- Boost-users - lists.boost.org

Building boost with custom gcc/g++
by arjun 10 Jan '13

10 Jan '13

Hi, I have issues in getting boost to build with a custom gcc installation in a separate folder, with default gcc in /usr/bin/. Describing the problem, and issues I am facing below. I would love to have a foolproof and right way to do this. I hope I can get some help here. Building boost with a custom compiler : Problem : 1. I have gcc/g++ in /usr/bin, and I have gcc 4.7.2 in /shared/tools/rtf/bin/gcc & /shared/tools/rtf/bin/g++ 2. I want to compile boost libraries using gcc 4.7.2 ONLY. Solutions : 1. If I export PATH as /shared/tools/rtf/bin:$PATH, so that "which gcc" picks the 4.7.2 one, then I run ./bootstrap.sh and ./b2 install, is that a correct way to do it? 2. Alternatively, would I have to change some .jam files like user-config.jam to point out gcc with using gcc : gcc-4.4 : /shared/tools/... gcc and then compile using Boost.build? Is this the recommended way? Issues/Confusions : 1. If I try to build using Boost.build way I first go to the directory tools/build/v2. Now 1st step to bootstrap, how to bootstrap it with /shared/tools/...gcc compiler? The only way I saw for that was to do like CC=/shared/tools/..gcc ./bootstrap.sh --with-toolset=cc as this actually picks from CC environment variable. Is this correct? 2. The subsequent step ./b2 install --prefix=INSTALL_PATH to build Boost.build fails saying cc.jam not present. How to get that jam file to be present? Should that be created manually? 3. I had also changed user-config.jam to have gcc-4.7 (using gcc : 4.7 : /shared/tools/..gcc). CC=/shared/tools/..gcc ./bootstrap.sh --with-toolset=cc toolset=gcc-4.7 Is this used during bootstrap.sh step? Or only when we give the actual boost build (running b2) 4. Another issue was when I run the bootstrap.sh inside tools/build/v2 (to get Boost.build built), If I give a --with-toolset option, I get the b2 and bjam executables in the current directory. If I do not give a with-toolset option, it automatically defaults to cc, in which case the b2,bjam are inside engine/bin.linux86_64 folder. Not able to understand this. Thanks, Arjun

4 8

Pushing iterators to far?
by Simon Knapp 10 Jan '13

10 Jan '13

Hi All, I'm trying to do some slightly hooky stuff with the iterators library and cannot get the following code to compile. When the second last line of main is uncommented ("ZIter iter(b);"), compilation fails. The error mesages suggest this is a problem with determining the appropriate traversal category of "ZIter". Can anyone advise if this is problem with my code, a bug, am I asking too much of the library, ...? Cheers, Simon Knapp. compiler: Visual Studio 2008 - version 9.0.30729.1 SP //---------------------------------------------------------------------------------------------------------------------------- // begin iterator_tests.cpp //---------------------------------------------------------------------------------------------------------------------------- #include <vector> #include <boost/iterator/filter_iterator.hpp> #include <boost/iterator/zip_iterator.hpp> #include <boost/tuple/tuple.hpp> typedef std::vector<long long> LLGrid; typedef std::vector<float> FloatGrid; typedef LLGrid::const_iterator LLGridIter; typedef FloatGrid::const_iterator FloatGridIter; typedef boost::tuples::tuple<LLGridIter, FloatGridIter> IDAndAreaTuple; typedef boost::zip_iterator<IDAndAreaTuple> IDAndAreaIterator; typedef std::iterator_traits<IDAndAreaIterator>::reference IDAndAreaRef; struct Fnctr { bool operator()(const IDAndAreaRef ref) const { return true; } }; typedef boost::filter_iterator<Fnctr, IDAndAreaIterator> FIter; typedef boost::tuples::tuple<FloatGridIter, FIter> ZIterTuple; typedef boost::zip_iterator<ZIterTuple> ZIter; static ZIterTuple makeZIterTuple( Fnctr const& fnctr, LLGridIter const& beginIdGrid, LLGridIter const& endIdGrid, FloatGridIter const& beginAreaGrid, FloatGridIter const& endAreaGrid, FloatGridIter const& gtgdIter ) { return boost::make_tuple( gtgdIter, boost::make_filter_iterator( fnctr, boost::make_zip_iterator(boost::make_tuple(beginIdGrid, beginAreaGrid)), boost::make_zip_iterator(boost::make_tuple(endIdGrid, endAreaGrid)) ) ); } int main(int argc, char* argv[]) { FloatGrid areaMask, regionStat; LLGrid regionMask; ZIterTuple b(makeZIterTuple(Fnctr(), regionMask.begin(), regionMask.end(), areaMask.begin(), areaMask.end(), regionStat.begin())); //ZIter iter(b); return 0; } //---------------------------------------------------------------------------------------------------------------------------- // end iterator_tests.cpp //----------------------------------------------------------------------------------------------------------------------------

