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

[boost-users] converting InputIterator to ForwardIterator
by Igor R 11 Mar '09

11 Mar '09

Hello, I'd like to do the subj, in order to be able to apply some std algorithms to files, like this: ifstream file("myfile", std::ios::binary|std::ios::in); std::string delim("\r\n\r\n"); // sequence to search typedef std::istreambuf_iterator<char> iterator; iterator begin(file), end; iterator pos = std::find_end(begin, end, delim.begin(), delim.end()); // won't work, because istreambuf_iterator is *not* ForwardIterator I found something called multi_pass in boost::spirit. However, I'm not sure if it's relevant or deprecated class (it seems to be a part of the "old" Spirit), and whether this's the only Boost utility I can use to perform the above conversion. Thanks.

1 0

[interprocess] Memory mapped file based allocator
by dhruva 11 Mar '09

11 Mar '09

Hi, I am looking for a memory mapped file based allocator that does not map the whole file into memory. The size may be huge and I would not want to map all the contents into process memory. Since my read and write access are sequential, I would like the mapping to happen on demand (based on access violation trying to read an unmapped region of memory). Is there something like that in Interprocess/Boost (since the implementation needs to work both on M$ and GNU/Linux)? regards, dhruva Bollywood news, movie reviews, film trailers and more! Go to http://in.movies.yahoo.com/

4 5

[Random] Sigma equal to zero in normal distribution.
by Rodolfo Federico Gamarra 11 Mar '09

11 Mar '09

Hi all, The documentation for the normal distribution (http://www.boost.org/doc/libs/1_38_0/libs/random/random-distributions.html#…) states: "Members explicit normal_distribution(const result_type& mean = 0, const result_type& sigma = 1); Requires: sigma > 0" In the actual code of this constructor (http://www.boost.org/doc/libs/1_38_0/boost/random/normal_distribution.hpp) it can be seen "assert(_sigma >= result_type(0));". That is, acepts sigma == 0. Indeed, there's no "division-by-zero" or so, because the method that is used ("Box-Muller") doesn't actually divide by sigma: return _cached_rho * (_valid ? cos(result_type(2)*pi*_r1) : sin(result_type(2)*pi*_r1)) * _sigma + _mean; So, in this case, the numbers will be all equal to _mean. From the mathematical point of view there's no such thing as a normal distribution with sigma equal to zero; but, computationally, may be handy to allow that. What's the expected behavior? Is there a typo in the docs? Or is it that the case sigma equal to zero is indeed undefined (so that anything could happen)? Thanks in advance, cheers. -- Rodolfo Federico Gamarra

2 1

Thread usage in boost other then boost.thread??
by Bill Taylor 11 Mar '09

11 Mar '09

Hello, I joined the list because I am helping a PHD candidate run a very long program with Condor. But the mode we need to use does not work if pthreads are used because we can not save the state of the job when it needs to run somewhere else. Is it known what components in the boost library use threads? Bill Taylor Condor Project

1 0

[asio] when to use io_service::post or deadline_timer::async_wait?
by Markus Werle 11 Mar '09

11 Mar '09

Hi! After reading the referece docs I got the impression that the only difference between io_service::post and deadline_timer::async_wait is the fact that the former triggers execution immediately (after the caller function returns?) while the latter waits until timer expires before execution of the function object. Is this really the only difference or is there any other issue to be considered? Markus

