July 2011 - Boost-users - lists.preview.boost.org

Metaprogamming question
by John M. Dlugosz 09 Jul '11

09 Jul '11

I roughed out a function that will be a template that dispatches to one of several different forms based on the nature of one of the arguments. It's fairly simple, each helper call asking one question to further divide it up. But, I want to write a dozen functions that all have similar (if not identical) issues. So I want to abstract out the mixmaster and code that once. I'm thinking that I map the argument to a "tag" that has many possible values, and have one set of overloaded functions. It is easy to write the back end, as each function's tag is clear what that case is for. Furthermore, I can make the tags empty structs that use an inheritance pattern, so any forms I don't supply for some function will simply pick the more general one. So, how do I write the metafunction that maps the type to the tag? I could follow my same approach of one step at a time. But I wonder if a "switch" statement of some kind would be clearer and easier. OTOH, I don't want to learn all of MPL just for one thing, and I don't want to bog down the compilation speed. Is there something specific someone can point me to? Second, my pseudocode has a metafunction for telling me whether a type is a resizable container. STL containers predate the modern style of metafunctions and use cumbersome traits instead, but is there something around that I can use to tell that something is supposed to be a STL container or specific categories of containers? I could just list them all one by one, but type_traits already did that for me for a lot of things, and I'd have to tell everyone how to meta-annotate their own classes that are supposed to work "like" STL containers for algorithms. Can I use something like enable_if to tell if a class has a push_back or resize function? Thanks everyone, --John

5 4

Boost.Operators on a POD class
by Krzysztof Czainski 08 Jul '11

08 Jul '11

Hello, A simplified example of what I am trying to accomplish follows: // needs to be a POD struct Point { double x, y; // some functions... }; Now I would like Point to be additive, and multiplicative by double, so normally I'd write: struct Point : additive1< Point, multiplicative2< Point, double > > { // supply the needed operators, and let Boost.Operators do the rest }; But Point needs to be POD, and inheritance kills POD-ness. So I'm wondering, can Boost help me here? Regards, Kris

1 0

Boost in X-Code 4
by Mark Schisler 08 Jul '11

08 Jul '11

When compiling for the iPhone simulator with the following command: ./bjam --prefix=/Downloads/boost_install/ toolset=darwin architecture=x86 target-os=iphone macosx-version=iphone-${SDK_VERSION} --with-date_time --with-regex --with-thread link=static install And the following user-config.jam: using darwin : 4.2.1~iphone : /Developer/Platforms/iPhoneOS.platform/Developer/usr/bin/llvm-g++-4.2 : <striper> <compileflags>"-arch armv7" <compileflags>-mthumb <compileflags>-D_LITTLE_ENDIAN : <architecture>arm <target-os>iphone ; using darwin : 4.2.1~iphonesim : /Developer/Platforms/iPhoneSimulator.platform/Developer/usr/bin/llvm-g++-4.2 : <striper> <compileflags>"-arch i386" <compileflags>-D_LITTLE_ENDIAN : <architecture>x86 <target-os>iphone ; using darwin : 4.2.1~darwin : /Developer/usr/bin/gcc : <striper> <compileflags>"-arch i386" <compileflags>"-arch x86_64" : <architecture>x86 <target-os>darwin I receive the following linker error when I attempt to link against the reg ex library: ld: bad codegen, pointer diff in boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<std::runtime_error> >::rethrow() constto global weak symbol vtable for boost::exceptionfor architecture i386 collect2: ld returned 1 exit status Command /Developer/Platforms/iPhoneSimulator.platform/Developer/usr/bin/g++-4.2 failed with exit code 1 I've noticed that many people have managed to get around this error by hiding boost's symbols, however, that doesn't seem to be the preferred approach as it requires hiding all of my projects symbols, which will prevent me from being able to handle exceptions according to Apple docs: > If your symbol uses runtime type identification (RTTI) information, > exceptions, or dynamic casts for an object that is defined in another > library, your symbol must be visible if it expects to handle requests > initiated by the other library. For example, if you define a catch > handler for a type in the C++ standard library, and you want to catch > exceptions of that type thrown by the C++ standard library, you must > make sure that your typeinfo object is visible. Is there another approach to fixing this linker error? Thanks, Mark

