December 2008 - Boost-users

[Serialization]Can not export polymorphic types
by Daniel Roberts 29 Dec '08

29 Dec '08

I made my own test program similar to test_no_rtti.cpp, only simpler. I have one base class and two classes that derive from it. All three classes are registered using BOOST_CLASS_EXPORT. I create an instance of each derived class and serialize them to an xml file. I then reload the file and try to serialize to a base class pointer and an "unregistered class" exception is thrown. I have stepped through my code and most of the boost serialization code and can not find a problem. The xml file looks good after creation but for some reason boost can not serialize it in properly. I have attached my short program. Can someone please tell me what I am doing wrong? Many thanks, Daniel Roberts

3 2

Re: [Boost-users] [Serialization]Serializing from base class pointers clips derived class
by Daniel Roberts 29 Dec '08

29 Dec '08

Thanks for the response. Under closer inspection, the serialize calls are never made for either the derived classes or the base class when using a base class pointer. This is the xml output after save_derived is called: <?xml version="1.0" encoding="UTF-8" standalone="yes" ?> <!DOCTYPE boost_serialization> <boost_serialization signature="serialization::archive" version="5"> <rd1 class_id="0" tracking_level="1" version="0" object_id="_0"> <polymorphic_base class_id="1" tracking_level="0" version="0"></polymorphic_base> </rd1> <rd2 class_id="2" tracking_level="1" version="0" object_id="_1"> <polymorphic_base></polymorphic_base> </rd2> <rb1 class_id_reference="0" object_id_reference="_0"></rb1> <rb2 class_id_reference="2" object_id_reference="_1"></rb2> </boost_serialization> I am not sure where rb1 and rb2 tags come from since the serialize functions never get called. load_derived also does not call the serialize functions when using the base class pointers. I have attached the code I am using. It's identical to test_no_rtti.cpp except for two items. I do not have test_tools.hpp so I commented corresponding functions out and switched to std::ofstream/std::ifstream and xml_oarchive/xml_iarchive. Also, my target is a win32 console application. Please let me know if you have any ideas. Thanks, Daniel Roberts

2 1

[mpl][variant] Visitor with typelists
by Tilo Nitzsche 29 Dec '08

29 Dec '08

What is the best (shortest/most elegant) way to create a visitor for a Variant, where there are subsets of the variant types, for which the operation is the same? typedef boost::mpl::vector<A, B, C> v1; typedef boost::mpl::vector<D, E> v2; typedef boost::mpl::joint_view<v1, v2> v; typedef boost::make_variant_over< v >::type v_union; struct visitor : public boost::static_visitor<int> { int operator()([T in v1] arg) { // do something for A, B, C return arg.a; } int operator()([T in v2] arg) { // do something for D, E return arg.b; } }; Thanks, Tilo

2 2

[thread]Is shared_mutex recursive with respect to shared locks?
by Ken Savela 29 Dec '08

29 Dec '08

In other words, should the following code result in a deadlock? #include <iostream> #include <boost/thread.hpp> int main(int argc, char* argv[]) { boost::shared_mutex smx; boost::upgrade_lock<boost::shared_mutex> lock_one(smx); std::cout << "Locked once." << std::endl; std::cout.flush(); boost::upgrade_lock<boost::shared_mutex> lock_two(smx); std::cout << "Locked again." << std::endl; std::cout.flush(); boost::upgrade_to_unique_lock<boost::shared_mutex> lock_three(lock_two); std::cout << "Write lock." << std::endl; return 0; } This example illustrates the problem I'm having in my app -- I have a situation where a single thread may end up performing more than a single shared lock. When it does so, it hangs on the nested upgrade_lock. Is this the expected behavior?

2 1

Lambda: Accessing fields of 'var'ed variable
by Swidan, Firas 29 Dec '08

29 Dec '08

Hi, I am trying to deal with the following situation: class tr { public: int i; int j; }; vector<tr> v(10); vector<uint> idx(5); for_each(idx.begin(), idx.end(), var(v)[_1].i = 1); This does not compile though, because lambda_functor has no member named i. My understanding is that I need the var before the v (also removing the var results in a compilation error). Am I on the right path? Thanks, Firas. ------------- Dr Firas Swidan, PhD Tel: +972.4.486.6278 E-mail: firas.swidan(a)intel.com<mailto:firas.swidan@intel.com> Intel Israel (74) Ltd. P.O.Box 1659, Haifa 31015, Israel --------------------------------------------------------------------- Intel Israel (74) Limited This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.