2 1

Generalised compile-time hash function
by Nikhilesh S 11 Mar '09

11 Mar '09

Hey guys! I'm not very familiar with boost.mpl, or boost.preprocessor, so I wanted to ask. I've written a compile-time hash function. Here's an example that uses it:- #include <iostream> using namespace std; //Compile-time hasher. #define HASH(partial,c) (((partial) << 1) ^ (c)) #define HASH_1(p1) HASH(0,(p1)) #define HASH_2(p1,p2) HASH(HASH_1(p1),(p2)) #define HASH_3(p1,p2,p3) HASH(HASH_2(p1,p2),(p3)) #define HASH_4(p1,p2,p3,p4) HASH(HASH_3(p1,p2,p3),(p4)) #define HASH_5(p1,p2,p3,p4,p5) HASH(HASH_4(p1,p2,p3,p4),(p5)) #define HASH_6(p1,p2,p3,p4,p5,p6) HASH(HASH_5(p1,p2,p3,p4,p5),(p6)) #define HASH_7(p1,p2,p3,p4,p5,p6,p7) HASH(HASH_6(p1,p2,p3,p4,p5,p6),(p7)) #define HASH_8(p1,p2,p3,p4,p5,p6,p7,p8) HASH(HASH_7(p1,p2,p3,p4,p5,p6,p7),(p8)) #define HASH_9(p1,p2,p3,p4,p5,p6,p7,p8,p9) HASH(HASH_8(p1,p2,p3,p4,p5,p6,p7,p8),(p9)) #define HASH_10(p1,p2,p3,p4,p5,p6,p7,p8,p9,p10) HASH(HASH_9(p1,p2,p3,p4,p5,p6,p7,p8,p9),(p10)) #define HASH_11(p1,p2,p3,p4,p5,p6,p7,p8,p9,p10,p11) HASH(HASH_10(p1,p2,p3,p4,p5,p6,p7,p8,p9,p10),(p11)) #define HASH_12(p1,p2,p3,p4,p5,p6,p7,p8,p9,p10,p11,p12) HASH(HASH_11(p1,p2,p3,p4,p5,p6,p7,p8,p9,p10,p11),(p12)) #define HASH_13(p1,p2,p3,p4,p5,p6,p7,p8,p9,p10,p11,p12,p13) HASH(HASH_12(p1,p2,p3,p4,p5,p6,p7,p8,p9,p10,p11,p12),(p13)) #define HASH_14(p1,p2,p3,p4,p5,p6,p7,p8,p9,p10,p11,p12,p13,p14) HASH(HASH_13(p1,p2,p3,p4,p5,p6,p7,p8,p9,p10,p11,p12,p13),(p14)) #define HASH_15(p1,p2,p3,p4,p5,p6,p7,p8,p9,p10,p11,p12,p13,p14,p15) HASH(HASH_14(p1,p2,p3,p4,p5,p6,p7,p8,p9,p10,p11,p12,p13,p14),(p15)) #define HASH_16(p1,p2,p3,p4,p5,p6,p7,p8,p9,p10,p11,p12,p13,p14,p15,p16) HASH(HASH_15(p1,p2,p3,p4,p5,p6,p7,p8,p9,p10,p11,p12,p13,p14,p15),(p16)) /* .. omitted the rest, but it's there till around HASH_30 .. */ //Runtime hasher. unsigned int hash(char const *p) { unsigned partial = 0; while (*p) { partial <<= 1; partial ^= *p; ++p; } return partial; } //Test it in a main() int main() { char str[31] = "thisisatest"; //Yay! Switch on strings! switch (hash(str)) { case HASH_11('t','h','i','s','i','s','a','t','e','s','t'): cout << "You said, 'this is a test'!"; break; case HASH_16('y','e','t','a','n','o','t','h','e','r','s','t','r','i','n','g'): cout << "You said, 'yet another string'!"; break; } return 0; } So, I was wondering how I could use boost.preprocessor to be able to write a more generic hash function, something like HASH_GENERIC(mystring), or at least HASH_GENERIC(8,mystring). BTW, here's a vim script that changes any thing like '@something@' on the curent line into "HASH_9('s','o','m','e','t','h','i','n','g')", when you type ',h':- map ,h :s/@$.\+$@/\='HASH_'.strlen(submatch(1))."('".join(split(submatch(1), '\zs'), "','")."')"/<CR> ;-) -- Nikhilesh S http://nikki.drupalsite.org

1 0

Making friend with boost::interprocess
by Oncaphillis 11 Mar '09

11 Mar '09

Hi, I'd like to implement a cross-process singleton which resides in a shared memory segment. So the basic idea is to provide a factory method that is the only way to construct the Singleton. e.g. class Foo { public: static Foo & GetInstance() { try { boost::interprocess::managed_shared_memory segment( boost::interprocess::open_or_create, "FooSegment", 65536); void_allocator_t alloc(segment.get_segment_manager()); return *(segment.find_or_construct<Foo>( "FooObject")(alloc)); } catch(std::exception & ex) { throw; } } private: Foo(const void_allocator_t & a) { ... ... } Foo(const Foo &); } Of course somewhere deep in the guts of boost I get an error since the Foo constructor is private. Who do I have to befriend in order to make this scheme work ? Thank you O.