2 1

[units]
by geocentrix 10 Jan '13

10 Jan '13

I am attempting to replace my own Quantity class with boost::units. One thing that is easy in my existing code is to convert a quantity from any unit of pressure into a specific unit, e.g. GPa (gigapascals), using a function ConvertedTo(). For example: Pressure converted_value = existing_value.ConvertedTo(GPa); How do you do this with boost:units.? I have the following things defined: typedef boost::units::quantity<boost::units::si::pressure> Pressure; static const Pressure Pa = 1.0 * boost::units::si::pascals; static const Pressure GPa = 1e9 * Pa; Pressure value(1.0^9 * Pa); How do I display value in GPa? I am missing something very simple and wondered if someone could advise, please? Thanks in advance. Andrew -- View this message in context: http://boost.2283326.n4.nabble.com/units-tp4641179.html Sent from the Boost - Users mailing list archive at Nabble.com.

2 1

[locale] how to sort utf8 std::string?
by Frédéric Bron 09 Jan '13

09 Jan '13

I read std::string from a file encoded in UTF-8. I would like to sort the strings. I think I have to first convert them to wide characters and then sort the resulting output. I have tried to do that with boost::local::conv::to_utf(utf8string, "UTF-8"); But it seems that all non ASCII characters are replaced by the question mark. Here is my program: #include <iostream> #include <fstream> #include <string> #include <boost/locale/encoding.hpp> int main(int argc, char **argv) { if (argc<2) return 0; std::ifstream is(argv[1], std::ios::binary); while (is) { std::string line; std::getline(is, line); if (not line.empty()) { try { std::wstring ws=boost::locale::conv::to_utf<wchar_t>(line, "UTF-8", boost::locale::conv::stop); std::wcout<<ws; } catch (...) { std::cout<<"exception\n"; } } } is.close(); return 0; } Frédéric

2 2

[astar search] Computation time for large graph
by Yann-Hervé Hellouvry 09 Jan '13

09 Jan '13

Hi, I try to use the "astar-cities" example but for graph of about 10000 nodes and edges (which are weighted) in order to find the fastest path between two random nodes... and I want to do it several times. I encountered very large computation time : about 4 min to perform 100 searches. Is it supposed to be slow for large graph? My machine is not very powerful but I think I've missed something...Is there ways to optimize? Can it be a problem from the weighted edges or from the heuristic distance ? my astar_search function call is : boost::astar_search(m_g, vertexStart, heuristicDistance, boost::predecessor_map(&p[0]) .distance_map(&d[0]) .visitor(astarVisitor) ); with m_g my graph which is a: typedef boost::adjacency_list<boost::listS, boost::vecS, boost::bidirectionalS, VertexProperties, boost::property<boost::edge_weight_t, float> > my_graph; heuristicDistance : same as the example (a simple euclidian distance) astarVisitor : same as the example any help is appreciated ! Thanks, Happy new year btw, YH

2 3

[Locale] as::ftime formatting
by Szymon Gatner 09 Jan '13

09 Jan '13

Hey, docs say: "There is a list of supported strftime flags by ICU backend:". I didn't build my boost/locale with ICU but still tried what happens when I try as::ftime("%A") an to my surprise I got correctly localized day of the week. My questions are: is it correct that it works without ICU? will it work on other machines (Win 32 bit)? What is the encoding of output string? I am seeing it correctly displayed as utf-8 but would like to know what is happening. Can I ensure utf-8 output? My test code: using namespace boost::locale; generator gen; locale loc=gen(""); std::ostringstream ostr; ostr.imbue(loc); ostr << as::datetime << as::ftime("%A") << info.time; std::String timeStr = ostr.str(); Cheers, Simon -- Szymon Gatner The Lordz Games Studio www.thelordzgamesstudio.com

2 1

boost serialization: can't deserialize pointer to abstract base class where derived classes overload load_construct_data
by Rebecca Lehman 09 Jan '13

