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

default copy constructors
by bringiton bringiton 28 Jul '06

28 Jul '06

I have a specific c++ question, and was wondering whether someone could offer some advice. Is is safe to assume that modern compilers will supply a default copy constructor and default assignment operator that will copy all attributes? Or may I run into undefined behaviour between compilers. class Test: public: int v1; char v2; float v3 std::string v4; std::vector<int> v5; boost::shared_ptr<int> v6; }; ie is it safe to assume that Test's copy constructor will call the copy constructor of all the above attributes. NOTE: i am only talking about attributes that are already copiable. i understand that if a class manages it's own memory, the copy constructor will need to be defined. **Just so we are using the same terminology, does c++ have a pattern to define concrete objects that are copiable? ie your basic types, int, char, float. and classes that have been designed to be copiable: ie std::string, std::vector etc. **Maybe it's called concrete type.

1 0

Re: [Boost-users] [license] Why is the Boost license not anofficial Open Source license?
by Sohail Somani 27 Jul '06

27 Jul '06

> -----Original Message----- > From: boost-users-bounces(a)lists.boost.org > [mailto:boost-users-bounces@lists.boost.org] On Behalf Of > Gennaro Prota > Sent: Thursday, July 27, 2006 10:05 AM > To: boost-users(a)lists.boost.org > Subject: Re: [Boost-users] [license] Why is the Boost license > not anofficial Open Source license? > > With a terminology stress which (to me) twists the meaning of "free > software" it could be classified as "non-copylefted free software" > (which basically means, anyone can redistribute it under a non free > license; in the case of boost they can do that only if they don't > redistribute the source code) So if you use boost in your interfaces, you've gotta redistribute your source code too?

3 2

Re: [Boost-users] [license] Why is the Boost license not anofficial Open Source license?
by Sohail Somani 27 Jul '06

27 Jul '06

> -----Original Message----- > From: boost-users-bounces(a)lists.boost.org > [mailto:boost-users-bounces@lists.boost.org] On Behalf Of > Sebastian Redl > Sohail Somani wrote: > > >So if you use boost in your interfaces, you've gotta > redistribute your > >source code too? > > > > > No. All he said is that you may not distribute the Boost > sources under > any license but the Boost license. It doesn't affect your own > stuff at all. Gotcha.

1 0

converting a string to an enum
by Elisabeth Benoit 27 Jul '06

27 Jul '06

Hi, Does boost offer some kind of functionality allowing to transform a string into an enum? thanks.

2 1

using boost
by Fred J. 27 Jul '06

27 Jul '06

Hi could some one point to a link how how to link/compile with boost? or tell me whats missing from the makefile below? thanks alot #include <converter_policies.hpp> using boost::floor; ... cout << floor(12.45) << endl; ... the make file; OBJS := $(patsubst %.cpp,%.o,$(wildcard *.cpp)) COMP = g++ proj: $(OBJS) $(COMP) -Wall -o proj $(OBJS) #-Wall turns on all warnings clean: rm -rf *.o proj --------------------------------- Do you Yahoo!? Next-gen email? Have it all with the all-new Yahoo! Mail Beta.

2 1

[Boost.Python/Bjam] How to wrap a binary only dll ?
by michael_gruner＠arcor.de 27 Jul '06

27 Jul '06

Hello, I'm new to boost and did sucessfully complete the installation of boost and the boost.python hello world example. However, I'm wondering how to wrap a shared library I get as a binary and a header file to access it. Can anyone point me to any documentation about this? little example: I own a library "mylib.dll" and a header file "mylib.h" without any source code for my platform. Now I'd like to build a python wrapper to access this mylib.dll. I already wrote a "mywrapper.cpp" that includes the mylib.h file. But however bjam tries to build a mylib.pyd instead of using just the dll. can anyone help me with this? thank you Michael

2 2

[regex] boost::match_extra vc 8.0 express access violation
by Ovanes Markarian 27 Jul '06

