- Boost-users - lists.preview.boost.org

version control choice
by Mihir Thakkar 12 Dec '13

12 Dec '13

Hey all, I am using Ubuntu and Git. Would it be alright if I work on Git? Is everything available on Git repository as well?

2 1

slow exception handling
by XIN LI 12 Dec '13

12 Dec '13

Hi, In my code, I add some protection to catch the exception from boost like "catch (boost::archive::archive_exception& ex)", strangely, it slow exception handling, causing one of our tests to go from around 30 seconds to greater than 5 minutes. I did profiling and found "linear_searc_fdes" spent too much time when process exception now. Does anyone have similar experience, why this happen, and how to fix the problem? Thanks a lot! Best wishes, Andy Li

3 4

How do I build libraries with binaries using two versions of the same toolset
by Ted Byers 11 Dec '13

11 Dec '13

On my OpenSuse box, version 12.3, the default version of gcc is 4.7. I just successfully built gcc 4.8.2, with the binary names ending with 48 (so I have gcc48, g++48, &c.). So, I have two versions of gcc on this machine, and I believe I can use either by setting CC to the one or the other in my make files. Hence my problem. I have built boost 1.55.0 on my Ubuntu 12.04 box (but with the default names, so that gcc -v tells me that I am using gcc 4.8.2: the default version of gcc on that version of Ubuntu is 4.6, which is too old for me to consider). So, a basic build of the boost libraries is simple, and something I have done before. But I do not know how to build boost twice, once with gcc 4.7 and once with gcc 4.8.2, and have them installed in such a way as to ensure that the two builds do not clash and so that if I set CC to the one or the other version of gcc, the right build of boost will automagically be selected (it would be appreciated if someone who knows Gnu make better than I do could show me how to effect that in my makefiles). Obviously I am assuming that a single build of boost will not work with both compilers (is that assumption correct?). Unfortunately, I have not yet found anything in the documentation that talks about building boost with different versions of a given toolset on the same machine. Any help would be appreciated. Thanks Ted

2 3

[spirit::karma | variant] Nested Variants
by Aaron_Wright＠selinc.com 11 Dec '13

11 Dec '13

I am trying to use spirit::karma to generate a CSV file. I want to output a vector of variants. Unfortunately, I also want to add one more type to the variant; a string. So I was thinking I could nest variants. I can get this code to compile, but only the strings print out. What am I doing wrong? template< typename Iterator > class Generator : public boost::spirit::karma::grammar< Iterator, std::vector< boost::variant< std::string, boost::variant< bool, int > > >() > { public: Generator() : Generator::base_type(start) { using boost::spirit::karma::string; using boost::spirit::karma::bool_; using boost::spirit::karma::int_; using boost::spirit::karma::eol; start = (string | bool_ | int_) % Separator() << eol; } boost::spirit::karma::rule< Iterator, Csv_generator_input() > start; }; int main(int argc, char** argv) { std::vector< boost::variant< std::string, boost::variant< bool, int > > > values; values.push_back(std::string("Hello")); values.push_back(true); values.push_back(false); values.push_back(10); values.push_back(std::string("World")); std::string generated; std::back_insert_iterator< std::string > sink(generated); boost::spirit::karma::generate(sink, Generaator(), values); // generated == "Hello,World", Expected "Hello,true,false,10,World" return 0; } --- Aaron Wright

1 0

Problem using a filtered graph in the Ressource constrained shortest path user defined extension and dominance functions
by Berit Dangaard Brouer 11 Dec '13

