- Boost-users - lists.boost.org

Re: [Boost-users] std::vector< boost::shared_ptr<int> >::pop_back()
by Andrew Holden 10 Jul '06

10 Jul '06

You would manually call destructors when you do your own memory management (like std::vector). In this case, you would also need to manually call constructors too. Here is a modified version of your test program that doesn't double-destruct the object. class A { public: A() { counter = 0; } ~A() { counter++; std::cout << "~A(): " << counter << std::endl; } int counter; }; int main(int argc, char* argv[]) { A *a; //Allocate memory using C functions, which do not call constructors a = (A*) malloc (sizeof (A)); cout << a << ' ' << a->counter << std::endl; //Manually call the constructor. The (a) tells the compiler which pointer we are initializing new (a) A (); cout << a << ' ' << a->counter << std::endl; //Manually call the destructor a->~A(); cout << a << ' ' << a->counter << std::endl; //Free memory using C functions, which do not call destructors free (a); return 0; } _____ From: bringiton bringiton [mailto:kneeride@gmail.com] Sent: Monday, July 10, 2006 10:28 AM To: boost-users(a)lists.boost.org Subject: Re: [Boost-users] std::vector< boost::shared_ptr<int> >::pop_back() holy! you learn something every day. you can call the destructor of an object (before destruction) this is a trap for me because i dont clean up my variables in destructors. class A { public: A() { counter = 0; } ~A() { counter++; std::cout << "~A(): " << counter << std::endl; } int counter; }; int main(int argc, char* argv[]) { A a; a.~A(); return 0; } NOTE: destructor gets called twice. (so it's probably a good idea to NULL pointers in destructors) On 7/11/06, bringiton bringiton <kneeride(a)gmail.com> wrote: >>just look at the include files: i looked at the header, but got a little confused. i don't see how you can call a destruct of a a memory space that still exists. ie a C version of a vector. shared_ptr<int> list[1024]; int n = 0; // add an item shared_ptr<int> newItem(new int(1)); list[n++] = newItem; // remove the tail item n--; // destructor never called is it possible to call a destructor of an object that exists? what then happens then the object goes out of scope? the destructor will be called twice. (or maybe i am completely missing the point) BTW: not sure if above code compiles (did it in my head) On 7/10/06, Boris Breidenbach < Boris.Breidenbach(a)physik.uni-erlangen.de <mailto:Boris.Breidenbach@physik.uni-erlangen.de> > wrote: On Mon, Jul 10, 2006 at 03:58:19PM +1000, bringiton bringiton wrote: > This question is based on curiosity. How does std::vector::pop_back() call > the destructor of the item getting removed? > > i understood std::vector to be a contigious array of memory, therefore an > item's memory does not go out of scope when being popped. ie the item goes > out of scope when the entire array goes out of scope. just look at the include files: bits/stl_vector.h says: void pop_back() { --this->_M_impl._M_finish; std::_Destroy(this->_M_impl._M_finish); } so the destructor is called in std::_Destroy. Which can be found in bits/stl_construct.h: /** * @if maint * Destroy the object pointed to by a pointer type. * @endif */ template<typename _Tp> inline void _Destroy(_Tp* __pointer) { __pointer->~_Tp(); } It just calls the destructor. _______________________________________________ Boost-users mailing list Boost-users(a)lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

3 2

Re: [Boost-users] std::vector< boost::shared_ptr<int> >::pop_back()
by Andrew Holden 10 Jul '06

10 Jul '06

True. I would not expect a guarantee that malloc/free would be suitable for objects, since they came from a language that doesn't have objects. I was merely looking for a way to demonstrate manually calling both the constructor and destructor without calling either twice, and that happened to work on my platform, and I think will probably work on most. Well, that and I didn't know you could use operator new and operator delete like this. We all learn something new every day. _____ From: Michael Nicolella [mailto:boost@mike-n.com] Sent: Monday, July 10, 2006 2:54 PM To: boost-users(a)lists.boost.org Subject: Re: [Boost-users] std::vector< boost::shared_ptr<int> >::pop_back() I would avoid malloc() and prefer the global operator new... I'm not sure if the standard guarantees that memory allocated with malloc is suitable for constructing an object in. My guess is there's no such guarantee, but there is that guarantee for operator new. Or maybe I'm off my rocker.

