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

[BOOST THREAD] Threads Spawning Unexpectedly
by Terrimane Pritchett 28 Apr '09

28 Apr '09

Hello, I have taken interest in using Boost Threads in my project. I have a strange issue that I am trying to debug. I spawn N boost::thread instances because I know exactly how many I need. My program crashes after a time on a boost::bad_lexical_cast exception. I have placed a wrapper around boost::lexical_cast which uses a boost::recursive_mutex and boost::lock_guard combination to prevent multiple threads from calling boost::lexical_cast at the same time. Still, my project invariably chokes on boost::bad_lexical_cast being thrown when I can verify the cast should work i.e. everything works as expected in my single threaded implementation. My real cause for alarm is this....when I act to debug my project I always discover there are more active boost::thread instances than I have explicitly created. I only create boost::thread instances in one location. I have perused the documentation but as of yet cannot find an explanation for this behavior. My assumption is that unless I explicitly instantiate a boost::thread it should not exist. However, I want to ask if this is indeed the case? If not, how so and what is the rationale behind this behavior? Here is what my code more or less looks like - this is all I do with boost::threads currently - how are more than N boost::threads spawning where N is the number of elements in an arbitrary stl container which itself is not allowed to change in size? typedef boost::shared_ptr<boost::thread> THREADHANDLE; typedef std::vector<THREADHANDLE> THREADHANDLECONTAINER; THREADHANDLECONTAINER workerThreads; workerThreads.reserve(some_container.size()); ... workerThreads.push_back( THREADHANDLE(new THREADHANDLE::element_type( Callable, arg1, boost::ref(arg2)); ... BOOST_FOREACH(const THREADHANDLE &workerThread,workerThreads) { workerThread->join(); }//end loop ------------------------------------------ Here is my attempt at using boost::lexical_cast in a generic thread safe manner. I am expecting exactly one thread my call boost::lexical_cast at a time with this wrapper yet my application blows up solely on boost::bad_lexical_cast whenever I introduce Boost Threads. Am I missing the elephant in the room somehow? //LexicalCaster.h #include <boost/thread.hpp> struct LexicalCaster { template<typename TargetType,typename SourceType> inline static TargetType LexicalCast(const SourceType &source); private: static boost::recursive_mutex& GetMutex(); static int& GetCount(); }; //LexicalCaster.cpp #include "LexicalCast.h" boost::recursive_mutex _mtx; int count(0); //here for debugging purposes boost::recursive_mutex& LexicalCaster::GetMutex() { return _mtx; } int& LexicalCaster::GetCount() { return count; } //LexicalCaster.inl #include <boost/lexical_cast.hpp> template<typename TargetType,typename SourceType> TargetType LexicalCaster::LexicalCast(const SourceType &source) { boost::lock_guard<boost::recursive_mutex> guardian(LexicalCaster::GetMutex()); std::cerr<<"Count: "<<LexicalCaster::GetCount()++<<std::endl; //the count will display sequentially if only //one thread of execution has access at a time return boost::lexical_cast<TargetType>(source);

4 12

Re: [Boost-users] Serialization problem with DLL & EXE
by Andrei Popa 28 Apr '09

28 Apr '09

Hi Robert, With a class derived from DbObject and defined/implemented in my EXE (that use the DLL) I got archive_exception::unregistered_class. I tried to register it or to export it, but I could not make it work. *** So you put BOOST_CLASS_EXPORT(DbObject) in to a *.cpp file which is compiled as part of the exe? Yes, I do. Any Ideas? I use Version 1.38 of the library & VC8(2005). When I use BOOST_CLASS_TYPE_INFO(as in test_no_rtti.cpp sample ) for my EXE resident class, the ::get_key is never called. // get the eti record for the exported type "polymorphic_derived2" boost::serialization::extended_type_info const * const d2_eti = boost::serialization::extended_type_info::find( "polymorphic_derived2" ); works too before I start to serialize the collection. *** is this an attempt the issure raised above or is this a separate question. No separate question at all, this is the way I tried figure out if my class is really (and correctly) registered. Is this a correct way to test? How can I debug what's wrong? Thanks Andrei. _________________________________________________________________ News, entertainment and everything you care about at Live.com. Get it now! http://www.live.com/getstarted.aspx

2 1

Serialization problem with DLL & EXE
by Andrei Popa 28 Apr '09

28 Apr '09

Hi there, I have a dll that contain an abstract base class called DbObject and a DbDatabase class that contains a collection of shared_ptr<DbObject> . It dose work as long all derived classes are defined in the same dll. With a class derived from DbObject and defined/implemented in my EXE (that use the DLL) I got archive_exception::unregistered_class. I tried to register it or to export it, but I could not make it work. Any Ideas? I use Version 1.38 of the library & VC8(2005). When I use BOOST_CLASS_TYPE_INFO(as in test_no_rtti.cpp sample ) for my EXE resident class, the ::get_key is never called. // get the eti record for the exported type "polymorphic_derived2" boost::serialization::extended_type_info const * const d2_eti = boost::serialization::extended_type_info::find( "polymorphic_derived2" ); works too before I start to serialize the collection. Thanks, Andrei _________________________________________________________________ Show them the way! Add maps and directions to your party invites. http://www.microsoft.com/windows/windowslive/products/events.aspx