11 Dec '13

Hi, In our use of the boost ressource constrained shortest path from Boost BGL, we define problem specific ressource extension and dominance function on a user defined graph type mcf_graph (source code is posted below) Our problem is the use of a filtered graph into the resource function. The forfeited edge filter stated in the code below, tries to remove some of the edges in order to speed up the algorithm. It seems that the resource and extension functions do not know how to deal with the filtered graph (I am guessing because they are not template functions), although the filtered graph contains the same Edge and Vertex definitions as the MCF graph, it is not typewise recognised as an mcf_graph, because it also contains the filter map. Is there any way I can work around this? I tried making the resource and extension functions into templates, but this is a problem because my edge bundled properties are problem specific. I would truly appreciate any suggestions as a filter is preferable to copying the graph. mcf_graph.h #ifndef MCF_GRAPH_HPP #define MCF_GRAPH_HPP #include <boost/config.hpp> #include <boost/graph/adjacency_list.hpp> #include <iostream> using namespace boost; using namespace std; struct Port { Port(int index=-1, int type=-1, int port_code=0, string UN="NN", string name="NN", int service_id=-1 ): m_index(index), m_type(type), m_port_code( port_code), UNLOCODE(UN), m_name(name), m_service_id(service_id) {} int m_index; int m_type;//0 port, 1 port call, 2 transhipment int m_port_code; string UNLOCODE; string m_name; int m_service_id; }; struct Edge { Edge( int id=-1, int type=-1, double weight=0.0, double cost=0.0, double dual_cost=0.0, double util=0.0, double cap=0.0, double r_cap=0.0, double tt=0.0 , int rot=-1) : m_idx(id), m_type(type), m_weight(weight) , m_cost(cost), m_dual_cost(dual_cost), m_utilization(util), m_capacity(cap), m_res_cap(r_cap), m_transittime(tt), m_rot(rot) {} unsigned int m_idx; int m_type;//0 is load edge, 1 is voyage edge, 2 forfeited edge (commodity link), 3 transhipment edge, 4 transhipment ring edge double m_weight; //distance double m_cost; //cost - only on load/unload/transhipment edges double m_dual_cost; double m_utilization; double m_capacity; double m_res_cap; //residual capacity double m_transittime; // transit time on edge with current speed double m_rot; // current rotation }; //Graph typedef adjacency_list< vecS , vecS , bidirectionalS, Port, Edge> mcf_graph; //descriptor typedefs typedef boost::graph_traits<mcf_graph>::vertex_descriptor vertex_descriptor; typedef boost::graph_traits<mcf_graph>::edge_descriptor edge_descriptor; //iterator typedefs typedef boost::graph_traits<mcf_graph>::vertex_iterator Viterator; typedef boost::graph_traits<mcf_graph>::edge_iterator Eiterator; typedef boost::graph_traits<mcf_graph>::in_edge_iterator IEiterator; typedef boost::graph_traits<mcf_graph>::out_edge_iterator OEiterator; template <typename edge_type_map> struct forfeited_edge_type { forfeited_edge_type() { } forfeited_edge_type( edge_type_map type) : forfeited(type) { } bool operator()(const edge_descriptor& e) const { return forfeited[e]==2; } edge_type_map forfeited; }; #endif And resource_extension and dominance functions here in res_path.h #ifndef RESPATH_H_ #define RESPATH_H_ #include <boost/config.hpp> #include <boost/graph/adjacency_list.hpp> #include <boost/graph/r_c_shortest_paths.hpp> #include <iostream> #include "mcf_graph.h" #include "data.h" #include <boost/graph/graphviz.hpp> namespace std { class respath { public: respath(mcf_graph & m, Eiterator & comm); virtual ~respath(); private: mcf_graph & m_res_graph; Eiterator & commodities; }; // data structures for shortest path problem with transit time constraints (spptt) // ResourceContainer model struct spp_spptt_res_cont { spp_spptt_res_cont( double c = 0, double t = 0 ) : cost( c ), time( t ) {} spp_spptt_res_cont& operator=( const spp_spptt_res_cont& other ) { if( this == &other ) return *this; this->~spp_spptt_res_cont(); new( this ) spp_spptt_res_cont( other ); return *this; } double cost; double time; }; bool operator==( const spp_spptt_res_cont& res_cont_1,const spp_spptt_res_cont& res_cont_2 ){ return ( res_cont_1.cost == res_cont_2.cost && res_cont_1.time == res_cont_2.time ); } bool operator<( const spp_spptt_res_cont& res_cont_1,const spp_spptt_res_cont& res_cont_2 ) { if( res_cont_1.cost > res_cont_2.cost ) return false; if( res_cont_1.cost == res_cont_2.cost ) return res_cont_1.time < res_cont_2.time; return true; } // ResourceExtensionFunction model class ref_spptt { static double maxtt; public: static void setmaxtt(double tt){ maxtt=tt; } inline bool operator()( const mcf_graph& g, spp_spptt_res_cont& new_cont, const spp_spptt_res_cont& old_cont, boost::graph_traits <mcf_graph>::edge_descriptor ed) const { const Edge& arc_prop = get( edge_bundle, g )[ed]; new_cont.cost = old_cont.cost + arc_prop.m_cost; new_cont.time=old_cont.time + arc_prop.m_transittime; double& i_time = new_cont.time; return i_time <= (maxtt) ? true : false; } }; double ref_spptt::maxtt=0; // Dominance Function class dom_spptt { public: inline bool operator()( const spp_spptt_res_cont& res_cont_1, const spp_spptt_res_cont& res_cont_2 ) const { return res_cont_1.cost <= res_cont_2.cost && res_cont_1.time <= res_cont_2.time; } }; } /* namespace std */ #endif /* RESPATH_H_ */ -- Best Regards, Ph.D. Berit Dangaard Brouer Post Doc. DTU Management Engineering, Management Science Technical University of Denmark Department of Management Engineering Produktionstorvet, Bygning 426 2800 Kgs. Lyngby blof(a)dtu.dk <mailto:blof@man.dtu.dk> www.ms.man.dtu.dk <http://www.ms.man.dtu.dk> http://www.maersk.com/innovation/leadingthroughinnovation/enerplan/ Please read about the ENERPLAN project here <http://www.maersk.com/innovation/leadingthroughinnovation/enerplan/>