1 0

[serialization] save_construct_data problem
by Hugh Hoover 10 Jul '06

10 Jul '06

I have a structure I'm trying to serialize that contains an STL vector of pointers to an abstract base type. When I serialize it, the save_construct_data free function is NOT called - the inline default version is always used. On de-serializing, the load_construct_data IS called! but, of course, fails because the data for it isn't there (although it throws archive_exception (archive_exception::unregistered_class) when it fails). I'm using declarations only for all serialization code, with all definitions in .s.cpp files along with explicit instantiations. Not sure if that affects the problem here. Just in case, I've tried all combinations of save_construct_data with (const)? (Base|Derived) (*|&), and none of them are called. I've also tried putting my declarations before and after the serialization.hpp header with the default declaration/definition. I'm using gcc4 currently. Any notion what might be wrong here?

3 5

Missing example in boost::graph doc
by Loïc Joly 10 Jul '06

10 Jul '06

Hello, In boost::graph doc, I can read (http://www.boost.org/libs/graph/doc/file_dependency_example.html) > A compilation system such as make has to be able to answer a number of questions: > > 1. If we need to compile (or recompile) all of the files, what order should that be done it? > 2. What files can be compiled in parallel? > 3. If a file is changed, which files must be recompiled? > 4. Are there any cycles in the dependencies? (which means the user has made a mistake and an error should be emitted) > I'd like to do something related to the point #3, and looked in the doc, but could not find it. Since I'm not at all a graph expert, I'd like to know whether there is a basic graph algorithm that can be used to remove from a graph all nodes who are not marked as modified and whose all dependencies (recursively) are not marked as modified either. -- Loïc

1 0

memory leaks with signal::connect()
by Elisabeth Benoit 10 Jul '06

10 Jul '06

Hello, I'm sending this message a second time (I didn't receive any response and still have this problem) I'm using boost::bind and boost::signal with VC6. VC6 is detecting some memory leak for a certain type of signal.connect(). I've done a little test program to check if the problem was due to some bug or special configuration in my main program, but the memory leaks are still detected in the test situation. Let me explain the structure of my test program. I have a main, and in the main, an object CCloneAppRecipe named clone. The class CCloneAppRecipe contains a member m_PtrTestMem (a shared pointer, but a raw point or an object does the same thing) of the class CTestMem. CTestMem contains 3 instances of the class CTestTab, and the class CTestTab contains a boost signal typedef boost::signal0<void> TypeSignalTest; TypeSignalTest m_SignalTest; In my main, I invoke clone.connectPtrTestMem(). clone.connectPtrTestMem() then invokes CTestMem::ConnectSignalTriplePtrTab(boost::bind(&CloneAppRecipe::test, this)); ConnectSignalTriplePtrTab() then invokes ConnectSignal(TypeSignalTest::slot_type slot) for each of the 3 instances of the class it contains. ConnectSignal then calls m_SignalTest.connect(slot). And I have a memory leak: 2 blocks of 16 bytes. If I connect only one instance of CTestTab, no leak. Two instances, one leak of 16 bytes. Three, two leaks of 16 bytes, and so on. I connect the m_SignalTest with a CCloneAppRecipe's slot. If I connect it to a CTestMem slot, I have no leak. Or if I place 3 instances of CTestTab directly in the CCloneAppRecipe class (so I don't use CTestMem anymore), I have no leak. But that's not what I need to do. I've tested plenty of variations (shared_pointer or object, etc.) Does anybody have a clue. Thanks #include "stdafx.h" #include "CloneAppRecipe.h" #ifdef _DEBUG #define new DEBUG_NEW #undef THIS_FILE static char THIS_FILE[] = __FILE__; #endif int main() { CloneAppRecipe t; // cree un leak t.connectPtrTestMem(); getchar(); return 0; } void CloneAppRecipe::connectPtrTestMem() { //m_ptrTestMem is a boost::shared_ptr m_ptrTestMem->ConnectSignalTriplePtrTab(boost::bind(&CloneAppRecipe::test, this)); } void CloneAppRecipe::test() const { } void CTestMem::ConnectSignalTriplePtrTab(const TypeSignaTest::slot_type& slot) { //m_ptrTestTab are boost::shared_ptr m_ptrTestTab1->ConnectSignal(slot); m_ptrTestTab2->ConnectSignal(slot); m_ptrTestTab3->ConnectSignal(slot); } void CTestTab::ConnectSignal(const TypeSignalTest::slot_type& slot) { m_SignalTest->connect(slot); } CTestTab::~CTestTab() { m_SignalTest->disconnect_all_slots(); }

