabling the value returned by the
 * thread function represented by this Cgu::Thread::Future object to
 * be dealt with asynchronously, which takes a Releaser object for
 * automatic disconnection of the callback passed as an argument to
 * this method (referred to below as the 'when' callback), if the
 * object having the 'when' callback function as a member is
 * destroyed.  For this to be race free, the lifetime of that object
 * must be controlled by the thread in whose main loop the 'when'
 * callback will execute.
 *
 * If the 'when' callback has not been released, this method causes it
 * to be executed by a thread's main loop if and when the thread
 * function represented by this Cgu::Thread::Future object finishes
 * correctly - the 'when' callback is passed that thread function's
 * return value when it is invoked.  This method is thread safe, and
 * may be called by any thread.
 *
 * This functionality is implemented by connecting an internal
 * dispatching callback to the done_emitter object.
 *
 * The 'when' callback should take a single unbound argument
 * comprising a const reference to the return type of the thread
 * function represented by this Cgu::Thread::Future object.  (So, in
 * the case of a Future<int> object, the callback function should take
 * a const int& argument as the unbound argument.)  The 'when'
 * callback can have any number of bound arguments, except that a
 * bound argument may not include a copy of this Cgu::Thread::Future
 * object held by intrusive pointer as returned by the make() methods
 * (that would result in this Cgu::Thread::Future object owning, via
 * done_emitter, a reference to itself and so become incapable of
 * being freed).  The 'when' callback may, however, take a pointer to
 * this Cgu::Thread::Future object, as obtained by the
 * Cgu::IntrusivePtr::get() method, because this Cgu::Thread::Future
 * object is guaranteed to remain in existence until the callback has
 * completed executing.
 *
 * This method cannot be called after the thread function represented
 * by this Cgu::Thread::Future object has completed (either
 * successfully or unsuccessfully) so that is_done() would return
 * true, and if this is attempted a Cgu::Thread::FutureWhenError
 * exception will be thrown.  Therefore, generally this method should
 * be called before the run() method has been called.
 *
 * The documentation for the version of this method which does not
 * take a Releaser object gives further details of how this method is
 * used.
 *
 * If glib < 2.32 is used, the glib main loop must have been made
 * thread-safe by a call to g_thread_init() before this function is
 * called.  glib >= 2.32 does not require g_thread_init() to be called
 * in order to be thread safe.
 *
 * @param cb The 'when' callback (the callback to be executed when the
 * function represented by this Cgu::Thread::Future object has
 * successfully completed).  Ownership is taken of this object, and it
 * will be deleted when it has been finished with.
 * @param r A Releaser object for automatic disconnection of the
 * 'when' callback before it executes in a main loop (mainly relevant
 * if the callback represents a non-static member function of an
 * object which may be destroyed before the callback executes).
 * @param priority The priority to be given to the 'when' callback in
 * the main loop after the thread function represented by this
 * Cgu::Thread::Future object has successfully completed.  In
 * ascending order of priorities, priorities are G_PRIORITY_LOW,
 * G_PRIORITY_DEFAULT_IDLE, G_PRIORITY_HIGH_IDLE, G_PRIORITY_DEFAULT
 * and G_PRIORITY_HIGH. The default is G_PRIORITY_DEFAULT.  This
 * determines the order in which the callback will appear in the event
 * list in the main loop, not the priority which the OS will adopt.
 * @param context The glib main context of the thread in whose main
 * loop the 'when' callback is to be executed (the default of NULL
 * will cause the callback to be executed in the main program loop).
 * @return The internal dispatching callback created by this method
 * and connected to done_emitter.  It is made available as a return
 * value so that if wanted it can be disconnected programmatically
 * from done_emitter, or block()/unblock() can be called on it (but if
 * that is to be done, it must be done before the thread function
 * represented by this Cgu::Thread::Future object has completed in
 * order for it to be effective).
 * @exception Cgu::Thread::FutureWhenError This method will throw
 * Cgu::Thread::FutureWhenError if it is called after the thread
 * function represented by this Cgu::Thread::Future object has
 * completed.  If it does so, the 'when' callback will be disposed of.
 * @exception std::bad_alloc This method might throw std::bad_alloc if
 * memory is exhausted and the system throws in that case.  If it does
 * so, the 'when' callback will be disposed of.
 * @exception Cgu::Thread::MutexError This method will throw
 * Cgu:Thread::MutexError if initialisation of the mutex in a
 * SafeEmitterArg object constructed by this method fails.  If it does
 * so, the 'when' callback will be disposed of.  (It is often not
 * worth checking for this, as it means either memory is exhausted or
 * pthread has run out of other resources to create new mutexes.)
 * @note 1. The return value of the function represented by this
 * Cgu::Thread::Future object is stored and passed as an argument to
 * the 'when' callback by const reference.  If other threads might
 * concurrently call this object's get() method, which copies the
 * stored value, the stored type's copy constructor must be thread
 * safe with respect to the stored type's const methods.  This would
 * be relevant if the stored type has data members declared mutable
 * which would be copied by its copy constructor.
 * @note 2. By virtue of the Releaser object, it is in theory possible
 * (if memory is exhausted and the system throws in that case) that an
 * internal SafeEmitterArg object will throw std::bad_alloc when
 * emitting/executing the 'when' callback in the glib main loop, with
 * the result that the relevant callback will not execute (instead the
 * exception will be consumed and a g_critical() warning will be
 * issued).  This is rarely of any relevance because glib will abort
 * the program if it is itself unable to obtain memory from the
 * operating system.  However, where it is relevant, design the
 * program so that it is not necessary to provide a releaser object.
 *
 * Since 1.2.16
 */
  Cgu::Callback::SafeFunctor when(const Cgu::Callback::CallbackArg<const Val&>* cb,
                  Cgu::Releaser& r,
                  gint priority = G_PRIORITY_DEFAULT,
                  GMainContext* context = 0);
 
