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

shared_from_this
by Dragan Milenkovic 14 Oct '03

14 Oct '03

Hello, people, I've been using enable_shared_from_this a LOT, mostly for implementing the Observer pattern (an object holds a list of weak pointers to it's observers). I wanted to know why are shared_from_this member functions public instead of being protected? Also, I've used the following expression a few times: shared_polymorphic_cast<This>(shared_from_this()) Any thoughts on putting it inside the enable_shared_from_this... shared_from_base<This>() or polymorphic_shared_from_this<This>() or such? Also, any comments on the designs I used? Better observers? Best regards, Dragan Milenkovic

2 1

Regex++ multiline matching
by Ben Hutchings 14 Oct '03

14 Oct '03

I see that Regex++ always treats ^ and $ as start and end of line rather than start and end of buffer. Perl does this only if the /m switch is used. Is there a simple way to switch between these two interpretations when compiling a regex?

2 1

boost::function compilation
by Terence Ou 14 Oct '03

14 Oct '03

Hi, I'm currently trying to use boost::function with VC++ 7.0 and I get the compilation error stating that function is not a member of boost. I've already included the header <boost/function.hpp> as well as the library path in the project settings but it just doesn't seem to be able to compile. The documentation specifies things about the preferred and portable syntax. Is there even some setting choose the desired syntax, becoz there are no hints or instructions that mention anything on it in the the documentation and I can't find anything in the source files as well. Thanks. Cheers, Terence

2 1

[BLL] bind(data_member_ptr,_1) functor won't accept temporaries as arguments
by Eelis van der Weegen 13 Oct '03

13 Oct '03

When Lambda's bind() is used to create a functor for dereferencing a stored data member pointer given an instance pointer, it seems it is not possible to pass the resulting functor a temporary instance pointer. For example: #include <boost/lambda/bind.hpp> #include <boost/lambda/lambda.hpp> struct S { int i; }; S * f (); void g (S * p) { using namespace boost::lambda; bind(&S::i, _1)(p); // ok bind(&S::i, _1)(p + 0); // not ok, "no match for call ..." bind(&S::i, _1)(f()); // not ok, "no match for call ..." } Does this restriction have a purpose? If not, is this something that needs to be (and can be) fixed? Regards, Eelis van der Weegen

2 1

enable_shared_from_this::_internal_weak_this initialization
by Cody Batt 13 Oct '03

13 Oct '03

Hello, It seems that classes which inherit from enable_shared_from_this are required to explicitly initialize _internal_weak_this. 1) Is that correct? If not, how is it initialized? 2) Is this detail mentioned anywhere in the documentation? (I couldn't find it.) Thanks, Cody

2 1

Stubbing for unit testing, what approach should I use
by Andrew Syrov 10 Oct '03

10 Oct '03

I am looking for set of techniques for isolation testing. As applications are designed with massive number of dependencies between classes and these dependencies end-up in external systems such as databases and OS, it's hard to extract function or part of class for efficient unit test. Basically I need an advice on what coding techniques should be used to allow flexible (generic?) stubbing: - function of a class may need to be tested in isolation from world (members could be global data for the purpose) - class may be tested with stubbed set functions; - data members could be stubbed, so no dependensies on classes which do not participate in test; Please advice. Andrew. _________________________________________________________________ Compare Cable, DSL or Satellite plans: As low as $29.95. https://broadband.msn.com

1 0

Enhancement to shared_pointer to support pointers to contained objects.
by Ricardo Capurro 09 Oct '03

