April 2006 - Boost-users - lists.preview.boost.org

Serialization of adjacency_list
by Anatoli Sakhnik 29 Apr '06

29 Apr '06

Hello! I slightely modified kevin-bacon2.cpp from examples of the BGL. Having tried to serialize the graph g back to the text and print it to stdout, I got a static assertion failure. Is serialization of the adjacency_list fully implemented indeed? Or is it a trouble with boost::serialization? -- Anatoli Sakhnik P.S.: diff kevin-bacon2.old.cpp kevin-bacon2.cpp 18a19 > #include <boost/archive/text_oarchive.hpp> 88a90,91 > archive::text_oarchive oa(std::cout); > oa << g;

3 4

Serialization difficulties
by Joseph Hosteny 29 Apr '06

29 Apr '06

Hi all, I'm a relative newbie at using the serialization libraries, so please go easy on me, but I'm having some troubles which I can't seem to work out. I have a class hierarchy that looks something like this: A -> B -> C -> D There is a corresponding header, and at least one source file, for each of the four classes. The base class, A, contains only methods, but is not a pure virtual class. Class B contains the first set of data members in the chain that must be serialized. Classes C and D contain some more methods and data. All stores/loads will be handled via pointers to instances of class D. Now, the header file for class B contains the following: #include <boost/archive/text_iarchive.hpp> #include <boost/archive/text_oarchive.hpp> #include <boost/serialization/access.hpp> #include <boost/serialization/base_object.hpp> Classes B, C and D define the following as public: friend class boost::serialization::access; template<class Archive> void serialize(Archive &ar, const unsigned int version); The serialize methods for classes C and D call the base class implementations of these methods via: ar & boost::serialization::base_object<base_class>(*this); Finally, the source files for classes B, C and D contain: #include <boost/serialization/export.hpp> BOOST_CLASS_EXPORT(class_name); It's not clear to me, though, that I should need the EXPORT macro since all loads/stores are done with the most derived class' pointers. However, when I run this code, I get an exception at oserialization.hpp line 394, which is of type archive_exception::unregistered type. I've looked through a lot of the documentation, but can't seem to find out what I'm doing wrong. I've also tried explicit calls to boost::serialization::void_cast_register<base_class, derived_class> (), but that didn't seem to work either. Since this is being retrofitted to a large design, I tried to distill what I was trying to do into a sample project with just empty classes with the same hierarchy and boot methods/macros. When I do this, everything seems to work fine. Does anyone have any suggestions on what I should look for that may be screwing things up for me? I can send the sample files I used, but since these work, I'm not sure if they will be of much help. Any pointers will be greatly appreciated! Thanks, Joe

2 1

[serialization] Empty save_override for tracking_type, object_id_type etc
by Denis Kuznetsov 29 Apr '06

29 Apr '06

Hi, I want to write my own archive class for reading/writing to some defined binary protocol. I'm deriving my archive class from the binary_oarchive_impl template and overriding save_override functions. As protocol is defined, I don't need any additional info(bytes). So I do smth like this: class fast_binary_oarchive : // don't derive from binary_oarchive !!! public binary_oarchive_impl<fast_binary_oarchive> { typedef fast_binary_oarchive derived_t; #ifndef BOOST_NO_MEMBER_TEMPLATE_FRIENDS public: #else friend class boost::archive::detail::interface_oarchive<derived_t>; friend class basic_binary_oarchive<derived_t>; friend class basic_binary_oprimitive<derived_t, std::ostream>; friend class boost::archive::save_access; #endif // add base class to the places considered when matching // save function to a specific set of arguments. Note, this didn't // work on my MSVC 7.0 system // using binary_oarchive_impl<derived_t>::load_override; // so we use the sure-fire method below. This failed to work as well template<class T> void save_override(T & t, BOOST_PFTO int){ binary_oarchive_impl<fast_binary_oarchive>::save_override(t, 0); // verify that this program is in fact working by making sure // that arrays are getting passed here BOOST_STATIC_ASSERT(! (boost::is_array<T>::value) ); } // all the functions have emtpy body void save_override(const class_id_optional_type &, int) {} void save_override(const version_type&, int) {} void save_override(const class_id_type&, int) {} void save_override(const class_id_reference_type&, int) {} void save_override(const object_id_type&, int) {} void save_override(const object_reference_type&, int) {} void save_override(const tracking_type&, int) {} void save_override(const class_name_type&, int) {} public: fast_binary_oarchive(std::ostream & os, unsigned flags = no_header | no_codecvt | no_xml_tag_checking | no_tracking) : binary_oarchive_impl<derived_t>(os, flags) {} }; Is this correct? What features of the serialization library will not I be able to use?