/**
 * A utility intended to be used where relevant in conjunction with
 * the when() methods.  It enables a callback to be executed in a glib
 * main loop (referred to below as the 'fail' callback) if memory is
 * exhausted and std::bad_alloc was thrown by the thread wrapper of
 * Cgu::Thread::Future after calling run() or by done_emitter when
 * emitting, or if the thread function represented by this
 * Cgu::Thread::Future object threw Cgu::Thread::Exit, exited with an
 * uncaught exception deriving from std::exception or was cancelled
 * (or that function took an argument of class type by value whose
 * copy constructor threw such an exception or had a return value of
 * class type whose assignment operator threw such an exception), or
 * any callback connected to done_emitter exited with an uncaught
 * exception.  It therefore enables errors to be detected and acted on
 * without having a thread wait on the get() method in order to test
 * is_error() or is_emitter_error().
 *
 * It is implemented by attaching a timeout to the main loop which
 * polls at 100 millisecond intervals and tests is_done()/is_error()
 * and is_emitter_done()/is_emitter_error().  The timeout is
 * automatically removed by the implementation once it has been
 * detected that an error has occurred and the 'fail' callback is
 * executed, or if the thread function represented by this Cgu::Future
 * object and all done_emitter emissions (including execution of any
 * 'when' callback) have completed successfully.
 *
 * This method can be called before or after the run() method has been
 * called, and whether or not the thread function represented by this
 * Cgu::Thread::Future object has completed.
 *
 * Once this method has been called, this Cgu::Thread::Future object
 * will always stay in existence until the timeout has been
 * automatically removed by the implementation.  Accordingly it is
 * safe to use this method even if the intrusive pointer object
 * returned by the make() methods will go out of scope before the
 * 'fail' callback has executed: the callback will execute correctly
 * irrespective of that.
 *
 * This method does not have a priority argument: as a polling timeout
 * is created, a particular priority will normally have no
 * significance (in fact, the 'fail' callback will execute in the main
 * loop with a priority of G_PRIORITY_DEFAULT).  If in a special case
 * a different polling interval than 100 milliseconds or a different
 * priority is required, users can attach their own polling timeouts
 * to a main loop and carry out the tests by hand.
 *
 * Four other points should be noted.  First, if as well as the when()
 * method being called some other callback has been connected to
 * done_emitter, and that other callback throws, the 'fail' callback
 * will execute.  Therefore, if the particular program design requires
 * that the 'fail' callback should only execute if the 'when' callback
 * is not executed (and the 'when' callback only execute if the 'fail'
 * callback does not execute), no other callbacks which throw should
 * be connected to done_emitter.
 *
 * Secondly, as mentioned in the documentation on the when() method,
 * if the 'when' callback exits with an uncaught exception upon being
 * executed by the main loop or the function it represents takes an
 * argument by value whose copy constructor throws, the 'fail'
 * callback will not execute (the exception will have been consumed
 * internally in order to protect the main loop and a g_critical
 * message issued).
 *
 * Thirdly, avoid if possible having the 'fail' callback represent a
 * function which either might throw or takes an argument by value
 * whose copy constructor might throw (such an exception would be
 * consumed internally in order to protect the main loop and a
 * g_critical message issued, but no other error indication apart from
 * the g_critical message will be provided).
 *
 * Fourthly, unlike the 'when' callback, a copy of this
 * Cgu::Thread::Future object held by intrusive pointer as returned by
 * the make() methods may safely be bound to the 'fail' callback,
 * which would enable the 'fail' callback to determine whether it is
 * is_error() or is_emitter_error() which returns false.
 *
 * If glib < 2.32 is used, the glib main loop must have been made
 * thread-safe by a call to g_thread_init() before this function is
 * called.  glib >= 2.32 does not require g_thread_init() to be called
 * in order to be thread safe.
 *
 * @param cb The 'fail' callback (the callback to be executed if the
 * thread function represented by this Cgu::Thread::Future object or a
 * done_emitter emission has failed to complete).  Ownership is taken
 * of this object, and it will be deleted when it has been finished
 * with.
 * @param context The glib main context of the thread in whose main
 * loop the 'fail' callback is to be executed (the default of NULL
 * will cause the functor to be executed in the main program loop).
 * @exception std::bad_alloc This method might throw std::bad_alloc if
 * memory is exhausted and the system throws in that case.  If it does
 * so, the 'fail' callback will be disposed of.
 *
 * Since 1.2.16
 */
  void fail(const Cgu::Callback::Callback* cb,
        GMainContext* context = 0);
 