1 0

multi_index in managed_shared_memory
by Sebastien Nobert 11 Dec '13

11 Dec '13

Hi All, I'm trying to build a multi_index in managed shared memory with two (or more) hashed index but I get an "is an amiguous base of" error. I've already done that with one hash_uniq and a sequenced index and everything work fine. With two hashed index i get. boost/interprocess/detail/mpl.hpp:74:55: error: > ‘boost::multi_index::detail::hashed_index_node_impl<boost::interprocess::allocator<char, > boost::interprocess::segment_manager<char, > boost::interprocess::rbtree_best_fit<boost::interprocess::mutex_family>, > boost::interprocess::iset_index> > >’ is an ambiguous base of > ‘boost::multi_index::detail::hashed_index_node<boost::multi_index::detail::hashed_index_node<boost::multi_index::detail::index_node_base<element, > boost::interprocess::allocator<element, > boost::interprocess::segment_manager<char, > boost::interprocess::rbtree_best_fit<boost::interprocess::mutex_family>, > boost::interprocess::iset_index> > > > >’ > static const bool value = sizeof(dispatch(trigger())) == > sizeof(true_t); > With a hashed and an ordered index i get. boost/multi_index_container.hpp:161:52: error: no type named > ‘final_allocator_type’ in ‘boost::multi_index::multi_index_container< ... > and it ask me to attach the preprocessed file to my bug report. This is compiled on linux with gcc version 4.8.2 20131017 (Red Hat 4.8.2-1) (GCC) Should I open a bug report ? Am I doing something wrong ? Thanks a lot. Seb

3 4

[Log] Attribute formatting tricks ineffective -- my enums are printed as numbers.
by Andres Jaan Tack 11 Dec '13

