October 2007 - Boost-users - lists.preview.boost.org

[filesystem] vacpp errors building 1_34_1 filesystem library on AIX
by Marty Fried 15 Oct '07

15 Oct '07

I've built all the boost libraries on Windows and Red Hat, but there are compiler errors building on AIX using Visual Age (xlC) for the filesystem library. It builds using gcc, but not vacpp, which is the AIX standard compiler. We are using the latest version, V. 8. There is more than one error, but I think they are all similar. I believe that the code envolved is not fully standard compliant. For example, in operations.hpp, there is an error on line 732, col 72; the full error is (I tried to wrap as best I could): The declaration of "end_itr" uses the undefined class "boost::filesystem:: basic_directory_iterator<boost::filesystem::basic_path <std::basic_string<char,std::char_traits<char>,std::allocator<char> >, boost::filesystem::path_traits> >" when the class must be complete. The code is called by line 730, called by 473, called by 689. The problem, I think, is that basic_directory_iterator is not defined until line 822, after all these calls. A coworker created a simple test that reproduced this error, fed it into an online ANSI/ISO C++ Standard test utility by Comeau computing, and it flagged it as an error. Has anyone compiled 1_34_1 using xlC/vacpp? If you need more information, or need the other errors, let me know. -- Marty Fried Left Coast, USA

1 0

boost-build-2.0-m11 and gcc-4.2.2
by Michael Albert 15 Oct '07

15 Oct '07

Greetings. If in "jam_src" I attempt to compile with gcc-4.2.2, the "bjam" executable compiles (using ./build.sh), but "bjam --version" fails on a segmentation fault. I can compile with gcc 3.4.6. Any thoughts? Anyone else seen this behavior? Thanks! -Mike

3 2

[ Boost-users] [preprocessor] - introduce mpl { ... } like asm { ... }
by Frank Bergemann 15 Oct '07

15 Oct '07

Hi, i posted this to comp.lang.c++.moderated: ------------------------------------------------------- Hi, i am just starting to work with "C++ Template Metaprogramming" - using the book of David Abrahams & Aleksey Gurtovoy. And you know, there are things like mpl::plus<> for some operator, mpl::if_<> for a control-struct, mpl:insert<>, mpl:erase<>, mpl:clear<> for STL container like operations, etc. Lots of things re-modeled the "unaccustomed template-syntax" - way for types, which are there in C/C++ language definition already for operating on variables and values. So i wonder if it would be a good option to introduce some mpl { ... } in remembrance of asm { ... } This time not for going down to the assembler level, but going up to the meta level. But with the goal to drop the template fashioned syntax for meta programming by a syntax already familiar in C/C++ language without templates (if this is possible - i don't now enough about that). Well, this is just a gut feeling, that there might be a chance this way to more align "standard" programming with "meta" programming, which then might enable to easier catch the the gray area of borderline between compilation and runtime execution and how to shift that back and forth. What do you think? rgds! Frank ------------------------------------------------------- Then i was thinking, that instead of extending the C++ language, it might be sufficient to use an own preprocessor before handing over to g++ (e.g.). For this maybe using boost::spirit? Finally i was remembering, that there is already some 'preprocessor' project in boost itself. But having a 1st look at it, it doesn't seem to do what i'd like to have. Just as an example - see this extract from boost::enable_if documentation: template <class T> typename enable_if_c<boost::is_arithmetic<T>::value, T>::type foo(T t) { return t; } I'd like to make it s.th. like: mpl { template <class T> if (is_arithmetic(T)) { T foo (T t) { return t;} } } This is just a simple example. But just imagine use of mpl metafunctions like mpl::and_ mpl::or_, mpl::plus, mpl::minus, mpl::equal_to, mpl::less, etc, that could come back to the "natural" &&, ||, +, -, ==, <, etc. [ hmm, i think, i mixed up mpl::enable_if with mpl::if_ - so my example is not correct - but i hope you get my point here ] Them maybe it would also be possible somehow: mpl { template <class T> if (is_arithmetic(T)) { T foo (T t) { return t;} } else { #error "instantiation of foo() not supported for given type T" } } ... to generate usefulcompiler warning/errors. The main point for me is to get find an equivalent to the "strange" template syntax that is more C/C++ alike way to deal with variables and values. Not because template syntax is "strange". But in order to re-align again. rgds! Frank

3 2

[mpl][fusion] making a heterogeneous collection of classes from a list of types.
by Michael Marcin 15 Oct '07