2 1

Re: [Boost-users] [WAVE] bug: generic lexing error
by Hartmut Kaiser 28 Apr '06

28 Apr '06

> Andreas Sæbjørnsen wrote: > > > I still experience the same problems as before with regards to this > > memory bug. I am not able to test this with the wave driver > though as > > it compiles with a lot of errors in a lot of header files with gcc > > 3.4.3 on linux, but it shows up in the lexed_tokens example. > > That's puzzling me. Why the Wave driver doesn't compile for you? > > Are you sure to really have the corrected version (I've > replaced all memcpy's in cpp_re.cpp with memmove's, which > should avoid the problem reported by valgrind)? Seems that the public CVS currently is a bit out of sync if compared to the 'real' CVS database. At least ViewCVS shows a state which is 5 weeks old. This would explain your compilation problems. And I don't know where you may get a current CVS version, sorry. Regards Hartmut > > Am I correct to believe that HEAD is the default CVS version for > > boost-cvs? I tested this on a default checkout from the boost-cvs. > > HEAD should be the default. > > Regards Hartmut >

2 3

Multi_array: A "proper" iterator over a N>1 Multi_array?
by Greg Link 28 Apr '06

28 Apr '06

I'm using multi_array types as containers for some things I'm working on, and the number of dimensions can get quite large - often over 7, sometimes as high as 11. I'd like to do things such as initialize these multi_arrays, access all elements within them, and other such 'reasonable' operations. Thing is, the only iterator interface to multi_array returns a reference to the nested containers. Thus, to iterate over all sub- elements, I'm forced to have 7-11 nested for-loops to iterate over the entire thing. Is there an existent means of simply iterating over the entire set of data in the multi_array? Has anyone implemented such a thing, or is such a thing something to be desired within boost::multi_array? It seems like such a iterator would be especially useful if also implemented/supported within subgroups, allowing the user to define sub-ranges of the multi_array, then iterate over each element therein, without having to worry about dimensionality, et cetera. Any thoughts, ideas, or pointers to how to easily (without tons of code) iterate over an 11-dimensioned object is appreciated. In conclusion then, I've got two problems. One is "How can I simplify my each-element access to a high-dimensional multi_array", and the second is "Should there be a direct-to-element iterator in multi_array, and does the documentation adequately address the issue of whether or not an iterator points to elements directly or to a sub- graph?" Thanks for your time and troubles, - Greg Link

