- Boost - lists.boost.org

[serialization] "stream error" when port from boost_1.33 to boost_1.36
by ipangth 26 Sep '08

26 Sep '08

I just port a project using boost_1.33_1 to use boost_1.36. The project has a Server class using a std::set<Server> serialization, the serialization save and load functions code see behind. I use gcc 4.1 and run on FC6. The porting problem is, when the old boost_1.33 generates(save) a xml file, then use boost_1.36 load the xml throws an exception "stream error"! the exception error code line : boost::archive::xml_iarchive ia(ifs); How can i sovle it? class Server { public: int id; std::string ip; int port; private: static std::set<Server> lst_; static int dirty_; }; #endif #ifdef USE_SERVER_SERIALIZATION #ifndef _SERVER_SERIALIZATION_ #define _SERVER_SERIALIZATION_ BOOST_CLASS_VERSION(Server, 1) namespace boost { namespace serialization { template <class archive> void serialize(archive & ar, Server & t, const unsigned int ver) { ar & BOOST_SERIALIZATION_NVP(t.id); ar & BOOST_SERIALIZATION_NVP(t.ip); ar & BOOST_SERIALIZATION_NVP(t.port); } } } #endif #endif int Server::save( const std::string & file) { try { std::ofstream ofs(file.c_str(), std::ios::binary); { boost::archive::xml_oarchive oa(ofs); oa << BOOST_SERIALIZATION_NVP(lst_); } } catch(...) { return ERROR_ERROR; } return ERROR_SUCCEED; } int Server::load(const std::string & file) { std::set<Server> Server::lst_; try { std::ifstream ifs(file.c_str(), std::ios::binary); { boost::archive::xml_iarchive ia(ifs); ia >> BOOST_SERIALIZATION_NVP(lst_); } } catch(std::exception & e) { std::cout << e.what() << std::endl; } return 0; } -- Best regards, PangYongQiang

2 5

[boos][switch] What do you think?
by Kasra Nassiri(Math & ComSci) 26 Sep '08

26 Sep '08

Hi all, I have been writing a BOOST_SWITCH(...) statement, which tries to emulate the C++ switch() for C++ objects. string variable = "some string"; string string0 = "0", string1 = "1", string2 = "2"; BOOST_SWITCH( variable , BOOST_CASE( string0, std::cout << "error" << std::endl; BOOST_BREAK; ); BOOST_CASE( string1, std::cout << "error" << std::endl; BOOST_BREAK; ); BOOST_DEFAULT( BOOST_SWITCH( string2, BOOST_DEFAULT( std::cout << "embedded switch statments work" << std::endl; ); ); std::cout << "end of secound switch statment" << std::endl; std::cout << "reached default" << std::endl; ); ); The above code would work correctly (msvc). However, for "const char *", the compiler could not deduce template. I have written an specialized "const char *" template however, still could not deduce template. Any suggestions? template<class T> switch_; template<> switch_<const char *>; With best regards Kasra Nassiri

1 0

Re: [boost] Phoenix review
by Alp Mestan 26 Sep '08

26 Sep '08

I really want to give my approval, for what it is worth... I vote to accept Phoenix into Boost. However, I think it is important to keep Boost.Bind as many of Boost users don't know (yet) Phoenix but intensively use Bind, that's to say they only use Bind because they just need it, no more. The documentation is excellent - I had to use it for some personal projects - and I got used to Phoenix very quickly. The syntax is quite simple compared to the work done in background by the library. Thank to Phoenix, we can now write FP code in C++ quickly, efficiently. Thanks guys.

