December 2009 - Boost-users - lists.preview.boost.org

Function template specialization
by Surya Kiran Gullapalli 18 Dec '09

18 Dec '09

Hello all, The following question may not be related to boost, but if a solution can be found with boost or otherwise would be great. I've a base class with a function template. I derive from base class and try to have a specialization for the function template in derived class I did something like this. ////////////////////////// Header file begin ////////////////////////// class Base { .. template <typename T> fun (T arg) { ... } }; class Derived : public Base { ... } ; template <> Derived::fun(int arg); ///////////////////////// Header file end //////////////////////////// and in .cpp file I've provided implementation for the template specialization. This works fine with MSVC 8.0 and g++ complains about lack of function declaration fun in Derived class. I do not know which compiler is behaving correctly. Any help in this is greatly appreciated. Thanks in advance, Surya

2 1

boost lambda and bind
by Conoscenza Silente 18 Dec '09

18 Dec '09

Hi All I am trying to use boost lambda for looping over a std::map to get each element and pass those to a class function using lambda functions. I don't know why but the binding inside my std::for_each is not compiling and I have 2 errors pointing to the line where boost::assign is used. 'invalid initialization of reference of type int&' 'invalid initialization of reference of type std::basic_string<>&' Do you know how to fix this and why it is happening Thank you very much Regards CS //Boost //#include<boost/assign/std/map.hpp> #include<boost/assign/list_of.hpp> #include<boost/lambda/lambda.hpp> #include<boost/lambda/core.hpp> #include<boost/lambda/bind.hpp> #include<boost/lambda/construct.hpp> class CInnerClassHandler{ public: void DoSomething( int aIntValue, const std::string& aStringValue ){ std::cout << "INT:" << aIntValue << "STRING:" << aStringValue << std::endl; // here we just trace } }; void map_lambda_test(){ typedef std::pair< int, std::string > this_pair; std::map< int, std::string > dictionary = boost::assign::map_list_of(3,"A")(4,"B")(5,"C"); // Errors points to // this line of code CInnerClassHandler obj; std::for_each( dictionary.begin(), dictionary.end(), boost::lambda::bind( &CInnerClassHandler::DoSomething, &obj, boost::lambda::bind(&this_pair::first, boost::lambda::_1), boost::lambda::bind(&this_pair::second,boost::lambda::_1)) ); }

2 2

Build boost for i686 (instead of i386)
by Christoph Duelli 18 Dec '09

18 Dec '09

What do I have to pass to bjam to build Boost with i686 instead of the default i386 as cpu? (I am using Linux, with gcc 4.3.2) So far I am calling ./bjam architecture=x86 address-model=32 Best regards Christoph

2 1

Re: [Boost-users] [Boost.Test] valgrind complains about invalid reads
by Scott Banachowski 18 Dec '09

18 Dec '09

Thanks for the reply Gennadiy, see inline for comments: --- On Wed, 12/16/09, Gennadiy Rozental <rogeeff(a)gmail.com> wrote: > From: Gennadiy Rozental <rogeeff(a)gmail.com> > Subject: Re: [Boost-users] [Boost.Test] valgrind complains about invalid reads > To: boost-users(a)lists.boost.org > Date: Wednesday, December 16, 2009, 9:32 PM > Scott Banachowski <sbanacho > <at> yahoo.com> writes: > > > > > > > I agree with Peter here, the destructor for test_unit > should be virtual. > > 'Should' is a strong statement. Do you have anything to > based it on? > Ok, should is too strong. Change that to "it would be nice to", because the possibility that the base class destructor may be called (even if not by your code, possibly by others). > > I find that it is being deleted when > > test_unit* is a stored in reference counted pointer, > > Where? My mistake, it's held in a map. See the next comment: > > > and at program exit it is being deleted by calling > ~test_unit(). > > Where do I call destructor explicitly? You don't. It's being called implicitly at program exit, by valgrind. I think what is happening is that valgrind replaces the delete operator with its own, which is "operator delete(void*)" The typeinfo is likely lost by this replacement of delete. Just because you don't call the destructor, it doesn't mean others won't. If a base class is derived from, best practice is to declare its destructor virtual. It is relatively common to replace allocators (e.g. tcmalloc). > > > If you use a derived class (test_suite), this is > leading to an error. > > Where? As already mentioned, reported by valgrind. > > > Also, in the code I'm looking at (1.40.0), the > test_suite class has a > > virtual destructor, but its base class test_unit > does not. > > Not making the base class virtual seems to be > defeating the purpose. > > What purpose? The only reason that test_suite destructor is > virtual is because > this class in fact is being polymorphically. It has > subclass master_test_suite. > By defeating the purpose, I mean it forces code like the following: if( ut_detail::test_id_2_unit_type( tu.second->p_id ) == tut_suite ) delete static_cast<test_suite const*>(tu.second); else delete static_cast<test_case const*>(tu.second); A virtual base case simplifies this code (especially if in the future other classes are added, or in the case of replacement allocator it does the right thing). thanks, Scott > Gennadiy > > _______________________________________________ > Boost-users mailing list > Boost-users(a)lists.boost.org > http://lists.boost.org/mailman/listinfo.cgi/boost-users > __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com