2 5

std::vector< boost::shared_ptr<int> >::pop_back()
by bringiton bringiton 10 Jul '06

10 Jul '06

This question is based on curiosity. How does std::vector::pop_back() call the destructor of the item getting removed? i understood std::vector to be a contigious array of memory, therefore an item's memory does not go out of scope when being popped. ie the item goes out of scope when the entire array goes out of scope. example code: the reference count is decremented when the item is popped from the vector. boost::shared_ptr<int> item1(new int(1)); std::vector< boost::shared_ptr<int> > vect1; vect1.push_back(item1); std::cout << item1.use_count() << std::endl; vect1.pop_back(); std::cout << item1.use_count() << std::endl;

3 5

vc-8_0-stlport target : Building Boost with VS2005 & STLPort 5 ?
by Sagnes, Frederic 10 Jul '06

10 Jul '06

Hello, Now that STLPort works for VS2005, I'd like to know if a BJam target for Visual Studio 2005 and STLPort 5.0.2 is available (as a patch for the 1.33 Boost distribution). Is STLPort 5.0 too young for inclusion, or is this target to be available only in the 1.34 version of the Boost libraries distribution? Regards _______________________________________________ Frédéric Sagnes, A&D AS RD DH K3 Siemens AG Karlsruhe

1 0

[boost.python] use of multi_array in python
by Sebastien Fortier 10 Jul '06

10 Jul '06

I'm having some problems with the use of multi_array's with python. What I'm trying to do is to export some functions that create multi_array's and pass these multi_array's to other functions that will extract other array's from them and display them. Lets say I have this as an example. //ArrayRef2D.h typedef for a 2D multi_array of doubles #ifndef ARRAYREF2D_H_ #define ARRAYREF2D_H_ #include <boost/multi_array.hpp> typedef boost::const_multi_array_ref < double, 2 > const_array_type2D_ref_d_t; typedef boost::multi_array < double, 2 > array_type2D_d_t; #endif /*ARRAYREF2D_H_*/ //Fstd.h 2D multi_array generator #ifndef FSTD_H_ #define FSTD_H_ #include "ArrayRef2D.h" #include <string> using namespace std; class Fstd { public: Fstd(string filename) : filename_(filename){} array_type2D_d_t get2Dfield() { array_type2D_d_t matrix(boost::extents[2][2]); matrix[0][0] = 0; matrix[0][1] = -1; matrix[1][0] = 20; matrix[1][1] = -30; return matrix; } private: string filename_; }; #endif /*FSTD_H_*/ //MatrixAlgorithm2D.h provides methods for displaying and manipulating matrices #ifndef MATRIXALGORITHM2D_H_ #define MATRIXALGORITHM2D_H_ #include "ArrayRef2D.h" #include <cassert> #include <cmath> #include <iostream> using namespace std; class MatrixAlgorithm2D { public: static array_type2D_d_t windChill(const_array_type2D_ref_d_t TT, const_array_type2D_ref_d_t UV) { assert((TT.shape()[0] == UV.shape()[0]) && (TT.shape()[1] == UV.shape()[1])); array_type2D_d_t RE(boost::extents[TT.shape()[0]][TT.shape()[1]]); for (unsigned int x = 0; x < TT.shape()[0]; x++) { for (unsigned int y = 0; y < TT.shape()[1]; y++) { if ((TT[x][y] <= 0.0) && (UV[x][y] >= 5.0)) { RE[x][y] = ( 13.1200 + 0.6215 * (TT[x][y]) + (0.3965 * (TT[x][y]) - 11.3700) * (pow((UV[x][y]),0.16))); } else { RE[x][y] = (TT[x][y]); } } } return RE; } static void display_d(const_array_type2D_ref_d_t a) { for (unsigned int x = 0; x < a.shape()[0]; x++) { for (unsigned int y = 0; y < a.shape()[1]; y++) { cout << a[x][y]; if(y < a.shape()[1] - 1) { cout << ", "; } } cout<<endl; } } }; #endif /*MATRIXALGORITHM2D_H_*/ //matrix.cpp boost.python interface file generated with pyste // Boost Includes ============================================================== #include "boost/python.hpp" #include "boost/cstdint.hpp" // Includes ==================================================================== #include "Fstd.h" #include "MatrixAlgorithm2D.h" // Using ======================================================================= using namespace boost::python; // Module ====================================================================== BOOST_PYTHON_MODULE(matrix) { class_< Fstd >("Fstd", init< const Fstd& >()) .def(init< std::string >()) .def("get2Dfield", &Fstd::get2Dfield) ; class_< MatrixAlgorithm2D >("MatrixAlgorithm2D", init< >()) .def(init< const MatrixAlgorithm2D& >()) .def("windChill", &MatrixAlgorithm2D::windChill) .def("display_d", &MatrixAlgorithm2D::display_d) .staticmethod("windChill") .staticmethod("display_d") ; } I generated my so file with bjam and then Python 2.3.5 (#2, Sep 4 2005, 22:01:42) [GCC 3.3.5 (Debian 1:3.3.5-13)] on linux2 Type "help", "copyright", "credits" or "license" for more information. >>> import matrix >>> dir(matrix) ['Fstd', 'MatrixAlgorithm2D', '__doc__', '__file__', '__name__'] >>> m = matrix.Fstd("HELLO") >>> l = m.get2Dfield() Traceback (most recent call last): File "<stdin>", line 1, in ? TypeError: No to_python (by-value) converter found for C++ type: boost::multi_array<double, 2, std::allocator<double> > So you see I can't use my code, what am I doing wrong? I would have liked to declare 2 arrays with get2Dfield() and pass these array to my windChill function to get another 2d matrix which I could then display with my display_d function. -- Sébastien Fortier