1 0

pure virtual function call error
by e120281 11 Mar '09

11 Mar '09

class A { public: int i; A( int i ) { this->i = i; } ~A() {} template<class Archive> void serialize(Archive &ar, const unsigned int version) { ar & i; } }; class B { public: B() {} ~B() { std::string filename = "C:\\a2.txt"; const A a( 2 ); std::ofstream ofs( filename.c_str(), std::ios::out | std::ios::binary ); boost::archive::binary_oarchive boa( ofs ); boa << a; } static B* getInstance() { static B instance; return &instance; } void serialize() { std::string filename = "C:\\a1.txt"; const A a( 1 ); std::ofstream ofs( filename.c_str(), std::ios::out | std::ios::binary ); boost::archive::binary_oarchive boa( ofs ); boa << a; } }; int main() { B* b = B::getInstance(); b->serialize(); } When I run the above code in visual studio 2005 I get pure virtual function call error. If I comment out the line b->serialize(); then everything is ok. What is wrong with the above code??? -- View this message in context: http://www.nabble.com/pure-virtual-function-call-error-tp22434024p22434024.… Sent from the Boost - Users mailing list archive at Nabble.com.

1 0

How to build specified library for install?
by Dancefire 11 Mar '09

11 Mar '09

Hi, I am going to build boost::iostreams and install to some location. I tried following command: C:\boost_1_38_0>bjam toolset=msvc variant=debug,release link=static threading=multi runtime-link=shared -sZLIB_SOURCE=E:\zlib123 -sBZIP2_SOURCE=E:\bzip2-1.0.5 --prefix="E:\Program Files\boost\boost_1_38_0\" --with-iostreams install However, I got following error: error: wrong library name 'iostreams install' in the --with-<library> option. It looks like that bjam consider install as a part of the library name. Is anything wrong with me? or it is a bug of bjam? Tao Wang

1 1

[Q] boost class to handle SQL Decimal (19,6)
by V S P 11 Mar '09

11 Mar '09

Hello, I have been searching and searching but could not find a Boost (or Boost-friendly) extension that handles SQL Decimal datatype (in my case decimal (19,6) I did find in Boost's vault http://www.boostpro.com/vault/index.php mp_math_v04 by Baraclese but not clear if it supports decimal point (I did not see any documentation) thanks in advance for any help, Vlad -- V S P toreason(a)fastmail.fm -- http://www.fastmail.fm - Or how I learned to stop worrying and love email again

1 0

MPL design question
by Jesse Perla 11 Mar '09

11 Mar '09

I have a class that is getting a huge number of template parameters (mostly a bunch of size_t used for vector sizes). They have a natural organization into groups, but almost all of them have a default as 0. How do you suggest that I design the template parameter for this? Is there any way to use some kind of associative container so I could look these up by "name" and default to 0 if they aren't there? I can't see how to use mpl::map to do this kind of stuff. Simplified Pseudocode: -------------------- What I have: template<std::size_t IndividualSize = 0, std::size_t AggregateSize> //Many others class my_problem{...}; What I want: template<class StateSizes> class MyProblem{...} Called something like: typedef my_problem<map<AggregateSize, 2> > my_problem_type; where AggregateSize is an enum, placeholder struct, etc. And in this case I would default IndividualSize to 0. -Jesse

2 1

Re: [Boost-users] [wave] assertion failed - multithread issue?
by Ein Held 10 Mar '09