2 1

po::program_options
by Zarkvon 28 Apr '09

28 Apr '09

Hello, During using library 'program_options' we noticed some issues. The first one is that class 'duplicate_option_error' never uses. /** Class thrown when duplicate option description is found. */ class BOOST_PROGRAM_OPTIONS_DECL duplicate_option_error : public error { public: duplicate_option_error(const std::string& what) : error(what) {} }; And the second one which is critical for us. Is that all errors are forming as one sentence in ENGLISH. And it's an issue to get which exception exactly occured to form your own message in your desired language. For instance class - validation_error. I guess it's really hard and inconvinient to get what message should be thrown. So, what will you advise to do to make this library more "international", so all exception messages are put out in any language which programmer wants.

1 0

Boost Graph Library mysterious compile errors
by Lindley M French 28 Apr '09

28 Apr '09

Follow-up: It appears to be related to the edge_weight property. Judging by the documentation I was under the impression that this was a default property that all edges would have. Is this not the case? How would I add it? I already tried putting a "double edge_weight" field in my EdgeProperties struct, didn't help.

2 1

[graph] ColorMap
by liquid 28 Apr '09

28 Apr '09

Please, help me to declare correctly the ColorMap as the 3-rd parameter of depth_first_search function. Documentation does not explain it clearly. Regards liquid

2 1

Boost & Continuous Integration
by Robert Dailey 28 Apr '09

28 Apr '09

Hey guys, One of the requirements for a continuous integration process is that the build for a specific target be quick. Something under 10 minutes is ideal. I find that when not using precompiled headers, boost can significantly impact the compilation time of the project. I've experienced 20 minute compiles when using boost fairly extensively throughout the code base. So far I only see two obvious solutions to this problem. First, use precompiled headers. I really don't want to do this because it causes issues with include dependencies and makes the code not reusable as a result. The second option is to beef up the machines doing the compiles, however this can have diminishing returns. I am sure that a lot of people in the community have had this specific issue with Boost's compile times, as well as in many other areas. What would you guys recommend? Is there any other solution beyond the obvious? Thanks for your time.

9 12

Boost Graph Library mysterious compile errors
by Lindley M French 28 Apr '09

28 Apr '09

I'm getting some mysterious compile errors relating to boost::detail::error_property_not_found. Truely massive errors, well too large to get any useful information out of.....ah, templates. I eventually figured out that these sorts of errors are usually related to BGL algorithm calls which can't find the properties they need. (Wow, thanks for pointing me right at the source, template error...) Well, the only algorithms I'm calling right now are connected_component and kruskal_minimum_spanning_tree. One such instance occurs with a particular connected_component call: typedef boost::adjacency_list< boost::listS, boost::vecS, boost::undirectedS, RegionProperties, EdgeProperties> Graph; template <typename Graph> struct edge_on { edge_on(): m_G(NULL) { } edge_on(const Graph &G) : m_G(&G) { } template <typename Edge> bool operator()(const Edge& e) const { return (*m_G)[e].on; } const Graph *m_G; }; ... //function arg ....(Graph &G,... filtered_graph<Graph, edge_on<Graph> > G_on_edges(G, edge_on<Graph>(G)); std::vector<int> component(num_vertices(G_on_edges)); int num_comps = connected_components(G_on_edges, &component[0]); Do you see anything wrong there? The kruskel call (which is producing more such errors) is Graph comp_G; .... vector<Edge> mstedges; mstedges.reserve(num_vertices(comp_G)); kruskal_minimum_spanning_tree(comp_G,std::back_inserter(mstedges)); I really have no idea what might be wrong here.

1 0

filtered_graph question
by Lindley M French 28 Apr '09

28 Apr '09

If I construct a filtered_graph on some underlying graph, and then (either through it or via another interface) modify the underlying graph to change which edges make it through the filter....when is this change reflected? Do I have to construct another filtered_graph? Or is it immediate? Will it invalidate iterators or anything like that? I'm just trying to get my head around the conceptual workings here.

1 0

Use of Boost [threadpool] library w/ Windows Service Process
by Art Beall 27 Apr '09

27 Apr '09

Hello, Our application was in need of a thread pool for processing socket messages. During implementation and testing we were using the [threadpool] library along with Boost v36. We are using Visual Studio 2003 for building our application. Our target systems are running Windows Server 2003 or Windows XP Pro. The windows console version of our application runs great. No problems with thread pooling. However, once we use it inside a windows service, the application fails to load. No code is actually executed. After doing a lot of research into the problem, I narrowed it down to the inclusion of the threadpool object into the code. One caveat with using the application as a windows service, the main project is compiled /clr (managed code). However, I've placed all code that refers to boost in a separate library compiled as unmanaged code. This library is linked with both the console application (working), and the service application (won't load). Does anyone have any ideas what would cause this ? Thanks! Art Beall

1 0