27 Jul '06

Hi! I am trying to match a function body and find out if the params or return type have specific class name patterns. Therefore I did a really huge regex... sorry if is not so readable, but I used defines to make as little errors as possible... //non marking group #define NMG_S(s) "(?:"+s+")" #define NMG_S_0N(s) NMG_S(s)"*" #define NMG_S_1N(s) NMG_S(s)"+" #define NMG_S_01(s) NMG_S(s)"?" #define NMG(s) "(?:"s")" #define NMG_0N(s) NMG(s)"*" #define NMG_1N(s) NMG(s)"+" #define NMG_01(s) NMG(s)"?" #define RE_OR "|" #define RE_SPACE "[[:space:]]" #define RE_SPACE_01 RE_SPACE "?" #define RE_SPACE_0N RE_SPACE "*" #define RE_SPACE_1N RE_SPACE "+" #define RE_START_WORD "\\<" #define RE_END_WORD "\\>" #define RE_ID "[[:alpha:]]" NMG_0N("[_a-zA-Z0-9]") #define RE_CONST "const" #define RE_VOLATILE "volatile" #define RE_ANY "." #define RE_ANY_0N ".*" #define RE_ANY_1N ".+" #define RE_ANY_01 ".?" #define RE_PTR "\\*" #define RE_PTR_0N RE_PTR "*" #define RE_PTR_1N RE_PTR "+" #define RE_PTR_01 RE_PTR "?" #define RE_REF "&" #define RE_REF_0N RE_REF "*" #define RE_REF_1N RE_REF "+" #define RE_REF_01 RE_REF "?" #define RE_LINE_START "^" #define ONE_OF_PROTOCOLS NMG_S(strings::Name1) RE_OR NMG_S(strings::Name2) RE_OR NMG_S(strings::Name3) RE_OR \ NMG_S(strings::Name4) RE_OR NMG_S(strings::Name5) RE_OR NMG_S(strings::Name6) RE_OR \ NMG_S(strings::Name7) RE_OR NMG_S(strings::Name8) RE_OR NMG_S(strings::Name9) RE_OR \ NMG_S(strings::Name10) #define MATCH_PROTOCOL RE_START_WORD \ NMG_01(RE_ANY_0N "(" ONE_OF_PROTOCOLS ")" RE_ANY_0N) RE_OR \ NMG_01(RE_ANY_1N) \ RE_END_WORD RE_SPACE_0N Name1 to Name10 are static const std::string definitions within a class. A have the following regex defined to match params and function return type const boost::regex strings::regex_fct ( //possibly leading whitespace or const RE_LINE_START RE_SPACE_0N NMG_01(RE_CONST RE_SPACE_1N) //result type MATCH_PROTOCOL //return type can be a pointer or reference with const qualifier NMG_01(RE_CONST RE_SPACE_0N) RE_PTR_0N RE_SPACE_0N RE_REF_01 RE_SPACE_0N NMG_01(RE_CONST) RE_SPACE_1N //function name RE_START_WORD NMG_01(RE_ID) RE_END_WORD RE_SPACE_0N //function params "\$" //open function params brace //param type (can be repeated 0 to N times) NMG_0N( RE_SPACE_0N MATCH_PROTOCOL RE_SPACE_0N //may be a ptr or ref type NMG_01(RE_CONST RE_SPACE_0N) RE_PTR_0N RE_SPACE_0N RE_REF_01 RE_SPACE_0N NMG_01(RE_CONST) RE_SPACE_1N //(?: still open //param name and default initializations are not important RE_ANY_0N //params can be comma separated in case of more then 1 param NMG_01(",") //close NMG_0N ) //terminate in "\$" ); Now I have the following test string: using namespace std; string search_where_fct1("CName1_Setup const* someFct(){}"); set<string> uniques_names; find_protocols(search_where_fct1, strings::regex_fct, uniques_names); void find_protocols(const std::string& str, const boost::regex& expression, std::set<std::string>& log) { boost::smatch what; IdentifyProtocolUsage usage(log); if(boost::regex_match(str, what, expression, boost::match_extra)) //<<<<< crash!!!! { usage(what); } } This is the boost::regex function where it crashes: template <class BidiIterator, class Allocator, class traits> bool perl_matcher<BidiIterator, Allocator, traits>::match_imp() { // initialise our stack if we are non-recursive: #ifdef BOOST_REGEX_NON_RECURSIVE save_state_init init(&m_stack_base, &m_backup_state); used_block_count = BOOST_REGEX_MAX_BLOCKS; #if !defined(BOOST_NO_EXCEPTIONS) try{ #endif #endif // reset our state machine: position = base; search_base = base; state_count = 0; m_match_flags |= regex_constants::match_all; //LINE 162 perl_matcher_common.hpp !!! m_presult->set_size((m_match_flags & match_nosubs) ? 1 : re.mark_count(), search_base, last); m_presult->set_base(base); if(m_match_flags & match_posix) m_result = *m_presult; verify_options(re.flags(), m_match_flags); if(0 == match_prefix()) return false; return m_result[0].second == last; #if defined(BOOST_REGEX_NON_RECURSIVE) && !defined(BOOST_NO_EXCEPTIONS) } catch(...) { // unwind all pushed states, apart from anything else this // ensures that all the states are correctly destructed // not just the memory freed. while(unwind(true)){} throw; } #endif } Call from this line causes access violation: Unhandled exception at 0x104817fd in some_app.exe: 0xC0000005: Access violation writing location 0x56394ab7. I re-built regex with BOOST_REGEX_MATCH_EXTRA uncommented in user.hpp. Would be really nice to receive some comments or suggestions. With Kind Regards, Ovanes Markarian