09 Oct '03

Hello, We are really very happy and impressed with boost's smart_pointers, and we found the section "Smart Pointer Programming Techniques" (http://www.boost.org/libs/smart_ptr/sp_techniques.html) in boost smart_pointer documentation very useful, but we are having this problem: We have two classes X and Y, and each object of class X holds an object of class Y, and we need to implement a method in class X to get a shared_ptr<Y> so we can access Y instance inside X, and we need that X objects (that are too referenced by shared_ptr<X>) don't be deleted until all references to the Y object are destroyed. See the problem in this code: class Y {}; class X : public enable_shared_from_this<X> { Y y; public: shared_ptr<Y> ptr_y(); // Oops, how do we do here??? }; Our first approach was "Using a shared_ptr to hold a pointer to a statically allocated object" (http://www.boost.org/libs/smart_ptr/sp_techniques.html#static) but this doesn't work because X objects can be destroyed and we still have pointers to contained Y object, so we try a custom deletor like this: template <typename Z> class delegator_deletor { shared_ptr<Z> ptr; // Internal pointer to father X, wich contains Y public: delegator_deletor( shared_ptr<Z> const & ptrz ) : ptr( ptrz ) {} template <typename T> void operator()( T* ) { ptr.reset(); } // Null deletor, only resets pointer to father X object }; template <typename Z> delegator_deletor<Z> make_delegator_deletor( shared_ptr<Z> ptrz ) { return delegator_deletor<Z>( ptrz ); } class X : public enable_shared_from_this<X> { Y y; public: shared_ptr<Y> ptr_y() { return shared_ptr<Y>( &y, make_delegator_deletor( shared_from_this() ) ); } }; And now the problem was resolved but with one little drawback, we are not sharing the same counter between shared_ptr<X> and shared_ptr<Y>, so start looking boost's code to shange it in a way that it satisfies our needs, and we added this constructor to shared_ptr class: template<class Y,class Z> shared_ptr(Y* y, shared_ptr<Z> const & z): px(z.px ? y : 0), pn( y ? z.pn : detail::shared_count() ) // never throws { detail::sp_enable_shared_from_this(y, y, pn); } then we change X::ptr_y() this way: shared_ptr<Y> X::ptr_y() { return shared_ptr<Y>( &y, shared_from_this() ); } This new constructor instructs to create a shared_ptr<> that uses the same counter that the second's shared_ptr<> parameter, but point's to the first raw pointer paramenter. With this change everithing worked fine. Finally our question is: Is there an easyer way to solve our problem? If the answer is yes we want to know it and if the answer is no we want to know if you can enhace shared_ptr to solve this kind of problem. Thank you very much Ricardo #include <iostream> #include <string> #include <boost/smart_ptr.hpp> #include <boost/enable_shared_from_this.hpp> using namespace ::std; using namespace ::boost; #if !defined( SOLUTION1 ) && !defined( SOLUTION2 ) && !defined( SOLUTION3 ) && !defined( SOLUTION4 ) #define SOLUTION3 #endif #if defined( SOLUTION1 ) class null_deleter { public: template <typename T> void operator()( T* ) {} }; #elif defined( SOLUTION2 ) || defined( SOLUTION3 ) template <typename Z> class delegator_deletor { shared_ptr<Z> ptr; public: delegator_deletor( shared_ptr<Z> const & ptrz ) : ptr( ptrz ) { cerr << "delegator_deletor<" << typeid( Z ).name() << ">()" << endl; } template <typename T> void operator()( T* ) { ptr.reset(); cerr << "delegator_deletor<" << typeid( Z ).name() << ">()( " << typeid( T ).name() << "* )" << endl; } }; template <typename Z> delegator_deletor<Z> make_delegator_deletor( shared_ptr<Z> const & ptrz ) { return delegator_deletor<Z>( ptrz ); } #endif class Y : public enable_shared_from_this<Y> { Y( Y const & ); Y& operator=( Y const & ); public: Y() { cerr << "Y::Y()" << endl; } ~Y() { cerr << "Y::~Y()" << endl; } }; class X : public enable_shared_from_this<X> { Y y; int i; X( X const & ); X& operator=( X const & ); public: X() : y(), i( 7 ) { cerr << "X::X()" << endl; } ~X() { cerr << "X::~X()" << endl; } shared_ptr<Y> ptr_y() { #if defined( SOLUTION1 ) return shared_ptr<Y>( &y, null_deleter() ); #elif defined( SOLUTION2 ) return shared_ptr<Y>( &y, make_delegator_deletor( shared_from_this() ) ); #elif defined( SOLUTION3 ) shared_ptr<Y> py; try { py = y.shared_from_this(); } catch ( const bad_weak_ptr& ) { py.reset( &y, make_delegator_deletor( shared_from_this() ) ); } return py; #elif defined( SOLUTION4 ) return shared_ptr<Y>( &y, shared_from_this() ); #endif } shared_ptr<int> ptr_i() { #if defined( SOLUTION1 ) return shared_ptr<int>( &i, null_deleter() ); #elif defined( SOLUTION2 ) || defined( SOLUTION3 ) return shared_ptr<int>( &i, make_delegator_deletor( shared_from_this() ) ); #elif defined( SOLUTION4 ) return shared_ptr<int>( &i, shared_from_this() ); #endif } }; int main( const int, const char** ) { try { cerr << "(1)" << endl; shared_ptr<X> px( new X() ); cerr << "(2)" << endl; shared_ptr<Y> py = px->ptr_y(); cerr << "(3)" << endl; shared_ptr<int> pi = px->ptr_i(); cerr << "(4)" << endl; px.reset(); cerr << "(5)" << endl; py.reset(); cerr << "(6)" << endl; pi.reset(); cerr << "(7)" << endl; } catch ( const exception& e ) { cerr << "Exception caught: " << e.what() << endl; } catch ( ... ) { cerr << "Exception caught ..." << endl; } return 0; }