2 8

[shmem] Bug I can't workaround
by Berenguer Blasi 10 Jul '06

10 Jul '06

Hi, first of all congratulations to Ion for this great library. I am trying to use shmem for a very large project and I have stumbled several bugs (you can read about them at the sourceforge bugtracker) I have workarounded so far, but there's one I can't workaround. Maybe somebody has found a solution? How to reproduce: 1- Any sample code will do (you can use the sample source at 'Quick Guide for the Impatient' -> 'Using STL containers and mapping the memory at a fixed address') 2- Change to a fixed_named segment, if it isn´t so already in your sample, and increase the segment size to 104857600. I guess it's not a matter of this specific size but something that happens for large segments. 3- Leave process A running in the background, run process B and you get a segmentation fault... 4- You won't have the problem for sizes: 1048576 or 10485760. Has anybody experienced something similar? Thanks in advance. It's driving me crazy! and I am so close to finishing my code. Only this bug pending! Environment: FC 5, gcc4.1.1-1

1 0

[variant].[apply_visitor] Binary visitation: Build related problem.
by Prashant Thakre 10 Jul '06

10 Jul '06

Hi, Compiler instantiates the second Binary visitor instead of the first one even though the template parameters are same. regards, Prashant --------------------------------------------------------------------------- struct add_tuple_visitor: public boost::static_visitor<bool> { // Expected this to be instantiated by the compiler. // 1. template <typename T> bool operator()( T & lhs, const T & rhs) const { lhs += rhs ; return true; } // Compiler considers this as an appropriate fit .. // 2. template <typename T, typename U> bool operator()( T & lhs, const U & rhs) const { lhs += rhs ; return true; } }; struct agg_calc { public: template <typename T> bool operator () ( T & lhs, const T & rhs ) const { for( std::size_t curr_oper = 0; curr_oper < num_operations; ++v ) { switch( agg_operators.at(curr_oper) ) { case '+' : boost::apply_visitor( agg_visitor, lhs.at(curr_oper), //vector of variants rhs.at(curr_oper) //vector of variants ); break; default: break; }; } // for loop return true; } //Store the aggregation operators agg_types agg_operators; std::size_t num_operations; // Create visitor objects for delayed form of apply_visitor add_tuple_visitor agg_visitor; }; ---------------------------------------------------------------------------

1 0