10 Mar '09

Hello Hartmut Using multiple threads, wave crashes randomly after a few seconds of running. The crash is actually an assert in test.exe!std::list<boost::wave::cpplexer::lex_token<boost::wave::util::file_position<boost::wave::util::flex_string<char,std::char_traits<char>,std::allocator<char>,boost::wave::util::CowString<boost::wave::util::AllocatorStringStorage<char,std::allocator<char> >,char *> > > >,boost::fast_pool_allocator<boost::wave::cpplexer::lex_token<boost::wave::util::file_position<boost::wave::util::flex_string<char,std::char_traits<char>,std::allocator<char>,boost::wave::util::CowString<boost::wave::util::AllocatorStringStorage<char,std::allocator<char> >,char *> > > >,boost::default_user_allocator_new_delete,boost::details::pool::win32_mutex,32> >::_Const_iterator<1>::operator*() Line 218 + 0x17 bytes C++ indicating that the iterator is not dereferencable. I am using an VC 9 x86 debug build on Vista x64. Here's the code: #include <fstream> #include <string> //Wave does not like the new interface :( //#define BOOST_FILESYSTEM_NO_DEPRECATED #include <boost/thread.hpp> #include <boost/wave.hpp> #include <boost/wave/cpplexer/cpp_lex_token.hpp> #include <boost/wave/cpplexer/cpp_lex_iterator.hpp> std::string file; const char* file_name="C:\\Program Files (x86)\\Microsoft SDKs\\Windows\\v5.0\\Include\\Windows.h"; void test() { for(;;) { typedef boost::wave::cpplexer::lex_token<> token_type; typedef boost::wave::cpplexer::lex_iterator<token_type> lex_iterator_type; typedef boost::wave::context<std::string::iterator, lex_iterator_type> context_type; context_type ctx (file.begin(), file.end(), file_name); ctx.set_language(static_cast<boost::wave::language_support>(boost::wave::support_cpp|boost::wave::support_option_variadics|boost::wave::support_option_include_guard_detection)); ctx.add_sysinclude_path("C:\\Program Files (x86)\\Microsoft SDKs\\Windows\\v5.0\\Include\\"); ctx.add_sysinclude_path("C:\\Program Files (x86)\\Microsoft Visual Studio 9.0\\VC\\include\\"); ctx.add_macro_definition("_CPPRTTI",true); ctx.add_macro_definition("_CPPUNWIND",true); ctx.add_macro_definition("_MT",true); ctx.add_macro_definition("_MSC_VER=1600",true); ctx.add_macro_definition("_M_IX86=600",true); ctx.add_macro_definition("_WIN32=600",true); context_type::iterator_type first = ctx.begin(); context_type::iterator_type last = ctx.end(); while (first != last) { ++first; } } } int main() { std::ifstream instream(file_name); file = std::string(std::istreambuf_iterator<char>(instream.rdbuf()),std::istreambuf_iterator<char>()); instream.close(); for(int i=0;i<10;++i) { boost::thread t(test); } test(); } Thank you

4 5

boost/math/special_functions/legendre.hpp causing compile errors in other headers
by Jeremy Conlin 10 Mar '09

10 Mar '09

I recently started using the Boost Math libraries because I need legendre polynomials. However I cam across a problem I was hoping someone on this list could help me with. My personal code compiles fine with no problems. When I add the line #include "boost/math/special_functions/legendre.hpp" then I get compiler errors attributed to one of my header files. Specifically I get: my_header.hh:109 error: ISO C++ forbids declaration of ‘tuple’ with no type As I mentioned before, I don't believe the problem is in my header file because it compiles just fine without the legendre.hpp file. Has anyone seen this problem before? Do you know how to fix it? I am running gcc 4.2.1 on Mac OS X 10.5.6. Thanks, Jeremy

2 2

[mpl] Trivial metafunctions and placeholders don't mix ?
by jean-louis.a.leroy＠fortis.com 10 Mar '09

10 Mar '09

Hello. I'm trying to use a pair<> to pass values inside a fold<> : template<typename Seq> struct remove_duplicates // even if not consecutive; also keep order { typedef typename mpl::second< mpl::fold< Seq, mpl::pair< mpl::set<>, mpl::vector<> >, typedef typename mpl::if_< mpl::has_key<mpl::first<mpl::_1>, mpl::_2>, mpl::_1, mpl::pair< typename mpl::insert< mpl::first<mpl::_1>, void, mpl::_2 >::type, // insert mpl::push_back< mpl::second<mpl::_1>, _2 >::type > // pair > >::type > type; }; It looks like first<> and second<> don't work on placeholders. My compiler - Visual Studio 2008 - complains that "'first' : is not a member of 'boost::mpl::arg<1>". When I look at the definition of first<> I see a "BOOST_MPL_AUX_LAMBDA_SUPPORT(1,first,(P))" that should take care of placeholders I guess. But when I look at the definition of the macro, alas : #if !defined(BOOST_MPL_CFG_NO_FULL_LAMBDA_SUPPORT) # define BOOST_MPL_AUX_LAMBDA_SUPPORT_SPEC(i, name, params) /**/ # define BOOST_MPL_AUX_LAMBDA_SUPPORT(i,name,params) /**/ What's the story here? My compiler is not good enough? Could be... While we're at it, should I put a BOOST_MPL_AUX_LAMBDA_SUPPORT(1,remove_duplicates,(P)) at the bottom of remove_duplicates<> ? Would it make my metafunction work with lambda placeholders? Thanks, Jean-Louis Leroy = = = = = = = = = = = = = = = = = = = = = = = = = Fortis disclaimer : http://www.fortis.be/legal/disclaimer.htm Privacy policy related to banking activities of Fortis: http://www.fortis.be/legal/privacy_policy.htm = = = = = = = = = = = = = = = = = = = = = = = = =