2 1

[graph] MAXINT problem
by Janek Kozicki 25 Sep '08

25 Sep '08

Hi, I just noticed that in file http://svn.boost.org/svn/boost/trunk/boost/graph/sloan_ordering.hpp there is this line: #define MAXINT 2147483647 I'm afraid that it can conflict with MAXINT defined elsewhere. Or at least cause some confusion between MAXINT and INT_MAX from <limits.h> Wouldn't it be better to just #include <limits.h> and use INT_MAX defined there? -- Janek Kozicki |

3 3

boost::signal poor performance
by Hansi 25 Sep '08

25 Sep '08

Hello, I have posted this already to the user list, maybe wrong forum... I have made now a small test and I don't know if I make something wrong. I have the following performance comparsion between a boost::function and a boost::signal. Okay, I know that boost::signal has a lot of more functionality and for that it is for sure slower than boost::function. But 100times slower? May be I do something wrong? On my pc I have for boost::function a runtime of 728ms and for boost::signals a runtime of 64128ms There is one slot used for signals. Can anyone give me some hint what I do wrong? my test program was: class TestBind { public: int counter; void OnEventReceived(char b) { counter++; } }; class TestSignal { public: int counter; void OnEventReceived(char b) { counter++; } }; int _tmain(int argc, _TCHAR* argv[]) { TestBind testBind; PerformanceCounter timer; boost::function<void (char)> f = boost::bind(&TestBind::OnEventReceived,&testBind,_1); for(int i = 0; i < 100000000; i++) { f('a'); } TestSignal testSignal; boost::signal<void (char)> sig; sig.connect(boost::bind(&TestSignal::OnEventReceived,&testSignal,_1));; for(int i = 0; i < 100000000; i++) { sig('a'); } } Best regards Hansjörg

6 8

function_traits and functors
by Michael Marcin 24 Sep '08

24 Sep '08

Is it possible to use function_traits to decompose functors? Specifically is this supposed to work? void test_helper( int a ) { } template< class Handler > void test( Handler h ) { typedef typename boost::function_traits<Handler>::arg1_type ArgType; } int main() { test( boost::bind(&test_helper,_1) ); } Thanks, -- Michael Marcin

2 2

[boos][threadalert] What about a submision to boost?
by vicente.botet 24 Sep '08

24 Sep '08

Hello Ronald, I need a transactional thread specific context which can also be accesible from other threads during consistency checking. I've found in your library threadalert (available on the boost vault) the thread_member_ptr which corresponds completely to my needs. I'think that independently of the alert library the thread_member_ptr might be a good candidate to Boost. thread_specific_ptr behaves as a private member of a thread and thread_member_ptr as a public member. This seems a natural extension. BTW, it would be nice to be able to associate a cleanup function as it is the case of the thread_specific_ptr. If you need some help I would like to collaborate with you. Please, let me know what do you think. Best regards, Vicente Juan Botet Escribá

1 0

feature request for building boost libraries for msvc
by Hansi 24 Sep '08

24 Sep '08

Hello, I have a question to building boost. I have read now a lot of posts where it is written that with msvc the SECURE option is really slow. In some posts I have read that this option breaks the ABI and for that it is needed to build also the boost library with the right compiler switch. isn't it? 1) How I can build it with a custom compiler switch? sorry if it is documented, but I don't have found it... 2) Is it possible to add a make option(complete build) which builds a library with this switch and against which boost links automatically when the application defines _SECURE_SCL=0? Best regards Hansjörg

1 0

boost::atomic<T> foo;
by Janek Kozicki 24 Sep '08

24 Sep '08

Hi, I looked through Boost.Threads and Boost.Interprocess libraries, but (as far as I can tell) none of them offers me simple: class Foo { boost::atomic<int> bar; } which would mean that *ALL* attempts to access bar would be atomic, eg. bar=10; Am I left with troublesome usage of boost::mutex to block all that access attempts by myself? -- Janek Kozicki |

6 8

[Filesystem] has_root_path()
by Robert Jones 24 Sep '08

24 Sep '08

Boost 1.33.1 used to support has_root_path() on directory_iterators. 1.36 seems to use a slightly different mechanism for directory_iterators, based on instantiating basic_directory_iterator, and I'm struggling to see what has happened to has_root_path(). It's there somewhere, but not accessible in quite the same way as it was. How do I express directory_iterator -> has_root_path() in 1.36.0 ? Thanks - Rob.

1 0