11 Dec '13

I'm having some trouble with attribute formatting in boost::log that I haven't managed to figure out on my own. The case seems basic: I want to format my own severity enum as a string, for which I've produced the following (failing) example. Tested on clang-3.3 and clang-3.4 with boost-1.54 from macports, the minimal failing test produces the following output: $ clang++ --std=c++11 -c -DBOOST_ALL_DYN_LINK -I/opt/local/include/boost fail.cxx $ g++ -o fail -L/opt/local/lib -lboost_log-mt -lboost_log_setup-mt -lboost_system-mt -lboost_thread-mt fail.o $ ./fail 1: Booo! I would have expected "INFO: Booo!" according to my code. I've attached the complete fail.cxx here, in which I do all of: 1) Employ BOOST_LOG_ATTRIBUTE_KEYWORD, 2) Overload operator<< (std::ostream&, const my::severity&), 3) Overload operator<< (logging::formatting_ostream&, const logging::to_log_manip<...>&), and 4) Invoke register_simple_formatter_factory I can't find any guidance in the documentation or elsewhere on the internet that has managed to solve this -- Am I missing something obvious, or have I found a failing test case? -- Andres

1 0

Introduction
by Mihir Thakkar 10 Dec '13

10 Dec '13

Hello! I am new to open source, I know c++ and i am learning neural networks. I would like to contribute to boost c++ libraries. Please help me to get started. :)

2 1

Problem with linking files using boost::log with CMake Buildfile
by David Georg Reichelt 10 Dec '13

10 Dec '13

Hi, I've got problems compiling a program using boost with CMake (using Windows and MinGW 4.7.2, with -G "MinGW Makefiles"). Boost is compiled normaly with b2.exe --toolset=gcc --prefix=C:\cppLib\boost_1_55_0 and the folder seems to contain the right files. The original CMake-File compiled fine using Ubuntu. I've got a code using only boost::log and a CMake File containing the following boost-related-code: ADD_DEFINITIONS("-std=c++0x") SET(BOOST_ROOT "C:/cppLib/boost_1_55_0/") link_libraries (${Boost_LIBRARIES}) find_package( Boost 1.54.0 COMPONENTS log) if(Boost_FOUND) MESSAGE("Boost found") include_directories(${Boost_INCLUDE_DIRS}) add_executable(Hasselbrack ${AUTO_SOURCES}) endif() This code works right, but when I call make, I get the following errors: In file included from C:/cppLib/boost_1_55_0/boost/log/expressions/formatters/c_decorator.hpp:21:0, from C:/cppLib/boost_1_55_0/boost/log/expressions/formatters.hpp:29, from C:/cppLib/boost_1_55_0/boost/log/expressions.hpp:26, from C:\Users\DaGeRe\cpp\Minimalprojekt\main.cpp:3: C:/cppLib/boost_1_55_0/boost/log/detail/snprintf.hpp:45:9: Fehler: ╗::swprintf½ wurde nicht deklariert C:/cppLib/boost_1_55_0/boost/log/detail/snprintf.hpp:46:9: Fehler: ╗::vswprintf½ wurde nicht deklariert In file included from C:/cppLib/boost_1_55_0/boost/log/expressions/formatters.hpp:29:0, from C:/cppLib/boost_1_55_0/boost/log/expressions.hpp:26, from C:\Users\DaGeRe\cpp\Minimalprojekt\main.cpp:3: C:/cppLib/boost_1_55_0/boost/log/expressions/formatters/c_decorator.hpp: In statischer Elementfunktion ╗static std::size_t boost::log::v2s_mt_nt5::exp ressions::aux::c_decorator_traits<wchar_t>::print_escaped(wchar_t (&)[N], wchar_t)½: C:/cppLib/boost_1_55_0/boost/log/expressions/formatters/c_decorator.hpp:110:43: Fehler: ╗swprintf½ ist kein Element von ╗boost::log::v2s_mt_nt5::aux½ I tried to figure it out, the lines in snprintf are the following: using ::snprintf; using ::vsnprintf; but I can't find an error in this. The Include in the beginning of the file: #ifdef BOOST_LOG_USE_WCHAR_T #include <wchar.h> #endif // BOOST_LOG_USE_WCHAR_T seems to be exectued right (according to http://pubs.opengroup.org/onlinepubs/7990989775/xsh/wchar.h.html, snprintf is declared here). If I don't use ADD_DEFINITIONS("-std=c++0x"), I got the following errors: CMakeFiles\Hasselbrack.dir/objects.a(main.cpp.obj):main.cpp:(.text+0x12c): undefined reference to `boost::log::v2s_mt_nt5::core::get()' CMakeFiles\Hasselbrack.dir/objects.a(main.cpp.obj):main.cpp:(.text+0x143): undefined reference to `boost::log::v2s_mt_nt5::core::set_filter(boost::log ::v2s_mt_nt5::filter const&)' CMakeFiles\Hasselbrack.dir/objects.a(main.cpp.obj):main.cpp:(.text+0x1c7): undefined reference to `boost::log::v2s_mt_nt5::trivial::logger::get()' CMakeFiles\Hasselbrack.dir/objects.a(main.cpp.obj):main.cpp:(.text+0x20b): undefined reference to `boost::log::v2s_mt_nt5::trivial::logger::get()' CMakeFiles\Hasselbrack.dir/objects.a(main.cpp.obj):main.cpp:(.text+0x2e7): undefined reference to `std::ios_base::Init::~Init()' CMakeFiles\Hasselbrack.dir/objects.a(main.cpp.obj):main.cpp:(.text+0x308): undefined reference to `std::ios_base::Init::Init()' c:/mingw/bin/../lib/gcc/mingw32/4.7.2/../../../../mingw32/bin/ld.exe: CMakeFiles\Hasselbrack.dir/objects.a(main.cpp.obj): bad reloc address 0x1b in se ction `.text$_ZN5boost6detail15sp_counted_base7releaseEv[__ZN5boost6detail15sp_counted_base7releaseEv]' c:/mingw/bin/../lib/gcc/mingw32/4.7.2/../../../../mingw32/bin/ld.exe: final link failed: Invalid operation But nevertheless, boost is found and I can't figure out CMake seems to be unable to link it. The output of $(Boost_LIBRARIES) from CMake also seems correct: C:/cppLib/boost_1_55_0/stage/lib/libboost_log-mgw47-mt-1_55.dll . So I got no clue anymore what I could try. Has anybody an hint how to solve this? Best Regards, DaGeRe