15 Oct '07

Hello, I have the MPL book and I've used some metaprogramming before but I'm always at the edge of my understanding of the language when I do it. I have a problem right now which I suspect is trivial but I can't for the life of me figure out how to do it. I have a list of 25 resource types (class Sound, class Script, etc.) and since they are not mutable I want to share them so I have a resource cache template that looks something like: template< typename Resource > class Cache { public: typedef typename Tag<Resource>::type resource_tag; /// Retrieve a resource by id /// Throws cache_error if the resource couldn't be found in the Repository Resource& Retrieve( ResourceId id ) { // is the resource in the cache? typename ResourceMap::iterator it = m_resources.find( id ); if( it != m_resources.end() ) { return it->second; } else { return LoadResource( id ); } } /// Clears the cache and destructs all cached resources. void Clear() { m_resources.clear(); } private: Resource& LoadResource( ResourceId id ) { Stream stream( Repository::Instance().Retrieve( id, resource_tag::type::value ) ); if( !stream.fail() ) { Resource* res = new Resource(stream); m_resources.insert( id, res ); return *res; } else { throw cache_error( id, resource_tag::type::value ); } } typedef boost::ptr_map< ResourceId, Resource > ResourceMap; ResourceMap m_resources; }; I have a class which must contain a Cache for each of the 25 resource types and support calls to get a resource of a particular type given an id and be able to call Clear on all 25 caches. Something like class Manager { public: /// Returns Cache<Resource>::Retrieve( id ) template< typename Resource > const Resource& Get( ResourceId id ); /// Calls Cache::Clear for all resource types void Clear(); }; It would be pretty easy to just copy and paste it out 25 times but I would like to reduce code duplication and simplify maintenance through metaprogramming. I have a private boost::mpl::list25 named resource_list containing all the resource types. I think I need to transform that in to a boost::fusion::map of 25 Resource to Cache<Resource> pairs but I'm not sure that is what I really need to do or how to go about doing that. Thanks, Michael Marcin

2 6

Header error in boost/python/object/make_holder?
by Michel Valstar 15 Oct '07

15 Oct '07

Hello, I am trying to build code of a recently graduated student, who is now unavailable to us (read: ignores us). When running scons in the appropriate directory, a lot gets build, until I get the following error: /usr/include/boost/python/object/make_holder.hpp:81: error: 'allocate' is not a member of 'boost::python::objects::value_holder<std::pair<const std::basic_string<char, std::char_traits<char>, std::allocator<char> >, State> >' /usr/include/boost/python/object/make_holder.hpp:83: error: no matching function for call to 'boost::python::objects::value_holder<std::pair<const std::basic_string<char, std::char_traits<char>, std::allocator<char> >, State> >::value_holder(int*&)' /usr/include/boost/python/object/value_holder.hpp:44: note: candidates are: boost::python::objects::value_holder<std::pair<const std::basic_string<char, std::char_traits<char>, std::allocator<char> >, State> >::value_holder() /usr/include/boost/python/object/value_holder.hpp:44: note: boost::python::objects::value_holder<std::pair<const std::basic_string<char, std::char_traits<char>, std::allocator<char> >, State> >::value_holder(const boost::python::objects::value_holder<std::pair<const std::basic_string<char, std::char_traits<char>, std::allocator<char> >, State> >&) scons: *** [trackerbase/trackerbasewrapper.os] Error 1 scons: building terminated because of errors. This does not seem to be an error in his own code, rather an error in the boost.python code? This is the first time I am using boost, and, as usual, I need to get this program working yesterday. My machine is a recently installed Ubuntu Feisty (hence the need for re-compilation). I am using the latest version of Boost (boost_1_34_1) and python 2.5.1 Any help would be appreciated. Thanks, MIchel -- Michel Valstar, PhD candidate http://www.doc.ic.ac.uk/~mvalstar Department of Computing Imperial College London 180 Queens' Gate London SW7 2AZ, U.K. tel.: +44 20 7594 8225 fax: +44 20 7581 8024

2 1

[serialization]-- serialize a std::vector<>?
by Rich Peters 15 Oct '07

15 Oct '07

Hello, sorry for the newbie questions... I have some sample serialization code where I serialize a container. if that container is std::list<sometype> it works fine. the same code with a std::vector<sometype> fails to compile. are vectors supported for serialization? I cant find any specific mention in the documentation or via searching. thanks