4 3

[serialization] object tracking for shared strings
by Nils Krumnack 17 Dec '09

17 Dec '09

Hi all, I hope one of you will be able to help me with this: I am working with a data structure that contains a large number of (mostly identical) strings. To save memory I put together a shared memory string implementation, which ensures that each string is kept in memory only once (implemented through a global map with all strings). However, once I write them to disk, I again have n copies of each string. I would really prefer, if I could avoid the space overhead. At first glance, object tracking seems to be the way to go. If my strings were tracked, there would automatically be only one copy of each string. However I can't quite figure out how. I don't really want to change the global tracking behavior for std::string, since that might end up affecting code all over the place. Serializing pointers would be an option, but then I would have to deal with the issue of when to deallocate that object. I suppose I could serialize shared pointers, but then I would have to change my implementation to use shared pointers internally. Alternatively I could also wrap a struct around the strings internally and turn on tracking for that struct, but again I would have to complicate my implementation for the sake of serialization. What I was hoping to find was some way of telling the archive at the point where I am serializing the object, that it should track this one object. I am thinking of something like ar & BOOST_SERIALIZATION_TRACK_THIS (mystring); I was looking for something like that in the documentation, but wasn't very successful. Any ideas? Thanks in advance for any help, Nils

3 7

are boost::interprocess containers thread-safe?
by blp330 17 Dec '09

17 Dec '09

Hi, I must implement a multiple producer-consumer IPC application in win32 (but will port to linux in near future). I am wondering whether the boost::interprocess container like list, vector or map is thread-safe (actually, interprocess-safe?) or I need to use lock in my producer-consumer application? I mean, can Producer just simply put shared memory string to container and take it from Consumer without any lock? Or is there any multiple producer-consumer example that uses boost::interprocess? Thanks a lot. -- View this message in context: http://old.nabble.com/are-boost%3A%3Ainterprocess-containers-thread-safe--t… Sent from the Boost - Users mailing list archive at Nabble.com.

2 3

[proto] extends example not compiling
by alfC 17 Dec '09

17 Dec '09

Is this example, contained in the boost.Proto manual supposed to compile? #include<complex> #include <boost/proto/core.hpp> #include <boost/proto/context.hpp> #include <boost/proto/debug.hpp> #include<boost/proto/extends.hpp> template<class T, class Dummy = boost::proto::is_proto_expr> struct my_complex { BOOST_PROTO_EXTENDS( typename boost::proto::terminal<std::complex<T> >::type , my_complex<T> , boost::proto::default_domain ) }; int main() { my_complex<int> c0, c1; c0 + c1; // OK, operator+ found now! } I get this error (with gcc 4.3 and gcc 4.4): proto_complex.cpp:10: error: expected nested-name-specifier before ‘typename’ proto_complex.cpp:10: error: two or more data types in declaration of ‘proto_base_expr’ proto_complex.cpp:10: error: ‘proto_base_expr’ has not been declared proto_complex.cpp:10: error: ‘proto_base_expr’ has not been declared The strangest thing is that when I replace the macro by its explicit code the code seems to compile. Thank you, Alfredo

2 5

[Re: ] boost.plot: integrating plotting in a C++ native application: general questions about which tools to use
by Richard Ulrich 17 Dec '09

17 Dec '09

Hi Hicham, another possibility is to use a webapp like Wt from http://webtoolkit.eu It's great to work with and has good support for 2D plots. As for the 3D: I've been thinking about that as well for a whild especially embedded in an interactive webpage. I came across openVRML which looks promising, but is a hell of a beast to build (the supporting libraries). So no experiences yet. Rgds Richard

3 2

newbie - graph library - vertex properties
by List User 17 Dec '09

17 Dec '09

Hi 1. In the creation of adjacency list, vertex properties and edge properties could be vecS, listS etc.But how will one access this vertex_properties vector, for example? 2. In the kevin-bacon2.cpp example file available with boost distribution, a struct is used to store vertex properties. But that example does not seem to have used boost_install_property or property_map etc. I thought one would need to use them for having new properties for the vertex. 3. In the kevin-bacon2.cpp approach for adding new properties to the graph, the vertex_properties are read as adjacency_list[vertex_descriptior]. How will have a read only access in such a case? When I use this in a function which is defined as const in c++, the compiler cribs :)Thanks suresh

5 15

[asio] Socket closed notification?
by Nicola Michael Gutberlet 17 Dec '09

17 Dec '09

Hello everybody, I'm using TCP sockets and wonder, if there's a possibility to get an automatic notification when one side of a socket is closed (even if the other side is crashed / not closed properly). Thanks in advance for any help. Regards, Nicola

4 14