2 1

Re: [Boost-users] Re: Problem using Boost.thread : Class destructor calledseveral times!
by Huck finn 08 Oct '03

08 Oct '03

Ok, now I see that my object is being copied. But for the lines : FakeClass *item1 = new FakeClass(1);//,item2(2); group.create_thread(*item1); I get : FakeClass #1 says : I'm alive! FakeClass #1 says : Being Copied!! FakeClass #1 says : Being Copied!! FakeClass #1 says : Being Copied!! FakeClass #1 says : I'm being destroyed! FakeClass #1 says : I'm being destroyed! FakeClass #1 says : Being Copied!! FakeClass #1 says : I'm being destroyed! Thread for FakeClass #1 has been started! FakeClass #1 says : I'm being destroyed! main ending... The copy constructor is called 4 times!!! And it is called 3 times before the thread has even started! Is there a way to avoid that ? I was going to use the thread_group to manage the connection to a simple server applications, so that all the threads would hold info about the socket used for connection, etc... And that if the objectdestructor is called, the connection to the socket would be closed... With my object being destroyed 3 times before the thread has started, it gets a bit hard to manage that way... Any clue on how to use the threads so that it would work? thanks, Simon _________________________________________________________________ MSN Messenger : discutez en direct avec vos amis ! http://messenger.fr.msn.ca/

2 1

Re: compiler warnings...
by David Abrahams 08 Oct '03

08 Oct '03

David Ohlemacher <ohlemacher(a)bbn.com> writes: > Hi all from a boost newbie, > > Given: > Linux RH 8.0 > gcc 3.2 > BOOST_VERSION 103002 > > I created a boost app and it compiles and runs however I am getting > lots of warnings of the form > /usr/lib/boost/boost/iterator_adaptors.hpp:775:43: warning: > "BOOST_MSVC" is not defined. > > What am I not doing correctly? You're using -Wall ;-) We're currently discussing what to do about these warnings on the developers' list. Stay tuned, and until then turn off -Wall or live with the warnings. Sorry, -- Dave Abrahams Boost Consulting www.boost-consulting.com

3 2

compiler warnings...
by David Ohlemacher 08 Oct '03

08 Oct '03

Hi all from a boost newbie, Given: Linux RH 8.0 gcc 3.2 BOOST_VERSION 103002 I created a boost app and it compiles and runs however I am getting lots of warnings of the form /usr/lib/boost/boost/iterator_adaptors.hpp:775:43: warning: "BOOST_MSVC" is not defined. What am I not doing correctly? I have both included and not included <boost/config.hpp> and I read (http://www.boost.org/libs/config/config.htm#header) that I should not include this. Thanks, -d

