November 2005 - Boost-users

[Boost][serialization] Serialization and DLLs
by Topi Mäenpää 03 Nov '05

03 Nov '05

Hi, Some time ago, we had a discussion about serialization of dynamically loaded classes. I got my code work on Linux, but still cannot do the same on Windows. I'd like someone to elaborate the stuff below a bit: Robert Ramey wrote: > Another post has me convinced that the problem is that there are multiple > copies of a base class extended_type_info record which I believe > shouldn't occur. I can confirm that this indeed happens. > If you trap your debugger the serialization.dll (shared_lib) at line 58 of > extened_type_info.cpp you should trap every time an extended_type_info > record is "self_registered". Examination of the stack should reveal > where this is being called from. I expect that the base class is being > self registered from both the DLL as well as the main executable. Whenever I register a class from a DLL, its base classes get registered as well, although they already have been registered. If I load multiple DLLs that all contain derived classes, the base classes are registered many times, but only once for each DLL. > My current thinking is that this behavior should be prohibited and > detected with an exception. Other users have addressed this > by tweaking the code in extended_type_info.cpp to permit > multiple "self registration" of the same extended_type_info type. > This eliminated their symptoms and everything seems to work. > But from my understanding of the library, I believe this just > hides a deeper problem. Can someone point out the details? How should this be done? Any other suggestions? Best regards, Topi

2 1

Newbie help needed
by Ricardo Kirkner 03 Nov '05

03 Nov '05

Hi: I am trying to write python bindings for an already existant c++ library. I started using pyste, because I felt this was the easiest way to do it. However I keep getting errors (mostly I think because of string and vector arguments conversion). I just wanted to ask if I am on the right track here. Maybe someone can give me some pointers on how to make this task easier? (I was tempted to switch to pyrex, swig, etc, but everything seems to have its own problems.) Should I use pyste, or should I rather write boost.python code directly? Thank you, Ricardo

2 1

enable_if, type_traits and Visual C++ 7.0
by Ovanes Markarian 03 Nov '05

03 Nov '05

Hello all, I am trying to get enable if to work with Visual C++ 7.0 and boost 1.33. for the following member function implemenation I get an error: template<> typename boost::enable_if<boost::is_pointer<_ElemFct::result_type>, _ElemFct::result_type>::type operator->() { return get_at(ptr_coll, pos); } template<> typename boost::enable_if<boost::is_reference<_ElemFct::result_type>, boost::add_pointer<_ElemFct::result_type>::type>::type operator->() { return &(get_at(ptr_coll, pos)); } My compiler is in German. I have tried to translate the error messages... error C2504: 'boost::enable_if_does_not_work_on_this_compiler<T>' : Base class undefined with [ T=boost::function2<std::string &,MemberTest *,int,int>::result_type ] see reference to the instance of compiled class template 'boost::enable_if<Cond,T>' with [ Cond=boost::is_pointer<boost::function2<std::string &,MemberTest *,int,int>::result_type>, T=boost::function2<std::string &,MemberTest *,int,int>::result_type ] see reference to the instance of compiled class template 'generic_iterator<_Coll,_Ty,_ElemFct,_LenFct,_Diff,_Ptr,_Ref,_Cat>' with [ _Coll=MemberTest, _Ty=std::string, _ElemFct=element_retrieval, _LenFct=length_retrieval, _Diff=int, _Ptr=std::string *, _Ref=std::string &, _Cat=std::bidirectional_iterator_tag ] error C2039: 'type' : Is not an element of 'boost::enable_if<Cond,T>' with [ Cond=boost::is_pointer<boost::function2<std::string &,MemberTest *,int,int>::result_type>, T=boost::function2<std::string &,MemberTest *,int,int>::result_type ] With Kind Regards, Ovanes Markarian

2 1

[date-time] posix timezone string
by Stering Wolfram 03 Nov '05

03 Nov '05

Hi, I am implementing a year_based_date-derived date generator to calculate the easter sunday for a given year. Now, this is the easy part, and it is done already. But I wonder how to generate a meaningful posix timezone string that is expected to result from calling the generator's to_string() member. The documentation at the (not well formatted, b.t.w.) "Description" section of class posix_time_zone didn't give any hints for Easter Sunday either. Any ideas? regards, -wolfi -- DI Wolfram Stering (Entwicklung) HALE electronic GmbH., Salzburg, Austria Tel: +43 (662) 439011 550 Fax: +43 (662) 439011 9 Email: wolfram.stering(a)hale.at

2 1

How to apply -Winline with Bjam v2
by Jeffrey Holle 03 Nov '05

03 Nov '05

I am using gcc 4.0.2 and Boost.Jam Version 3.1.11. I've been told that the way to determine if an inline method is actually inline by the compiler is to compile with the "-Winline" option. When this is done, a warning is emmitted when a sugguested inline method wasn't inlined. I'm using bjam v.2. I have not found the equivalent to "features.jam" in the v2 documentation, so I tried "<cxxflags>-Wonline", which I got from /tools/build/v1/features.jam. This doesn't work. What is the way to do it in Bjam v2?