/**
 * This is a version of the fail() utility for use in conjunction with
 * the when() methods, which takes a Releaser object for automatic
 * disconnection of the callback functor passed as an argument to this
 * method if the object having the callback function as a member is
 * destroyed.  For this to be race free, the lifetime of that object
 * must be controlled by the thread in whose main loop the 'fail'
 * callback will execute.
 *
 * This method enables a callback to be executed in a glib main loop
 * if memory is exhausted and std::bad_alloc was thrown by the thread
 * wrapper of Cgu::Thread::Future after calling run() or by
 * done_emitter when emitting, or if the thread function represented
 * by this Cgu::Thread::Future object threw Cgu::Thread::Exit, exited
 * with an uncaught exception deriving from std::exception or was
 * cancelled (or that function took an argument of class type by value
 * whose copy constructor threw such an exception or had a return
 * value of class type whose assignment operator threw such an
 * exception), or any callback connected to done_emitter exited with
 * an uncaught exception.  It therefore enables errors to be detected
 * and acted on without having a thread wait on the get() method in
 * order to test is_error() or is_emitter_error().
 *
 * This method can be called before or after the run() method has been
 * called, and whether or not the thread function represented by this
 * Cgu::Thread::Future object has completed.
 *
 * The documentation for the version of this method which does not
 * take a Releaser object gives further details of how this method is
 * used.
 *
 * If glib < 2.32 is used, the glib main loop must have been made
 * thread-safe by a call to g_thread_init() before this function is
 * called.  glib >= 2.32 does not require g_thread_init() to be called
 * in order to be thread safe.
 *
 * @param cb The 'fail' callback (the callback to be executed if the
 * thread function represented by this Cgu::Thread::Future object or a
 * done_emitter emission has failed to complete).  Ownership is taken
 * of this object, and it will be deleted when it has been finished
 * with.
 * @param r A Releaser object for automatic disconnection of the
 * 'fail' callback before it executes in a main loop (mainly relevant
 * if the callback represents a non-static member function of an
 * object which may be destroyed before the callback executes).
 * @param context The glib main context of the thread in whose main
 * loop the 'fail' callback is to be executed (the default of NULL
 * will cause the functor to be executed in the main program loop).
 * @exception std::bad_alloc This method might throw std::bad_alloc if
 * memory is exhausted and the system throws in that case.  If it does
 * so, the 'fail' callback will be disposed of.
 * @exception Cgu::Thread::MutexError This method will throw
 * Cgu:Thread::MutexError if initialisation of the mutex in a
 * SafeEmitterArg object constructed by Cgu::start_timeout() fails.
 * If it does so, the 'fail' callback will be disposed of.  (It is
 * often not worth checking for this, as it means either memory is
 * exhausted or pthread has run out of other resources to create new
 * mutexes.)
 * @note By virtue of the Releaser object, it is in theory possible
 * (if memory is exhausted and the system throws in that case) that an
 * internal SafeEmitterArg object will throw std::bad_alloc when
 * emitting/executing the 'fail' callback in the glib main loop, with
 * the result that the relevant callback will not execute (instead the
 * exception will be consumed and a g_critical() warning will be
 * issued).  This is rarely of any relevance because glib will abort
 * the program if it is itself unable to obtain memory from the
 * operating system.  However, where it is relevant, design the
 * program so that it is not necessary to provide a releaser object.
 *
 * Since 1.2.16
 */
  void fail(const Cgu::Callback::Callback* cb,
        Cgu::Releaser& r,
        GMainContext* context = 0);
 