09 Jan '13

I have some pointers to instances of an abstract base class. Some of the derived classes have no default constructors. I can serialize instances of these derived classes by overloading load_construct_data and save_construct_data, but I always get segmentation faults when I try to serialize the base class pointers. Saving works fine, and loading will correctly load the instances of the derived classes, but it segfaults before initializing the pointers. It doesn't seem to matter whether I overload the load_construct_data and save_construct_data of the base class, or how I register my classes and the base/derived relationship, or whether the pointers are ordinary pointers or boost::shared_ptrs. (The details of the segfault messages vary, but it always segfaults). Does anyone have an example of how to serialize such pointers? Here's a trivial example of what I'm trying to do: class Base { Base() {} public: virtual int Apply()=0 const; protected: friend class boost::serialization::access; template<class Archive> void serialize(Archive &ar, const unsigned int version){ } }; BOOST_ASSUME_ABSTRACT(Base); class D1 { public: D1( int der_input): der_int(der_input){} int Apply() const { return der_int; } protected: const int & der_int; friend class boost::serialization::access; template<class Archive> void serialize(Archive &ar, const unsigned int version){ ar & boost::serialization::base_object<Base>(*this); // or boost::serialization::void_cast_register<D1, Base>( // static_cast<D1 *>(NULL), // static_cast<Base *>(NULL) // ) } }; BOOST_CLASS_EXPORT_GUID(D1, "D1"); class D2 { public: D2( int der_input): der_int(der_input){} int Apply() const { return 2*der_int; } protected: const int der_int; friend class boost::serialization::access; template<class Archive> void serialize(Archive &ar, const unsigned int version){ ar & boost::serialization::base_object<base_class_of_T>(*this); // save/load class member variables } }; BOOST_CLASS_EXPORT_GUID(D1, "D1"); namespace boost { namespace serialization { template<class Archive> inline void save_construct_data( Archive & ar, const D1* t, const unsigned int file_version ){ // save data required to construct instance ar << t->der_int; } template<class Archive> inline void load_construct_data( Archive & ar, D1 * t, const unsigned int file_version ){ // retrieve data from archive required to construct new instance int attribute; ar >> attribute; // invoke inplace constructor to initialize instance of my_class ::new(t)D1(attribute); } template<class Archive> inline void save_construct_data( Archive & ar, const D2* t, const unsigned int file_version ){ // save data required to construct instance ar << t->der_int; } template<class Archive> inline void load_construct_data( Archive & ar, D2 * t, const unsigned int file_version ){ // retrieve data from archive required to construct new instance int attribute; ar >> attribute; // invoke inplace constructor to initialize instance of my_class ::new(t)D2(attribute); } }} main(){ ... ar.template register_type<D1>(); ar.template register_type<D2>(); std::vector<Base *> b; ... ar & b; }

2 2

Boost.MSM: Executable size
by RaRi 08 Jan '13

08 Jan '13

Hi all, I'm using boost.MSM for a productive embedded system. My concern is, that my compiler produces a executable size of all most 2MB for a state machine consisting of 4 sub-machine and nearly 40 transition for the complete state machine, but with no action and guard code. My compilation system is as follows: - eclipse juno - mingw 4.7.0 - Compiler settings for the release of the binary are: "-O3 -Wall -c -fmessage-length=0" Is this a good size or should I be worried about this nearly 2 MB for the release executable? I'm asking as like stated in http://www.boost.org/doc/libs/1_52_0/libs/msm/doc/HTML/ch04s02.html <http://www.boost.org/doc/libs/1_52_0/libs/msm/doc/HTML/ch04s02.html> the executable size for a good sized state machine could reach a size of almost 600kB, and my state machine is still not finished (50 % more code has to be added for further submachine...) Using the "-Os" produces a executable size which is almost 20% smaller as with the "-O3". In case I compile the state machine with the default debug level (-g) the binary is allmost 35MB big.... The compilation takes around 1m:30sec ( using the *favor_compile_time* policy)... Best regard, RaRi -- View this message in context: http://boost.2283326.n4.nabble.com/Boost-MSM-Executable-size-tp4641132.html Sent from the Boost - Users mailing list archive at Nabble.com.

2 1

Problem with installation
by Evan Driscoll 08 Jan '13

08 Jan '13

Hi, I'm having problems with the installation procedure. I'm using the 1.53 beta. I have run the following: ./bootstrap.sh --prefix=/unsup/boost-1.53.0/beta/common --exec-prefix=/unsup/boost-1.53.0/beta/amd64_rhel6 ./b2 --prefix=/unsup/boost-1.53.0/beta/common --exec-prefix=/unsup/boost-1.53.0/beta/amd64_rhel6 -j12 ./b2 --prefix=/unsup/boost-1.53.0/beta/common --exec-prefix=/unsup/boost-1.53.0/beta/amd64_rhel6 -j12 install however it is ignoring my --exec-prefix and installing everything to /unsup/boost-1.53.0/beta/common. I actually haven't really used --exec-prefix before I think, and have just manually moved stuff around post-installation instead. But I figured I'd try to do it "right" this time and actually use the tools I've been given. :-) What am I doing wrong? [Disclaimer: I made a typo originally and ran both bootstrap.sh and b2 with both --prefix and --exec-prefix set to ..../common. I did run './b2 clean' after that though.] Evan

1 0

How to use geometry::within with a polygon that encompasses a pole
by Cameron Sabol 08 Jan '13

08 Jan '13

Hi, I'm have an issue with polygons from the geometry library using the spherical_equatorial coordinates when the polygon is meant to wrap a pole. e.g. namespace bg = boost::geometry; typedef bg::model::point < double, 2, bg::cs::spherical_equatorial<bg::degree> > LonLatPoint; bg::model::polygon < LonLatPoint> northPoleBoundary; double lon = -180; while (lon <= 180) { bg::append(northPoleBoundary, LonLatPoint(lon, 88)); lon += 5; } LonLatPoint northPole(90, 0); bool isInNorthPole = bg::within(northPole, northPoleBoundary); This code is giving me isInNorthPole == false. Am I constructing my polygon incorrectly? Is this a bug? Or is this behavior simply not implemented? Thanks for your help!

2 2