1 0

mpl + enum + VC = error
by nagual 06 Oct '03

06 Oct '03

Hello, All! I write simple code: =============mpl_enum.cpp============= #include <boost/mpl/range_c.hpp> #include <boost/mpl/for_each.hpp> #include <iostream> enum test_e {ONE, TWO, THREE}; void f(test_e e) { const char* names[] = {"ONE", "TWO", "THREE"}; std::cout<<names[e]<<std::endl; } typedef boost::mpl::range_c< test_e, ONE, static_cast<test_e>(THREE + 1) > range; int main() { boost::mpl::for_each<range>(f); return 0; } =============mpl_enum.cpp============= MSVC 7 (ver 13.00.9466) =============Command Line============== c:\cl /EHsc /GR /Zc:wchar_t /Zc:forScope /nologo mpl_enum.cpp =============Command Line============== =============Output===================== mpl_enum.cpp C:\Lang\Projects\boost\boost\mpl\aux_\integral_wrapper.hpp(78) : error C2440: 'return' : cannot convert from '' to 'test_e' Conversion to enumeration type requires an explicit cast (static_cast, C-style cast or function-style cast) C:\Lang\Projects\boost\boost\mpl\aux_\integral_wrapper.hpp(78) : while compiling class-template member function 'boost::mpl::integral_c<T,N>::operator`T'(void) const' with [ T=test_e, N=ONE ] C:\Lang\Projects\boost\boost\mpl\aux_\range_c\iterator.hpp(34) : see reference to class template instantiation 'boost::mpl::integral_c<T,N>' being compiled with [ T=test_e, N=ONE ] C:\Lang\Projects\boost\boost\mpl\for_each.hpp(70) : see reference to class template instantiation 'boost::mpl::range_c_iterator<N>' being compiled with [ N=boost::mpl::range_c<test_e,ONE,3>::start ] C:\Lang\Projects\boost\boost\mpl\for_each.hpp(101) : see reference to function template instantiation 'void boost::mpl::aux::for_each_impl<done>::execute(Iterator *,last *,transform_op *,void (__cdecl *)(test_e))' being compiled with [ done=false, Iterator=first ] C:\Lang\Projects\boost\boost\mpl\for_each.hpp(111) : see reference to function template instantiation 'void boost::mpl::for_each(F,range *,boost::mpl::identity<T> *)' being compiled with [ F=void (__cdecl *)(test_e), T=boost::mpl::void_ ] mpl_enum.cpp(15) : see reference to function template instantiation 'void boost::mpl::for_each(F,range *)' being compiled with [ F=void (__cdecl *)(test_e) ] =============Output===================== However gcc v3.2.3, icl v 7.1, bcc32 v5.6.4 compile it. (bcc32 show 3 equivalent warning). How use it in MSVC 7?

2 1

remove_pointer
by Leo Landa 05 Oct '03

05 Oct '03

Good day to everyone, I am trying to find a portable enough implementation of remove_pointer (MSVC 6, MSVC 7.1, G++ 2.95.3, G++ 2.96). The "usual" template technique relies on partial template specialization, which MSVC 6.0 does not support. As far as I can tell, Boost's version of remove_pointer is a little bit of a "cheat" - i.e. it becomes predefined only for primitive types. Am I right about that? I'm a bit of a newb in these parts of London ;-) If this is true, then I'm screwed - and I can't convert a generic pointer type to the pointee type. If it is not true, then what is the technique? Leo. P.S. Just in case - I know how to figure out _whether_ the type is a pointer, and now just trying to strip it.