/**
 * @return true if the function represented by this
 * Cgu::Thread::Future object has finished, either by returning
 * normally, by cancellation or by virtue of having thrown
 * Cgu::Thread::Exit or some exception derived from std::exception.
 * Once this method returns true, then it is guaranteed that the get()
 * method will not block (except as incidental to any contention with
 * other threads calling get()).  Once this method has returned true
 * or get() has unblocked, then the result of is_error() is
 * definitive.  This method is thread safe and may be called by any
 * thread.  It will not throw.
 * @note This method will return true even though any callbacks
 * connected to done_emitter are still executing or waiting to
 * execute.  From version 1.2.16 the is_emitter_done() method will
 * indicate when done_emitter callbacks (if any) have also completed.
 *
 * Since 1.0.2
 */
  bool is_done() const;
 
/**
 * @return true if both the function represented by this
 * Cgu::Thread::Future object has finished and any callbacks connected
 * to done_emitter have completed.  Once this method returns true,
 * then the result of is_emitter_error() is definitive.  This method
 * is thread safe and may be called by any thread.  It will not throw.
 * @note This method will return true automatically if is_error() and
 * is_done() return true, because if the function represented by this
 * Cgu::Thread::Future object was cancelled or exited with an uncaught
 * exception, done_emitter is never emitted.  In addition, if this
 * method returns true, then is_done() must also return true.
 *
 * Since 1.2.16
 */
  bool is_emitter_done() const;
 
/**
 * @return true if (a) a Cgu::Thread::Exit exception has been thrown
 * by the function represented by this Cgu::Thread::Future object
 * (which will have been consumed by this Cgu::Thread::Future object),
 * (b) an exception derived from std::exception has been thrown on
 * invoking that function which was not caught in that function (which
 * will have been consumed by this Cgu::Thread::Future object), (c)
 * any of those exceptions have been thrown either by the copy
 * constructor of an argument taken by value by that function, or by
 * the assignment operator of the return value of that function (which
 * will have been consumed by this Cgu::Thread::Future object), (d)
 * the worker thread in which that function runs was cancelled in
 * mid-course with a call to cancel() or (e) the thread wrapper
 * implementing the worker thread in this Cgu::Thread::Future object
 * threw and then consumed std::bad_alloc (this is different from the
 * run() method throwing std::bad_alloc).  In these cases the value
 * obtained by get() will not be valid (it will be a default
 * constructed object of the return type of the function represented
 * by this Cgu::Thread::Future object).  Otherwise this method returns
 * false.  The result of this method is definitive once get() has
 * unblocked or is_done() returns true.  This method is thread safe
 * and may be called by any thread.  It will not throw.
 *
 * Since 1.0.2
 */
  bool is_error() const;
 