3 5

[serialization] std::list compiler error
by Stephan Nies 28 Apr '06

28 Apr '06

Hello, I want to use boost/serialization combined with the MPI (Message Passing Interface) to add parallel computation to a traffic simulation i am working on. I want to serialize a STL container filled with custom Objects to send it to another process. If I don't set "boost::serialization::track_never" for every class i get a compiler error. The Example Code (without any MPI) below creates this error when compiled: g++ test3.cpp -I/usr/include -L/usr/lib -lboost_serialization -o test3 /usr/include/boost/archive/detail/oserializer.hpp: In function void boost::archive::save(Archive&, T&) [with Archive = boost::archive::text_oarchive, T = gps_packet]: /usr/include/boost/archive/basic_text_oarchive.hpp:78: instantiated from void boost::archive::basic_text_oarchive<Archive>::save_override(T&, int) [with T = gps_packet, Archive = boost::archive::text_oarchive] /usr/include/boost/archive/detail/interface_oarchive.hpp:78: instantiated from Archive& boost::archive::detail::interface_oarchive<Archive>::operator<<(T&) [with T = gps_packet, Archive = boost::archive::text_oarchive] test3.cpp:76: instantiated from here /usr/include/boost/archive/detail/oserializer.hpp:567: error: invalid application of sizeof to incomplete type boost::STATIC_ASSERTION_FAILURE<false> make: *** [test3] Error 1 Here is a short example that creates the above error based upon the example in the documentation: ---code start-------- #include <iostream> #include <sstream> #include <boost/archive/text_oarchive.hpp> #include <boost/archive/text_iarchive.hpp> #include <boost/serialization/list.hpp> #include <list> class gps_position { private: friend class boost::serialization::access; template<class Archive> void serialize(Archive & ar, const unsigned int version) { ar & degrees; ar & minutes; ar & seconds; } int degrees; int minutes; float seconds; public: gps_position(){}; gps_position(int d, int m, float s) : degrees(d), minutes(m), seconds(s) {} void pretty_print() { std::cout << "deg:" << degrees ; std::cout << " min:" << minutes ; std::cout << " sec:" << seconds << std::endl; } }; typedef std::list<gps_position> list_gps_position; class gps_packet { private: friend class boost::serialization::access; template<class Archive> void serialize(Archive & ar, const unsigned int version) { ar & gps_list; } public: gps_packet(){} list_gps_position gps_list; }; // only when I uncomment the following // boost::serialization::track_never // statements programm will compile //BOOST_CLASS_IMPLEMENTATION(gps_position,boost::serialization::object_serializable); //BOOST_CLASS_TRACKING(gps_position, boost::serialization::track_never); //BOOST_CLASS_IMPLEMENTATION(gps_packet,boost::serialization::object_serializable); //BOOST_CLASS_TRACKING(gps_packet, boost::serialization::track_never); int main() { // create come pseudo data gps_position g0(1, 2, 3.3); gps_position g1(11, 22, 33.33); gps_packet packet; packet.gps_list.push_back(g0); packet.gps_list.push_back(g1); // print before transfer std::cout << "Before:" << std::endl; packet.gps_list.front().pretty_print(); packet.gps_list.back().pretty_print(); // serialize to chars in order to use // the MPI::CHAR elementary datatype std::ostringstream ost; boost::archive::text_oarchive oa(ost); oa << packet; const char* chars = ost.str().data(); // assume some message passing // through MPI::CHAR to happen here std::cout << "Transfering:" << std::endl; std::cout << chars << std::endl; // reconstruction from chars for further // computation std::istringstream ist(chars); boost::archive::text_iarchive ia(ist); gps_packet packet_new; ia >> packet_new; // print transfer result std::cout << "After:" << std::endl; packet_new.gps_list.front().pretty_print(); packet_new.gps_list.back().pretty_print(); return 0; } ---code end-------- when i uncomment BOOST_CLASS_TRACKING(gps_position, boost::serialization::track_never); BOOST_CLASS_TRACKING(gps_packet, boost::serialization::track_never); i can compile and execute with this output: Before: deg:1 min:2 sec:3.3 deg:11 min:22 sec:33.33 Transfering: 22 serialization::archive 3 0 0 0 0 2 0 0 1 2 3.3 11 22 33.330002 After: deg:1 min:2 sec:3.3 deg:11 min:22 sec:33.33 Further Questions: 1. I would also like to use just list<gps_position> or the typedef, instead of the wrapper class gps_packet which i need to pass something to BOOST_CLASS_TRACKING(gps_packet, boost::serialization::track_never); 2. Is it a good Idea to use something like: BOOST_CLASS_IMPLEMENTATION(gps_position,boost::serialization::object_serializable); BOOST_CLASS_TRACKING(gps_position,boost::serialization::track_never); since I don't want to store but just pass the Information or are there any pitfalls? Thanks, Stephan Nies -- When you see this text i am using the webinterface of my mail provider and i apologize for any commercials they may attach below. Echte DSL-Flatrate dauerhaft für 0,- Euro*! "Feel free" mit GMX DSL! http://www.gmx.net/de/go/dsl

1 0

Visual C++ 8 - Example Crashes with Access Violation
by Brown Gabe 28 Apr '06

28 Apr '06

Hi Everyone, Hopefully this isn't some noob question, but here goes... I'm trying to run boost 1.3.1 in Visual C++ 8 VC8 and I keep getting an access violation at the termination of my thread. I've tried looking to see what the problem is but I can't seem to determine the problem. I'm using the plain example thread.cpp ====================[ EXAMPLE CODE ]======================= #include <boost/thread/thread.hpp> #include <boost/thread/xtime.hpp> #include <iostream> struct thread_alarm { thread_alarm(int secs) : m_secs(secs) { } void operator()() { boost::xtime xt; boost::xtime_get(&xt, boost::TIME_UTC); xt.sec += m_secs; boost::thread::sleep(xt); std::cout << "alarm sounded..." << std::endl; } int m_secs; }; int main(int argc, char* argv[]) { int secs = 5; std::cout << "setting alarm for 5 seconds..." << std::endl; thread_alarm alarm(secs); boost::thread thrd(alarm); thrd.join(); } ====================[ EXAMPLE CODE ]======================= I get the expected out from the application. Upon tracing my code, the exception only occurs when the worker thread exits. Exception occured in threadex.c on line 348. Unhandled exception at 0x7c9012b4 in test.exe: 0xC0000005: Access violation reading location 0xcccccccc. " __try { _endthreadex ( ( (unsigned (__CLR_OR_STD_CALL *)(void *))(((_ptiddata)ptd)->_initaddr) ) ( ((_ptiddata)ptd)->_initarg ) ) ;" Any thoughts on what might cause an error like this??? -- -Gabe