1 0

design question for boost::program_options and dynamic options
by boost＠nogga.de 02 Nov '05

02 Nov '05

Hello, I'm new to the boost::program_options library and tried to use it for a command pattern based software. The program is launched with the following commandline: > program command [Options] <Files> The program can use different output formatters with --formatter={XML|binary|text}. Every command and formatter specify their own set of options. I first tried to collect all options form the commands and formatters and parse the command line only once. This lead to the problem, that due to the different positional options information the commandline parser always threw an exception due to wrong command line formatting. Since I couldn't find a way to add mutliple optional option groups to the parser, I moved the parsing code into my command base class and query the derived commands for the different required information, like: class Command { public: vritual options_description GetOptionsDescription (); virtual options_description GetHiddenOptionsDescription (); virtual positional_program_options GetPositionalProgramOptions (); virtual Run (variables_map const & vm) = 0; int Execute (std::vector<std:.string> args) { options_description desc; descr.add (GetOptionsDescription ()).add (GetHiddenOptionsDescription ()); variables_map vm; store(command_line_parser(args) .options(desc). .positional(GetPositionalProgramOptions ()) .run(), vm); notify(vm); Run (vm); } } The execute function is called form the program main after the "command" has been stripped from the commandline and the concrete command was created by the factory. So far so good. This works for the different commands, but not for the formatters. How can I parse the command line dynamically? E.g. if I encounter the "--formatter=xml" argument, a new XMLFormatter is created and the options description from this formatter are added to list of valid options? I know, that there is the notify mechanism, but this only works after the commandline has been successfully parsed. And also the custom parsers do not solve my problem. Any idea how to solve this issue? Best regards Dirk

2 2

Tokenizer: how to eliminate unwanted null tokens?
by CN 02 Nov '05

02 Nov '05

Hi! The following code uses '\0' and ';' as seperators and produces the result: <X><Y><><Z><> string str=string("X;Y\0\0Z\0",7); typedef tokenizer<boost::escaped_list_separator<char> > Tok; escaped_list_separator<char> sep(string(),string("\0;",2),string()); Tok tokens(str, sep); for(Tok::iterator i = tokens.begin();i != tokens.end();++i) cout << "<" << *i << ">"; How to revise it in order to get this result? <X><Y><Z> Thank you in advance! CN -- http://www.fastmail.fm - Does exactly what it says on the tin

2 1

configuring a library to work with multiple versions of boost
by Ross Boylan 01 Nov '05

01 Nov '05

I have a library that uses boost and also uses a pre-release version of the ptr_container (header only) library that eventually made it into boost 1.33 (I think). I would like people to be able to compile my code with various versions of boost. I've had it working with 1.31 and 1.32, and expect to have it working with 1.33 when that reaches my distribution. I have a subdirectory with the pre-release ptr_container library and some supporting code from elsewhere in boost. This is in a "boost" subdirectory of my project, which has included the directory above that unconditionally in my compilations. With 1.33 I think I want to ignore my local stuff completely. I'm using autoconf and libtool. Are there autoconf macros available for detecting the boost libraries? Do people have any suggestions about the best way to proceed? One possibility that occurs to me is to test for the version of the library at configure time, and make the include path depend on the results. Thanks. Ross Boylan

2 2

ptr_map
by Oleg Smolsky 01 Nov '05

01 Nov '05

Hello Thorsten, thanks for the ptr_vector clarifications. I have another question about ptr_map: I can declare a map iterator in the following way: ptr_map<string, SomeClass>::iterator i = m.begin(); Now, I'd expect to be able to compile the following, as with a normal map: string s = i->first; i->second.SomeMethod(); ...assuming SomeClass::SomeMethod() exists and is declared as public. However, "first" and "second" don't seem to exist in ptr_map. Could you shed some light on how I can iterate your map please? Best regards, Oleg.

2 3

[test]
by Paul Giaccone 01 Nov '05

01 Nov '05

Is the use of BOOST_CHECK_EQUAL not permitted in init_unit_test_suite(int argc, char* argv[])? The following code throws an exception and reports the error "Boost.Test internal framework error: unknown reason" boost::unit_test::test_suite* init_unit_test_suite(int argc, char* argv[]) { BOOST_CHECK_EQUAL(foo(argc, argv), 0); boost::unit_test::test_suite* test = BOOST_TEST_SUITE("My test"); test -> add(BOOST_TEST_CASE(&test1)); test -> add(BOOST_TEST_CASE(&test2)); return test; } when the following works without any problems: boost::unit_test::test_suite* init_unit_test_suite(int argc, char* argv[]) { int result = foo(argc, argv); assert(result == 0); boost::unit_test::test_suite* test = BOOST_TEST_SUITE("My test"); test -> add(BOOST_TEST_CASE(&test1)); test -> add(BOOST_TEST_CASE(&test2)); return test; } I have no problem with using the latter code, but I am surprised that I get an exception in the former code and that it is caught by "..." and not by the library. If BOOST_CHECK_EQUAL (and BOOST_CHECK, BOOST_CHECK_CLOSE, etc) is not allowed in init_unit_test_suite(), is this fact documented? Should it generate an exception that can be caught by the library and reported with a useful message? Thanks, Paul

2 2