/**
 * @return true if an uncaught exception arose in emitting @ref
 * DoneEmitterAnchor "done_emitter" when executing callbacks connected
 * to it.  Otherwise this method returns false.  The result of this
 * method is definitive once is_emitter_done() returns true.  This
 * method is thread safe and may be called by any thread.  It will not
 * throw.
 * @note This method will return false automatically if is_error()
 * returns true, because if the function represented by this
 * Cgu::Thread::Future object was cancelled or exited with an uncaught
 * exception, done_emitter is never emitted.  It follows that if this
 * method returns true, is_error() must return false.
 *
 * Since 1.0.2
 */
  bool is_emitter_error() const;
 
/**
 * A Cgu::SafeEmitter object which is emitted when the function
 * represented by this Cgu::Thread::Future object finishes correctly
 * (that is, the function is not cancelled and does not throw any
 * uncaught exceptions).  By itself it does not do too much as it is
 * emitted (and connected callbacks execute in) the same worker thread
 * immediately after the Future function has completed.  However, any
 * thread can connect a Callback object to this Cgu::SafeEmitter
 * object and a connected callback can, say, cause another Callback to
 * be executed in a thread's main loop using Cgu::Callback::post(),
 * and from version 1.2.16 when() methods are provided which will do
 * this for users automatically.  Once the run() method has been
 * called, this Cgu::Thread::Future object (and so done_emitter) will
 * always stay in existence until the function represented by it has
 * completed (whether correctly, by cancellation or by a thrown
 * exception) and any callbacks connected to the done_emitter object
 * have completed, irrespective of whether the intrusive pointer
 * returned by the make() methods has gone out of scope.
 * @note 1. Cancellation is blocked while the Cgu::SafeEmitter object
 * emits and any connected callback executes.
 * @note 2. A connected callback can however terminate the worker
 * thread by throwing Cgu::Thread::Exit (in which case no subsequent
 * callbacks to be executed on that emission will execute either: the
 * worker thread will safely terminate and unwind the stack in so
 * doing).  In that event, the emitter_error flag will be set.
 * @note 3. All other uncaught exceptions which might be thrown by the
 * Cgu::SafeEmitter object emitting, or by a connected callback
 * function executing, are consumed to retain the integrity of the
 * Thread::Future object.  In the event of such an exception being
 * thrown, the emitter_error flag will be set.  In summary, the
 * emitter_error flag will be set if (a) a callback function throws
 * Cgu::Thread::Exit, (b) some other uncaught exception escapes from a
 * callback function or (c) Cgu::SafeEmitter::emit() throws
 * std::bad_alloc or the copy constructor of a bound argument which is
 * not a reference argument has thrown.  If the user knows that the
 * callback function does not throw Cgu::Thread::Exit and does not
 * allow any other exception to escape, then the cause must be a
 * std::bad_alloc memory exception in Cgu::SafeEmitter::emit() or the
 * copy constructor of a non-reference bound argument throwing.
 * @note 4. An emission is thread safe if the connected callback
 * functions are thread safe.
 * @note 5. This Cgu::Thread::Future object's mutex is released while
 * the Cgu::SafeEmitter object emits.  This means that any connected
 * callbacks can safely call, say, the Future object's get() or
 * is_error() methods.  However, a connected callback should not have
 * a bound argument comprising a copy of this Cgu::Thread::Future
 * object held by intrusive pointer as returned by the make() methods
 * (that would result in this Cgu::Thread::Future object owning, via
 * done_emitter, a reference to itself and so become incapable of
 * being freed).  The callback may, however, take a pointer to this
 * Cgu::Thread::Future object as a bound argument, as obtained by the
 * Cgu::IntrusivePtr::get() method, because this Cgu::Thread::Future
 * object is guaranteed to remain in existence until all callbacks
 * connected to done_emitter have completed executing.
 * @anchor DoneEmitterAnchor
 *
 * Since 1.0.2
 */
  SafeEmitter done_emitter;
 
/* Only has effect if --with-glib-memory-slices-compat or
 * --with-glib-memory-slices-no-compat option picked */
  CGU_GLIB_MEMORY_SLICES_FUNCS
};
 
} // namespace Thread
 
} // namespace Cgu
 
#include <c++-gtk-utils/future.tpp>
 
#endif