4 7

[boost::any] Newbie Question About Compiler Warning on HPUX with boost::any
by Richard Ward 28 Apr '06

28 Apr '06

Hello, I'm a new user to boost and have a question about a compiler warning I'm seeing. I'm currently using just the boost::any type and on HPUX v11.00 PA with aCC: HP ANSI C++ B3910B A.03.56 I'm seeing the following warning: Warning (anachronism) 823: "/Components/dev/Boost/v1.33.1/boost/type_traits/is_reference.hpp", line 38 # Redundant preprocessing concatenation operation results in two valid preprocessing tokens. Depending on this implementation defined behavior will result in non-portable code. BOOST_TT_AUX_BOOL_TRAIT_DEF1(is_reference,T,false) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Here's the compile line that generated the warning: aCC -AA -DACE_HAS_STANDARD_CPP_LIBRARY=1 -D_RWSTD_MULTI_THREAD -D_REENTRANT -U_INCLUDE__STDC_A1_SOURCE -DACE_HAS_THREADS -D_POSIX_C_SOURCE=199506L -DHPUX_VERS=1100 -DACE_LACKS_PRAGMA_ONCE +W302 +W930 +W849 -mt +DAportable +Z -D_LARGEFILE64_SOURCE=1 +O3 +Osize -DACE_NDEBUG -DACE_USE_RCSID=0 -DACE_HAS_EXCEPTIONS -D__ACE_INLINE__ -D__UNIX -c -o .objects/HP-UX.B.11.00_RISC_aCC.A.03.37/release//LicAgentSvc.o LicAgentSvc.cpp Because the boost::any header is included in one of my headers, I see this warning hundreds of times. While I could just suppress it I prefer to first try to figure out what the issue is. I dislike suppressing warnings I don't understand. Does anyone know what this warning means and/or have any suggestions for what to do about it? Thanks, Richard J. Ward Symantec Corp.

1 0

bjam, 64-bit builds, VC8
by Jarl Lindrud 28 Apr '06

28 Apr '06

Hi! I need to build the boost libraries, and eventually my own application, in 64 bi t mode on Windows, but unfortunately I haven't been able to find any information on how to do so. Does anyone here have any tips? Typically, for a 32 build of boost.thread, for instance, I would type "bjam --with-thread -sTOOLS=vc-8_0 stage", but I don't know what the invocation should be for a 64 bit build. I'm using Visual C++ 2005 (the full version, not "express"). Thanks! Jarl.

1 0

multi_index, best way to do insert_or_replace
by bert hubert 28 Apr '06

28 Apr '06

Hi, I misread (or didn't read) the boost::multi_index_container documentation and assumed 'insert' on a unique index would do 'replace', but of course it doesn't and shouldn't. So, I've made this change: - s_negcache.insert(ne); + + pair<negcache_t::iterator, bool> res=s_negcache.insert(ne); + if(!res.second) { + s_negcache.erase(res.first); + s_negcache.insert(ne); + } But I can't help feeling there is probably a better and prettier way. Do people have any suggestions? In this case, ne is a NegCacheEntry, defined as: struct NegCacheEntry { string d_name; QType d_qtype; string d_qname; uint32_t d_ttd; }; And the index: typedef multi_index_container < NegCacheEntry, indexed_by < ordered_unique< composite_key< NegCacheEntry, member<NegCacheEntry, string, &NegCacheEntry::d_name>, member<NegCacheEntry, QType, &NegCacheEntry::d_qtype> >, composite_key_compare<CIStringCompare, std::less<QType> > >, ordered_non_unique< member<NegCacheEntry, uint32_t, &NegCacheEntry::d_ttd> > > >negcache_t; Thanks! -- http://www.PowerDNS.com Open source, database driven DNS Software http://netherlabs.nl Open and Closed source services

2 2