2 1

Re: [Boost-users] [shared_ptr] Is it possible for a shared_ptr to
by Max 29 Dec '08

29 Dec '08

There seems to be a solution that can make the code a little bit cleaner and easier to maintain. #define DECLARE_PTR_AUTO_DELETOR(CLASS, VAR_PTR)\ boost::scoped_ptr<CLASS> auto__deletor__for__ ## VAR_PTR (VAR_PTR) AcDbCircle *pFace =new AcDbCircle (AcGePoint3d::kOrigin, AcGeVector3d::kZAxis, 1.0) ; AcDbCircle *pLeftEye =new AcDbCircle (AcGePoint3d (0.33, 0.25, 0.0), AcGeVector3d::kZAxis, 0.1) ; AcDbCircle *pRightEye =new AcDbCircle (AcGePoint3d (-0.33, 0.25, 0.0), AcGeVector3d::kZAxis, 0.1) ; AcDbArc *pMouth =new AcDbArc (AcGePoint3d (0, 0.5, 0), 1.0, 3.141592 + (3.141592 * 0.3), 3.141592 + (3.141592 * 0.7)) ; DECLARE_PTR_AUTO_DELETOR(AcDbCircle, pFace); DECLARE_PTR_AUTO_DELETOR(AcDbCirclep, LeftEye); DECLARE_PTR_AUTO_DELETOR(AcDbCircle, pRightEye); DECLARE_PTR_AUTO_DELETOR(AcDbArc, pMouth); // Set the color property. pFace->setColorIndex (2) ; pLeftEye->setColorIndex (5) ; pRightEye->setColorIndex (5) ; pMouth->setColorIndex (1) ; // add the entities to the new block table record if ( (es =pBlockTableRecord->appendAcDbEntity (pFace)) != Acad::eOk ) { //delete pFace ; // all these 'delete ...' not needed any more ... //delete pLeftEye ; //delete pRightEye ; //delete pMouth ; pBlockTableRecord->erase () ; pBlockTableRecord->close () ; return (es) ; } pFace->close () ; if ( (es =pBlockTableRecord->appendAcDbEntity (pLeftEye)) != Acad::eOk ) { //delete pLeftEye ; //delete pRightEye ; //delete pMouth ; pBlockTableRecord->erase () ; pBlockTableRecord->close () ; return (es) ; } pLeftEye->close () ; if ( (es =pBlockTableRecord->appendAcDbEntity (pRightEye)) != Acad::eOk ) { //delete pRightEye ; //delete pMouth ; pBlockTableRecord->erase () ; pBlockTableRecord->close () ; return (es) ; } pRightEye->close () ; if ( (es =pBlockTableRecord->appendAcDbEntity (pMouth)) != Acad::eOk ) { //delete pMouth ; pBlockTableRecord->erase () ; pBlockTableRecord->close () ; return (es) ; } pMouth->close () ; Thanks you all that hepled. B/Rgds Max ----- Original Message ----- From: Max <loadcom(a)sina.com> To: Emil Dotchevski<emildotchevski(a)gmail.com> Subject: Re: [Boost-users] [shared_ptr] Is it possible for a shared_ptr to Date: 2008-12-26 09:22:00 Hello Dotchevski, Thanks for your explanation. This question was pushed to me by seeing the following code in an Autodesk AutoCAD ObjectARX sample program. Don't pay much attention to the descrete classes, just to the new/delete operations. I found the if/delete guards to prevent memory leakage too tedious. I hope there's a smart_ptr-like technique that could help me circumvent this verbose coding style. In other words, my need could be described as: 1) allocate objects dynamically on heap, 2) send them to algorithms that are prone to failure 3) if the opreation succeeds, the objects will be taken care of by the receiver of these objects, or, otherwise, they will be automatically freed without having to delete these objects manually. AcDbCircle *pFace =new AcDbCircle (AcGePoint3d::kOrigin, AcGeVector3d::kZAxis, 1.0) ; AcDbCircle *pLeftEye =new AcDbCircle (AcGePoint3d (0.33, 0.25, 0.0), AcGeVector3d::kZAxis, 0.1) ; AcDbCircle *pRightEye =new AcDbCircle (AcGePoint3d (-0.33, 0.25, 0.0), AcGeVector3d::kZAxis, 0.1) ; AcDbArc *pMouth =new AcDbArc (AcGePoint3d (0, 0.5, 0), 1.0, 3.141592 + (3.141592 * 0.3), 3.141592 + (3.141592 * 0.7)) ; // Set the color property. pFace->setColorIndex (2) ; pLeftEye->setColorIndex (5) ; pRightEye->setColorIndex (5) ; pMouth->setColorIndex (1) ; // add the entities to the new block table record if ( (es =pBlockTableRecord->appendAcDbEntity (pFace)) != Acad::eOk ) { delete pFace ; delete pLeftEye ; delete pRightEye ; delete pMouth ; pBlockTableRecord->erase () ; pBlockTableRecord->close () ; return (es) ; } pFace->close () ; if ( (es =pBlockTableRecord->appendAcDbEntity (pLeftEye)) != Acad::eOk ) { delete pLeftEye ; delete pRightEye ; delete pMouth ; pBlockTableRecord->erase () ; pBlockTableRecord->close () ; return (es) ; } pLeftEye->close () ; if ( (es =pBlockTableRecord->appendAcDbEntity (pRightEye)) != Acad::eOk ) { delete pRightEye ; delete pMouth ; pBlockTableRecord->erase () ; pBlockTableRecord->close () ; return (es) ; } pRightEye->close () ; if ( (es =pBlockTableRecord->appendAcDbEntity (pMouth)) != Acad::eOk ) { delete pMouth ; pBlockTableRecord->erase () ; pBlockTableRecord->close () ; return (es) ; } pMouth->close () ; Thanks for your patience! Best regards Max ------------------------------------------------------------------- 新浪空间——与朋友开心分享网络新生活！(http://space.sina.com.cn/ )