2 1

[Boost compilation on SGI] Problems compiling boost on SGI
by Sebastien Fortier 27 Jul '06

27 Jul '06

Can anyone help with this, I have been trying to compile boost on an SGI machine. Boost source taken from cvs. IRIX version: IRIX Release 6.5 IP35 MIPSpro compiler version: MIPSpro Compilers: Version 7.4.2m attached is the log of my last compilation effort. Here is a short sorted list of the reoccurring errors ----------------------------------------------------------------------------------------------------------------------------------- cc-1004 CC: ERROR File = /data/cmdwx4/pst/package/boost_cvs/boost/type_traits/is_arithmetic.hpp, Line = 37 Not enough memory is available. BOOST_TT_AUX_BOOL_TRAIT_DEF1(is_arithmetic,T,::boost::detail::is_arithmetic_impl<T>::value) ^ 1 catastrophic error detected in the compilation of [...] ----------------------------------------------------------------------------------------------------------------------------------- cc-1004 CC: ERROR File = /data/cmdwx4/pst/package/boost_cvs/boost/type_traits/is_same.hpp, Line = 37 Not enough memory is available. BOOST_TT_AUX_BOOL_TRAIT_DEF2(is_same,T,U,false) ^ 1 catastrophic error detected in the compilation of [...] ----------------------------------------------------------------------------------------------------------------------------------- cc-1282 CC: ERROR File = /data/cmdwx4/pst/package/boost_cvs/boost/test/utils/named_params.hpp, Line = 307 More than one instance of overloaded function "boost::nfp::optionally_assign" matches the argument list. Function symbol function template "boost::nfp::optionally_assign(T &, boost::nfp::nfp_detail::nil)" is ambiguous by inheritance. Function symbol function template "boost::nfp::optionally_assign(T &, const Source &)" is ambiguous by inheritance. The argument types are: (std::size_t, boost::nfp::nfp_detail::nil ). optionally_assign( target, p[k] ); ^ ----------------------------------------------------------------------------------------------------------------------------------- cc-1384 CC: ERROR File = /data/cmdwx4/pst/package/boost_cvs/boost/test/impl/unit_test_suite.ipp, Line = 124 Protected member "boost::unit_test::class_property<boost::unit_test::test_unit_id>::va lue" is not accessible through a "boost::unit_test::readwrite_property<boost::unit_test::counter_t>" pointer or object. tu->p_expected_failures.value = expected_failures; ^ ----------------------------------------------------------------------------------------------------------------------------------- cc-1384 CC: ERROR File = /data/cmdwx4/pst/package/boost_cvs/boost/test/impl/unit_test_suite.ipp, Line = 126 Protected member "boost::unit_test::class_property<boost::unit_test::test_unit_id>::va lue" is not accessible through a "boost::unit_test::readwrite_property<boost::unit_test::counter_t>" pointer or object. p_expected_failures.value += tu->p_expected_failures; ^ These errors reoccur throughout the compilation process, is there a fix for this ? the skipped libraries are libboost_test_exec_monitor libboost_unit_test_framework libboost_wave I already compiled the library under Linux and everything worked fine. I was hoping to use the unit testing library from boost instead of cppunit but if I cant compile it I will have no choice in the matter. -- Sébastien Fortier

1 0

Re: [Boost-users] [boost] BGL - cycles in an undirected graph
by Aaron Windsor 27 Jul '06

27 Jul '06

On 7/20/06, Michal <simecek(a)gmx.de> wrote: > my task is to find cycles in an undirected graph and to assign a cycle id > to each > vertex in this cycle. It is still a bit more complicated, because I do not > need just the simple cycles, but all simple cycles which are sharing > edges, must > get the same cycle id. > > I think this problem is already solved somewhere, but I cann not find the > right > approach, thus I am asking here if somebody have an idea how to solve it > best > with BGL. Michal, If you want to label the edges of a graph such that (1) any two edges that are on a simple cycle together get the same label, and (2) any two edges not on any cycle together get distinct labels then you simply want to find the biconnected components of the graph - there's a way to do this in linear time using depth-first search, and the algorithm is in the BGL: http://www.boost.org/libs/graph/doc/biconnected_components.html However, the fact that you said that you want to label all vertices in the graph with a distinct label subject to condition (1) above is a little confusing - you can have two cycles that don't meet in any edges but share a common vertex, and in that case, which of the two labels would you want to use for that vertex? Aaron

2 1

[mpl] A partioning unique/non-unique algorithm
by Eric Lemings 27 Jul '06

27 Jul '06

Greetings, How would you combine the boost::mpl::partition and boost::mpl::unique algorithms such that an element that is not placed into the first inserter (for unique elements) is placed into the second inserter (for non-unique elements)? Thanks, Eric.

2 1