2 1

Trouble using boost::edge_range in Boost Graph library
by Artyom 10 Mar '09

10 Mar '09

Hi all, I am having trouble getting this function to work. I am using 'multisetS' for OutEdgeProperty as instructed in the documentation, however, my code refuses to compile with the following errors: http://rafb.net/p/LgT89332.html I am using Visual Studio 8.0 SP1 under Windows Vista x64. Regards, Artyom

2 2

[program_options] Getting access to the remaining argv
by Ruediger Berlich 10 Mar '09

10 Mar '09

Hi there, I have a program ("A") that calls other, external programs ("B"). A can get a number of command line options, which are parsed with the help of the program_options library. B may also accept options (which are possibly defined by other people and not known to me in advance). Thus I want to pass all those options that are not known to A on to B, in the hope that they are needed for that program. Is there a way to retrieve the remaining argv string, with all options "known" to A removed ? Or should I better invent a flag of the following kind for A --external-options="-a 1 -b 2 -c 3" whose value is passed on to the external program ? Thanks and Best Regards, Ruediger

3 2

BGL Graph support for named vertices in adjacency_list documentation?
by ljw＠onet.eu 10 Mar '09

10 Mar '09

Is there any documentation available for the support for named vertices in adjacency_list introduced with version 1.38 and announced in the release notes? Regards Łukasz Wojakowski

2 1

[asio] Synchronous/Asynchronous pattern
by Rutger ter Borg 10 Mar '09

10 Mar '09

In an application we would like to support synchronous operations in the application layer, whereas operations below it are done with asynchronous operations. Callback handlers are not appropriate in this case, To make the example more clear, suppose you have a remote_vector which you would like to iterate through, remote_vector a(10); for( int i=0; i<10; ++i ) { std::cout << a[i] << std::endl; } here, a[i] will cause messages to be sent to retrieve the actual vector data before printing it. The network layer is hidden from the user. What is the best way to "emulate" synchronous behaviour with asio? The thread running this iteration should not block, but do other work (at the io_service?) until the requested results are received. Many thanks, Kind regards, Rutger ter Borg

3 7

condition_variable::timed_wait doesn't wait
by Daniele Barzotti 10 Mar '09

10 Mar '09

Hi! I'm trying to use condition_variable::timed_wait without success. Using the following code, the timed_wait returns immediatly also if Receive() is never called. Where I wrong? void Receive() { boost::lock_guard<boost::mutex> lock(_mut); //... code ...// _data_ready = true; // Notify and unlock the mutex _cond.notify_one(); } bool Send() { boost::unique_lock<boost::mutex> lock(_mut); boost::xtime xt; xtime_get(&xt, boost::TIME_UTC); xt.sec += 5; _data_ready = false; //... send data ...// //wait for an answer.. bool ready = _cond.timed_wait(lock, xt, boost::lambda::var(_data_ready)); return ready; } I've also tried to use: ptime t = second_clock::universal_time(); ptime timeout = t + milliseconds(cmd_timeout); bool ready = _cond.timed_wait(lock, timeout, boost::lambda::var(_data_ready)); but with the same effect. Thanks, Daniele.

3 2

Re: [Boost-users] Pointer serialization and deserialization
by niranjan bangera 10 Mar '09

10 Mar '09

Robert, Yes the pointer is been initialized by NULL in the constructor of the class of that pointer. Thanks, Niranjan Do you know for a fact that the *pointer* has been initialized? Robert Ramey "niranjan bangera" <niranjannina@...<http://www.nabble.com/user/SendEmail.jtp?type=post&post=22416132&i=0>> wrote in message news:50a2f6f20903090457n37c0bfb2n3eec9d716a56c84c@...<news://50a2f6f20903090457n37c0bfb2n3eec9d716a56c84c@mail.gmail.com/> ... Hi all, Boost:1.38.0 compiler:gcc4.1.1 I am trying to serialize *and* deserialize a *pointer* to the class, but ended with the segmentation fault, Is that the null *pointer* *serialization * gives the segmentation fault?. I read in document that, library will maintain the Null *pointer*. if not how can i serialize? ,any docs or related refference is a;so acceptable. Thanks in advance.... one more thing: what -1 in archive(text_archive) indicates? Example ------------ #include <fstream> #include "boost/archive/text_iarchive.hpp" #include "boost/archive/text_oarchive.hpp" #include "boost/*serialization*/access.hpp" #include "boost/*serialization*/assume_abstract.hpp" class Dkpt { public: friend class boost::*serialization*::access; template<class Archive> void serialize(Archive & ar, const unsigned int file_version) { (void) file_version; ar & data_; ar & b_i; ar & f_er; ar & pa_er; ar & ct_c; } private: uint8_t data_; bool b_i; bool f_er; bool pa_er; bool ct_c; }; ---- class RB : public RM { public: virtual void checkpoint_save(boost::archive::text_oarchive & oa) { oa & dr; } virtual void checkpoint_retrive(boost::archive::text_iarchive & ia) { ia & dr; } private: Dkpt* dr; ;} Thanks. Niranjan