2 2

iterators
by Witz 05 Oct '03

05 Oct '03

Hi, I would like to raise a couple of points about the iterator adaptors. Firstly, is it really necessary to have the predicate as part of the template specification for the filter_iterator? Surely it can be assumed that it will always be convertible to boost::function<bool (const value_type&)> ? Secondly, by allowing the return type of the function used by the transform_iterator to be a reference can we not use it to implement a more generic form of indirect_iterator? The code below gives a flavour of the idea. struct dereference { typedef int result_type; int* data_; dereference(int* data):data_(data){} int operator()(int i) const { return data_[i]; } }; int main() { int data[] = { 1,2,3,4,5,6,7,8 }; int keys[] = { 0,3,1,6,5,4,2,7 }; typedef boost::transform_iterator_generator<dereference, int*>::type indirect_iterator; indirect_iterator i(keys, dereference(data)), i_end(keys+8,dereference(data)); while(i != i_end) std::cout << *i++ << ' '; return 0; } The output is: 1 4 2 7 6 5 3 8 Witz

3 2

boost library build consistency
by Thomas S. Urban 03 Oct '03

03 Oct '03

Is there any definition of how libraries are supposed to build? I've integrated the building of required boost libraries into the build system for our projects. There seems to be some inconsistency in what gets built and where it gets put. In particular, for the thread and regex libs, boost builds both shared and static versions of the libraries and puts them in a bin-stage subdirectory under the library build directory. However, for the filesystem library, only the static library gets built and the bin-stage/ directory is not used. Is there a standard behavior library developers are supposed to follow in this regard? Can I count on the bin-stage/ as a short-named location for the most recent build of any library? Are the boost or boost_filestsystem maintainers interested in the simple fix to the Jamfile to make that library behave like the other two I mentioned? Thanks Scott Urban

3 3

parse error in format library
by Alan Stewart 03 Oct '03

03 Oct '03

I've been successfully using the tokenizer library for some time, and am very grateful for it. The format library looks very userful - I need to convert a bunch of old printf's to ostreams. However, in attempting to compile with gcc 2.96 I'm finding that there are many substitutions that need to be made, such as replacing "char_traits" with "basic_string". I'm using Boost 1.30.0 in production, but have also tried 1.30.2 with the same results. I see that the format library headers include the header file "boost/config.hpp". Is it likely that I have some sort of configuration error? Where do I find information on toolset compatabiliy for specific Boost libraries, since this one does not seem to be compatible with mine? -- D. Alan Stewart Senior Software Developer Layton Graphics, Inc. 155 Woolco Drive Marietta, GA 30065 Voice: 770/973-4312 Fax: 800/367-8192 http://www.layton-graphics.com

2 1

Problem using Boost.thread : Class destructor called several times!
by Huck finn 02 Oct '03

02 Oct '03

Hi there, still a newbie question for you guys... I changed a bit the sample example from thread.cpp : here's the new code i'm trying: ----------------- boost::mutex io_mutex; // The iostreams are not guaranteed to be thread-safe! class FakeClass { public: FakeClass(int NewID):ID(NewID) { boost::mutex::scoped_lock scoped_lock(io_mutex); std::cout<<"FakeClass #"<<ID<<" says : "<<"I'm alive!"<<std::endl; } ~FakeClass() { boost::mutex::scoped_lock scoped_lock(io_mutex); std::cout<<"FakeClass #"<<ID<<" says : "<<"I'm being destroyed!"<<std::endl; } void operator()() { boost::mutex::scoped_lock scoped_lock(io_mutex); std::cout<<"Thread for FakeClass #"<<ID<<" has been started!"<<std::endl; } private: int ID; }; void main () { boost::thread_group group; FakeClass item1(1),item2(2); group.create_thread(item1); group.create_thread(item2); group.join_all(); std::cout<<" main ending..."<<std::endl; } ----------------------------- here's what I get when executing : FakeClass #1 says : I'm alive! FakeClass #2 says : I'm alive! FakeClass #1 says : I'm being destroyed! FakeClass #1 says : I'm being destroyed! FakeClass #1 says : I'm being destroyed! Thread for FakeClass #1 has been started! FakeClass #2 says : I'm being destroyed! FakeClass #1 says : I'm being destroyed! FakeClass #2 says : I'm being destroyed! FakeClass #2 says : I'm being destroyed! Thread for FakeClass #2 has been started! FakeClass #2 says : I'm being destroyed! main ending... FakeClass #2 says : I'm being destroyed! FakeClass #1 says : I'm being destroyed! Press any key to continue 1- I'm wondering why the destructor of my FakeClass is called 4 times for each items ??? 2- It says in the doc that the function that creates a thread "copies the function wrapped by threadfunc" and threadfunc would be in this case item1 and item2, does this means that the copy constructor is called every time a thread is created with an object? Thanks all! Simon _________________________________________________________________ MSN Messenger : discutez en direct avec vos amis ! http://messenger.fr.msn.ca/