2 1

[ptr_map]const_iterator does not represent value_type
by Bill Buklis 15 Oct '07

15 Oct '07

In ptr_map, const_iterator does not represent value_type. Given this map: typedef boost::ptr_map<DWORD, int> BOOSTMAP; BOOSTMAP::value_type equates to: boost::ptr_container_detail::ref_pair<DWORD, int* const> BOOSTMAP::iterator does return this, but const_iterator returns: boost::ptr_container_detail::ref_pair<DWORD, const int* const> which is unlike std::map which represents const value_type for const_iterator. Since this is a different structure, it's incompatible with BOOSTMAP::value_type. The following code works using std::map: typedef std::map<DWORD, int*> STDMAP; int sample = 5; STDMAP sm; sm.insert( STDMAP::value_type(10,&sample) ); STDMAP::iterator sfirst = sm.begin(); STDMAP::const_iterator scfirst = sm.begin(); STDMAP::value_type& svt = *sfirst; const STDMAP::value_type& scvt = *scfirst; But, the same fails on the last line with boost::ptr_map typedef boost::ptr_map<DWORD, int> BOOSTMAP; BOOSTMAP bm; DWORD key = 10; bm.insert( key, new int(5) ); BOOSTMAP::iterator bfirst = bm.begin(); BOOSTMAP::const_iterator bcfirst = bm.begin(); BOOSTMAP::value_type& bvt = *bfirst; const BOOSTMAP::value_type& bcvt = *bcfirst; Also, it is not possible to bind to one of the members BOOSTMAP::value_type. For example, this works with std::map, but fails with ptr_map: boost::bind( &BOOSTMAP::value_type::first, _1 ); -- Bill --

1 0

[tools/regression] Directory reorganization
by Beman Dawes 15 Oct '07

15 Oct '07

There's some reorganization in progress in the trunk/tools/regression directory. This shouldn't affect most developers or users, since any of our Boost scripts that depend on these directory locations are being updated too. But if you've got any of your own scripts, make files, or IDE projects that depend on these directory locations, you will need to update them accordingly. Thanks to Rene Rivera for cleaning up this aspect of our tools directory structure! --Beman

1 0

Build Problem: bjam + boost + python
by Michael Bradley Jr 15 Oct '07

15 Oct '07

Hi, I've done a svn check_out of the boost repository. I'm trying to build it with the underlaying options, but can't achieve that goal due to the underlaying error messages. Any help or hints will be appreciated Cheers Michael ps A couple of weeks ago, i was able to compile boost with the same options without a problem. "bin/bjam.exe" -d 0 stage --stagedir=. --toolset=msvc link=sta tic runtime-link=shared architecture=x86 release debug address-model=32 warning: Graph library does not contain optional GraphML reader. note: to enable GraphML support, set EXPAT_INCLUDE and EXPAT_LIBPATH to the note: directories containing the Expat headers and libraries, respectively. warning: skipping optional Message Passing Interface (MPI) library. note: to enable MPI support, add "using mpi ;" to user-config.jam. note: to suppress this message, pass "--without-mpi" to bjam. note: otherwise, you can safely ignore this message. Building Boost.Regex with the optional Unicode/ICU support disabled. Please refer to the Boost.Regex documentation for more information (don't panic: this is a strictly optional feature). error: Unable to find file or target named error: '/python//python_for_extensions' error: referred from project at error: 'libs/python/build'

1 0

[boost-users][system] Mixing non-errors and errors (status_code vs. error_code)
by Darren Garvey 15 Oct '07

15 Oct '07

Hello, I've been wondering for a while if it's sensible to use Boost.System for transporting status_codes, rather than strictly error_codes. I'm aware than I *may* be trying to misuse the system library here, so I'll get straight to an example: If I had an HTTP library that retrieved remote pages, I might want to use, as (a very crude) example: template<typename MutableBufferSequence> void get_page(http::connection& conn, MutableBufferSequence& buf) { boost::system::error_code ec; conn.get_page(buf, ec); if (ec) throw ec; } The problem with relying on error_code to carry the errors is that the various success codes (eg. 200 OK, 201 ACCEPTED, 202 CREATED, etc.) seem to get lost, unless you class them as errors. Then you'd need something like: - if (ec) throw ec; + if (ec >= http::ok || ec <= http::multi_status) throw ec; It's not a massive pain, but having `ec == true` even when you don't have an error isn't helpful. AFAICT, any non-zero error value is an error that'll give `ec == true`. I feel like I'm missing something here, please tell me if I am. Regards, Darren PS. I'm trying to avoid having to make something that lies on top of boost::system::error_code, like a status_code class.

1 0