4 8

Is the following use of boost::function<> safe?
by David Goodenough 10 Dec '13

10 Dec '13

Let me preface this with the note that the whole idea behind what I'm trying to do is invoke an arbitrary function in a new thread, optionally delaying the invocation by some number of milliseconds That said, ----- threadrun.h ----- #ifndef _THREADRUN_H_ #define _THREADRUN_H_ #include <boost/function.hpp> namespace Thread { void ThreadRun(boost::function<void()> const &job); void ThreadDelayedRun(boost::function<void()> const &job, unsigned int delay); }; #endif ----- threadrun.h ----- ----- threadrun.cpp ----- #include "threadrun.h" static void DelayedRun(boost::function<void()> job, unsigned int delay) { Sleepms(delay); // Routine available elsewhere that simply sleeps for delay ms job(); } namespace Thread { void ThreadRun(boost::function<void()> const &job) { boost::thread thr = boost::thread(job); thr.detach(); // Not really necessary, but it does make it explicit that we'll never wait for the thread to finish } void ThreadDelayedRun(boost::function<void()> const &job, unsigned int delay) { boost::thread thr = boost::thread(boost::bind(DelayedRun, job, delay)); thr.detach(); // Not really necessary, but it does make it explicit that we'll never wait for the thread to finish } }; ----- threadrun.cpp ----- Use cases would be: void myFunc(int a, int b) { ... } int myFunc1(int a, int b) { ... return x } Thread::ThreadRun(boost::bind(myFunc, 1000, 17)); Thread::ThreadDelayedRun(boost::bind(myFunc1, 2000, 42), 5000); This appears to work correctly, I'm currently using boost 1.54 - haven't build 1.55 yet, but will do so in the near future. My only concern is the second case where I have myFunc1() whose return type is int, yet in the prototype for ThreadDelayedRun() I'm making it look a whole lot like it's void. The return type ultimately doesn't matter since this is specifically designed to be a "fire and forget" system, which means that any return value will be lost, there's no way for the invoking thread to recover it. So, is there anything dangerous about what I'm doing?

3 2