3 2

Problem with is_class and is_enum
by Romain Lerallut 02 Oct '03

02 Oct '03

In the following code, struct foo2 is mistakenly taken for an enum, which breaks is_class, which leads to some unfortunate problems in boost::python. Problem appears on both Boost 1.30.0 and a one-month old CVS checkout, using gcc 3.3.2 and 2.95 and Intel icc 7.1 for linux. I have tried specializing 'is_enum', setting it by hand to return 'false', but though it works for this simple case, it does no good when the call to is_enum is buried deep inside boost. I rely heavily on being able to do these implicit conversions for foo2, so if anyone has an idea on how to make this work, it would be great. Thanks in advance, Romain ---------------------------------------------------------------- #include <boost/type_traits/is_enum.hpp> #include <iostream> struct foo1 // this one is not detected as an enum { operator int() { return 0; } }; struct foo2 // this one *is* detected as an enum { template<typename T> operator T(){return 0;} }; int main() { std::cout<<boost::is_enum<int>::value<<"\n"; std::cout<<boost::is_enum<foo1>::value<<"\n"; std::cout<<boost::is_enum<foo2>::value<<"\n"; } -------------------------------------------------------------- Expected Result: 0 0 0 Actual Result: 0 0 1

2 1

compilation on Win64 platform with MSVC7.1 for 64 bit
by bart.tanghe＠skynet.be 02 Oct '03

02 Oct '03

Hi, we're using Boost 1.30.0 with MSVC7.1 on a Win32 platform. Currently where working on a port to a Win64 platform using the MSVC7.1 for IA64 compiler. (13.10.2240.8 for IA64) I just compiled the code as is and unfortunately I encountered errors in boost headers. This probably is because of missing/bad defines for this 'unknown' platform in the config headers... I was wondering if somebody already tried to compile boost on a 64 bit platform, and if he/she succeeded? Or if there are known implementation details which would cause boost 'never' to work on 64 bit platforms? If there are no known 'showstoppers' then I suppose I'll have to go through the Boost Configuration Reference or should I look somewhere else? Thanks, Bart

2 1

Compile Boost.thread libraries has a static library (.lib) with msvc 6 ?
by Huck finn 01 Oct '03

01 Oct '03

Hi everyone, I'm a total newbie to boost, but I've got to developp a project that could use some of the powerful boost libraries, so I'm trying to make a move in that way... Ok, most of the boost libraries I can build in a static way and create (.lib) files with Microsoft Visual Studio 6.0 for my partners to use . But I am having a problem doing the same thing with the Boost.Thread libraries... First question : Is it possible to build this librairy in a (.lib) file with msvc 6.0? I'm getting tons of warnings concerning "dllexport" problems, but the (.lib) are still created. I was wondering if someone could help me understanding/fixing the warning problems : warning C4275: non dll-interface class 'boost::noncopyable' used as base for dll-interface class 'boost::detail::condition_impl' c:\wrk\common\boost-1.30.2\boost\noncopyable.hpp(22) : see declaration of 'noncopyable' warning C4273: 'thread::thread' : inconsistent dll linkage. dllexport assumed. I really need to avoid using .dll in this project, so builing this static library is very important to me... Thanks in advance, Simon _________________________________________________________________ MSN Search, le moteur de recherche qui pense comme vous ! http://fr.ca.search.msn.com/

5 7

bind and Constructor invocation?
by Jeff Flinn 30 Sep '03

30 Sep '03

I don't see any way to use bind to directly construct an object, other than calling a function that returns an object. Or am I missing something? Like something similar to Spirit::Phoenix construct_<>? This would avoid manually creating these factory functions. -- Jeff Flinn Applied Dynamics, International