1 0

[asio]: Having trouble getting started with ssl
by Roland Bock 29 Dec '08

29 Dec '08

Hi, I tried to follow the documentation for ssl in boost::asio (Overview/SSL), coming up with a few lines of code which fail to compile with boost-1.36 (gcc-4.2.4, Ubuntu8.04, 64bit) with the following message: sslTest.cpp:13: error: invalid initialization of reference of type ‘boost::asio::basic_stream_socket<boost::asio::ip::tcp, boost::asio::stream_socket_service<boost::asio::ip::tcp> >&’ from expression of type ‘boost::asio::basic_socket<boost::asio::ip::tcp, boost::asio::stream_socket_service<boost::asio::ip::tcp> >’ The message is pretty clear, but I have no idea yet how to fix the problem. Here is the code: // ----------------------------------------- #include <boost/asio.hpp> #include <boost/asio/ssl.hpp> using namespace std; using namespace boost::asio; int main() { io_service ioService; ssl::context ctx(ioService, ssl::context::sslv23); ssl::stream<ip::tcp::socket> sslSock(ioService, ctx); ip::tcp::socket& sock = sslSock.lowest_layer(); } // ----------------------------------------- Regards, Roland

2 3

Seeking example graph program for newbies
by Roger House 29 Dec '08

29 Dec '08

I've just purchased The Boost Graph Library book and downloaded the CD from the publisher's website onto my Ubuntu system. I had to go through a number of contortions to unpack the CD because it is in a "sit" file format and it seems that there is no readily available free software for operating on sit files. I was a bit surprised to discover that the CD does not seem to contain a complete example program which I can compile and run to see that I'm able to access the boost header files and libraries on my system. There is a pdf version of the book, but as far as I can see, there is no example of a complete, simple "hello, graph" program in the book or on the CD. Could someone please point me to such a program. Thank you, Roger House Software Developer

2 1

Re: [Boost-users] [shared_ptr] Is there an interface to set the ptr a
by Max 29 Dec '08

29 Dec '08

Sorry, this version is incorrect: it should be #define DECLARE_VAR_AUTO_CLEANER_BY_MEM_FN(CLASS, VAR, MEM_FN_CLEANER)\ boost::shared_ptr<CLASS> auto__cleaner__for__ ## VAR\ (&VAR, boost::bind(&CLASS::MEM_FN_CLEANER, _1)); #define DECLARE_PTR_AUTO_CLEANER_BY_MEM_FN(CLASS, VAR_PTR, MEM_FN_CLEANER)\ boost::shared_ptr<CLASS> auto__cleaner__for ## VAR_PTR\ (VAR_PTR,\ boost::lambda::if_then(boost::lambda::_1!=static_cast<CLASS*>(0),\ boost::lambda::bind(&CLASS::MEM_FN_CLEANER, boost::lambda::_1))) // use case: struct A { void close() { delete this; } }; { A *pa = new A(); DECLARE_PTR_AUTO_CLEANER_BY_MEM_FN(A, pa, close); } // pa is cleaned by its close() function automatically Max ============================================================= Further, I've written the following macros for such use case: #define DECLARE_VAR_PTR_WITH_AUTO_MEM_CLEANER(CLASS, VAR_PTR, MEM_FN_CLEANER)\ CLASS *VAR_PTR;\ boost::shared_ptr auto__cleaner__for__ ## VAR_PTR\ (VAR_PTR, boost::bind(&CLASS::MEM_FN_CLEANER, _1)); #define DECLARE_VAR_WITH_AUTO_MEM_CLEANER(CLASS, VAR, MEM_FN_CLEANER)\ CLASS VAR;\ boost::shared_ptr auto__cleaner__for__ ## VAR\ (&VAR, boost::bind(&CLASS::MEM_FN_CLEANER, _1)); Usage: DECLARE_VAR_PTR_WITH_AUTO_MEM_CLEANER(AcDbBlockTable, pBlockTable, close); Max ------------------------------------------------------------------- 新浪空间——与朋友开心分享网络新生活！(http://space.sina.com.cn/ )

1 0

[Serialization]Serializing from base class pointers clips derived class
by Daniel Roberts 29 Dec '08

29 Dec '08

The serialization example, test_no_rtti.cpp, appears to only write the base class XML element when saving from the base class pointer. Perhaps I am misunderstanding the point of the extended_type_info classes but I thought it was a way to serialize classes with a common base class without having to know the type of the derived class. The example in test_no_rtti.cpp clips the derived portion of the class when serializing from a base class pointer. Please advise. Thanks.

2 1