2 1

Re: [Boost-users] [wave] assertion failed - multithread issue?
by Ein Held 10 Mar '09

10 Mar '09

>> I use seperate wave instances for each thread. Anyway, I'll try to get >> gdb running and debug the problem. > >Thanks! Any hint is highly appreciated. > >> Nontheless there is another problem(?) in wave. Wave still uses the old >> filesystem interface, which makes it incompatible with code compiled >> with BOOST_FILESYSTEM_NO_DEPRECATED. That is a bug, right? > >Ohh, I was not aware of that. How can I reproduce this? > > > >Regards Hartmut Simply define BOOST_FILESYSTEM_NO_DEPRECATED before including boost.wave/boost.filesystem ;)

2 1

Re: [Boost-users] [wave] assertion failed - multithread issue?
by Ein Held 09 Mar '09

09 Mar '09

Hello Sohail actually test is never ment to return (check the for(;;) ;) - so this should not do any harm on the running program. This is definitely not a good coding style, but IMO acceptable for that simple non-productive example ;)

1 0

Static linking/GCC 4.3.2 on Linux (-fPIC missing?)
by Anteru 09 Mar '09

09 Mar '09

Hi, I've been using Boost 1.37 so far on Linux, and linked some libraries statically into a shared library. That is, I have a .so (Foo.so), which statically links in Boost.Filesystem (among others). Yesterday I tried to compile Boost 1.38 for Linux, everything worked fine, but as soon as it got to linking my .so, GCC stops as Boost.Filesystem has not been compiled with -fPIC. I don't remember I had to set -fPIC on 1.37 though. Setting cxxflags=-fPIC obviously solves the problem (along with a recompile of Boost), so my question is: Is this a regression? Did I miss something? The environment is Linux with GCC 4.3.2, on x64. Cheers, Anteru

3 6

multidimensional array query
by Sean Farrow 09 Mar '09

09 Mar '09

Hi: I need to define a multi-dimensional array with 4 collumns and two rows, What is the correct way to do this? My current definition is: boost::multi_array<std::wstring, 2> table(boost::extents[4][Lines.size()]); Is this correct, Lines is a vector decared prevously in the code. Any help appreciated Sean.

1 0

[serialization] redefinition of ‘struct boost::archive::detail::init_guid<T>
by Jan Boehme 09 Mar '09

09 Mar '09

Hi, I have something like the following object hierarchy: class Root {}; class A : Root {}; // BOOST_CLASS_EXPORT in header file class B : Root {}; // BOOST_CLASS_EXPORT in header file class U : A {}; // BOOST_CLASS_EXPORT in header file class V : A {}; // BOOST_CLASS_EXPORT in header file class X : B {}; // BOOST_CLASS_EXPORT in header file class Y : B {}; // BOOST_CLASS_EXPORT in header file There are a lot of shared pointers to the base classes Root, A and B which are to serialize.Boost 1.37 works fine while Boost 1.38 brings "redefinition of ‘struct boost::archive::detail::init_guid<A>" "redefinition of ‘struct boost::archive::detail::init_guid<B>" "redefinition of ‘struct boost::archive::detail::init_guid<U>" "redefinition of ‘struct boost::archive::detail::init_guid<V>" "redefinition of ‘struct boost::archive::detail::init_guid<X>" "redefinition of ‘struct boost::archive::detail::init_guid<Y>" after trying to link. It seems clear I have to put the BOOST_CLASS_EXPORT stuff in the belonging implementation files. But this corrupts the input of files which were created with the former binaries using the BOOST_CLASS_EXPORT in the header. I need an advice how and where the macro to place in terms of a long time decision. Below the class definition (which I find a good solution) seems no longer to work. Thank you very much, Jan.

2 1