1 0

BGL runtime crash with g++ -O2 optimizer.
by Franck.Caron＠infineon.com 29 Sep '03

29 Sep '03

Hi all, I have a problem while compiling BGL with gcc 3.1.1 on solaris7 with the -O2 optimizer option. All is ok with -g (debug) but -O2 produce a crash at runtime Here is a code that reproduce the problem : #include <boost/graph/adjacency_list.hpp> #include <boost/graph/properties.hpp> #include <iostream> #include <string> using namespace std; using namespace boost; struct VertexPropertyTag { typedef vertex_property_tag kind; }; typedef property<VertexPropertyTag,string> VertexProp; typedef adjacency_list<setS,vecS,bidirectionalS,VertexProp> Graph; int main(){ Graph graph; graph_traits<Graph>::vertex_descriptor v1 = add_vertex(string("s1"),graph); graph_traits<Graph>::vertex_descriptor v2 = add_vertex(string("s2"),graph); add_edge(v1,v2,graph); graph_traits<Graph>::edge_iterator eIt,eItEnd; tie(eIt,eItEnd) = edges(graph); for(;eIt != eItEnd;++eIt){ cout << "->" << *eIt << endl; } return 0; } With g++ -Wall -pedantic -fPIC -g -c -o solaris7/gcc3.1/Main.o Main.cc -I. -I/opt/boost_1.30.2 g++ -fPIC -o solaris7/gcc3.1/testbug.exe solaris7/gcc3.1/Main.o -L /opt/boost_1.30.2/libs/gcc3.1/release/single-thread -R -Wl,-Bdynamic Here is the result : shell> solaris7/gcc3.1/testbug.exe ->(0,1) It is OK With g++ -Wall -pedantic -fPIC -O2 -c -o solaris7/gcc3.1/Main.o Main.cc -I. -I/opt/boost_1.30.2 g++ -fPIC -o solaris7/gcc3.1/testbug.exe solaris7/gcc3.1/Main.o -L /opt/boost_1.30.2/libs/gcc3.1/release/single-thread -R -Wl,-Bdynamic Here is the result : shell> solaris7/gcc3.1/testbug.exe ->(0,1) ->(0,0) Bus error (core dumped) Does anybody already had this ? Is it a bug in my code, in BGL or in gcc ? Thanks for the help. Franck

1 0

Re: [Boost-users] Trying to get boost threads working under visual .NET
by John Maddock 29 Sep '03

29 Sep '03

> I'm trying to get the Boost thread library working under Microsoft .Net, > here is what I've done and the problems I'm seeing. > > I am currnetly using the regex lib and the date/time classes, which are > working. I now want to start creating threads so I follow the given > example, and started to get linker errors for xtime_get and > boost::thread::sleep. So I look around, and do the bjam -sTOOLS=msvc in > the appropriate threads directory and get no errors. It still won't > link. > I notice that the only boost*.lib files in my vc7\lib dir were from > boost_regexp, so I copy 4 files boost_thread(d).dll/lib into that > directory. I then explicitely add boost_threadd.lib into the project's > link. Ok, so now my application compiles AND links, but it wants > boost_threadd.dll when it executes. So I copy boost_threadd.dll into the > projects Debug directory wehre the .exe compiles to. Now when I run it, I > get a dialog box saying the application failed because > "boost_regex_vc7_mdid.dll" can not be loaded! the regex stuff was working > FINE and I don't think I did anyhting to change it. > > So can someone shed some light on this for me, please ? Why does it now > want the boost_regex dll? Thanks for any input.. if I've missed some > crucial documentation somewhere, sorry. If you changed your project build options from using one of the static C runtimes to using the dll runtime then that would be the problem, define BOOST_REGEX_STATIC_LINK when building if you want to static link to boost.regex. John. --------------------------------- Want to chat instantly with your online friends? Get the FREE Yahoo!Messenger