1 1

stop a socket
by Gianni Ambrosio 08 Jul '11

08 Jul '11

Hi All, I would like you to explain me a detail of io_service. In my case I attach a socket to the io_service but if the socket is still open the io_service.stop() does not work as expected. Is there a way to close the socket just through the related io_service? Regards Gianni

2 1

[pool] Memory usage with boost::pool
by Daniele Barzotti 08 Jul '11

08 Jul '11

Hi, in my project I try to use boost::pool to handle memory allocation/deallocation. I have a MemoryManager object with two simple methods: void* WaveStreamMM::AllocateBuffer(size_t numbytes, int memtype) { boost::mutex::scoped_lock lock(m_mutex); void *buf = m_pool.ordered_malloc(numbytes); memset(buf, 0, numbytes); return buf; } void WaveStreamMM::FreeBuffer(void *p) { boost::mutex::scoped_lock lock(m_mutex); m_pool.free(p); } If I use the code above, the memory usage of my application increase over the time, while if I substitute the allocation/deallocation respectively with: void *buf = malloc(numbytes); and free(p); the memory usage remains constant. I wrong something? Regards, Daniele.

1 0

boost::asio::tcp -> asyncconnect generates multiple sockets
by Hert 08 Jul '11

08 Jul '11

hello, I noticed that when asynchronously connecting to an acceptor on localhost using boost::asio::ip::tcp::socket generates 2 TCP sockets on the system. One to the acceptor and one to the other one. Stepping into the code I can see where it is done, but I don't understand why. So why is this extra TCP socket there? thanks in advance, Gert -- View this message in context: http://boost.2283326.n4.nabble.com/boost-asio-tcp-asyncconnect-generates-mu… Sent from the Boost - Users mailing list archive at Nabble.com.

2 2

[thread] set thread affinity
by James C. Sutherland 08 Jul '11

08 Jul '11

Is it possible to set thread affinity through boost::thread? I suspect that I am losing performance in my application because the OS is migrating threads and causing cache thrashing... James

2 1

ct_if.hpp
by Simon Knapp 08 Jul '11

08 Jul '11

Hi List, I was looking at the file boost/pending/detail/property.hpp and noted that boost::ct_if was used therein. I think this must have come from boost/pending/ct_if.hpp (but that header is not included?), which no longer exists in boost version 1.46.1. This is the only reference to boost::ct_if I can find within the boost distribution. I was wondering if this is just old code that is not used anywhere (any more), or if I am missing something?

2 2

[Graph] Assertion in two_bit_color_map
by Maxime van Noppen 07 Jul '11

07 Jul '11

Hi, I'm getting an assertion failure that I don't understand. /usr/include/boost/graph/two_bit_color_map.hpp:86: void boost::put(const boost::two_bit_color_map<IndexMap>&, typename boost::property_traits<PropMap>::key_type, boost::two_bit_color_type) [with IndexMap = boost::vec_adj_list_vertex_id_map<boost::property<boost::vertex_bundle_t, trade_graph::vertex, boost::no_property>, long unsigned int>, typename boost::property_traits<PropMap>::key_type = long unsigned int]: Assertion `(std::size_t)i < pm.n' failed. It's triggered by a call to dijkstra. I've attached the relevant part of the backtrace to this mail. There's a lot of code involved and can't reproduce a minimal test-case. Here's the call to dijkstra: cost_map_t cmap(boost::num_vertices(cg)); predecessor_map_t pmap(boost::num_vertices(cg)); boost::dijkstra_shortest_paths(cg, src_rp->position(), boost::predecessor_map(&pmap[0]). distance_map(&cmap[0]).weight_map( make_function_property_map<edge_t> (boost::bind(&transaction::edge_cost_, this, _1, boost::ref(g), quantity, src, dst)))); Any idea greatly appreciated, thanks! -- Maxime

1 1

Storing vertex descriptors when removing a vertex
by Orianne Siret 07 Jul '11

07 Jul '11

Hi all, I am using the boost graph library for my project and encountered a small problem I don't know how to solve. I had to keep in an extern storage some of the vertex_descriptor in order to have access to their adjacent vertices and their properties. But as it is written in the boost documentation, when I remove a vertex, the vertex_descriptors change and I am not able to keep the right ones in my storage. Is there something else apart from vertex_descriptor that won't change but from which I could also access adjacent vertices and properties ? Thank you in advance for your help. Orianne

2 1