1 0

regex and stream processing
by John Maddock 29 Sep '03

29 Sep '03

Message forwarded from Dave Gomboc: > We have a stream of text data (actually it is a chunked > response of a webserver, but it might be a file or > something...) The data size may be very large and we need to > do some find/replace in this stream using regular > expressions. It is not possible to load entire stream into > the memory. I'm not even hope that Regex++ could work with > such kind of streams, but maybe someone can suggest me a > solution to this problem? This is possible without much trouble _if_ the text strings you are trying to match will fit into main memory, even if all the data won't. Use a std::list as a buffer, so that deleting and inserting is inexpensive, and you don't have to perform copies and update iterators when reading in more input. Fill your (sliding window) buffer, then regex it, using the partial match feature whenever you have not already detected the end of the file. If a match occurs, any characters occuring before the match can be processed (serialized to disk, or whatever) and removed from the buffer. Repeat this looking for a match until you reach the end of current input. You'll either have "no match" or "partial match" at this point: remove non-matching characters (that could be all of them) at this point, then re-fill your buffer with more data and keep going. If you're just doing one substitution, that's fine, but if more, there are a few things to keep in mind. I have done over 60 distinct substitutions at a time on a stream of 0.1 gigabytes using this technique, and it works quite quickly. (I used a buffer that held about 1 million characters (2^20), but I also tested with buffers as small as 256 characters -- the strings I was matching were not large). Just read in data, process all the substitutions in order, write the data out and repeat. If you're doing multiple substitutions, you (of course) can't flush any data from memory until you've tried all the substitutions on the current region of interest. Also, for each later substituion, the current region of interest shrinks because you have to acknowledge partial matches from earlier subsitutions. For example, say that there is no partial match at the end of substituions 1 through 6 (there might have been full matches where substitution was performed, that's okay.) But substituion 7 has a partial match at the last five characters. Now, your region of interest for substitution 8 is five characters smaller, because you don't want to process those last five characters because substitution 7 might change them once you've read in more input. The order of your substituions is important if the outputs of some desired subsitutions overlap with the inputs of others. If you are not grokking my meaning, I can fish around for the source code. Perhaps it's easier to explain in code than in words, my code for the entire process is maybe 10 or 15 lines. An alternative is to try to do some streambuf hacking, but the above method was simpler and adequate for my purposes. Dave --------------------------------- Want to chat instantly with your online friends? Get the FREE Yahoo!Messenger

1 0

regex problem
by John Maddock 29 Sep '03

29 Sep '03

> The last assert always fails, because regex::match insists on including the > space after the ":" in the second match result, which seems wrong to me. > I've tried giving various different combinations of flags to the regex ctor, > although from what I can tell from the documentation the default flags > should work. This is with VC7 and boost 1.30.2. For comparison, perl > (5.6.1) does not seem to have this problem, as the following script > illustrates: > > #!/usr/bin/perl > print "matches: <$1> <$2>\n" if ('foo: bar' =~ /([\w]+):[\s]*(.*)/); > > Is there any way to get the behavior I want? This is the difference between perl and POSIX leftmost-longest matching rules showing itself (the next release will default to perl rules and do what you want, but of course that doesn't help you now), basically non-marked sections are neither greedy nor non-greedy under POSIX rules, so regex finds the match it does because it makes $2 further to the left than the alternative (without effecting the overall match). You can get the effect you want by making your expression more "precise": "([\\w]+):[\\s]*([^\\s].*)" for example, John. --------------------------------- Want to chat instantly with your online friends? Get the FREE Yahoo!Messenger

1 0

Re: Implementation of my first template class
by Zenaan Harkness 29 Sep '03

29 Sep '03

Take 3: /** * A class providing a combination of an input stream * and an input iterator. * I use for convenience when processing a part-ascii, * part-binary stream. */ template<typename charT, typename traits = std::char_traits<charT> > class basic_istringtraverser : public std::basic_istringstream<charT, traits> { public: basic_istringtraverser(std::basic_string<charT> s) : std::basic_istringstream<charT, traits>(s) {} charT operator*() {return (charT)(rdbuf()->sgetc());} basic_istringtraverser& operator++() { rdbuf()->sbumpc(); return *this; } }; /** Specialization for char type. */ typedef basic_istringtraverser<char> istringtraverser;

1 0