February 2008 - Boost-users

Using boost::barrier dynamically
by Gonzalo Garramuño 22 Feb '08

22 Feb '08

I'm trying to have several threads synced with a boost::barrier. But I'm finding its design very limiting (or, more likely, my understanding of its use is too poor). I want to have a barrier created dynamically (based on some parameters that determine how many threads run). And I also want to be able to delete the barrier at any point in time (from a thread that does not wait on the barrier). I expect this to force all threads that are waiting on the barrier to continue (ie. send a notify_all). However, boost::barrier's destructor just does nothing, so I get a hang. What's the right way to achieve this? PS. On a separate note... is there any reason why all of boost primitives have their variables private? I really would find it pretty useful to sometimes be able to query, for example, the lock count of some objects. -- Gonzalo Garramuño ggarra(a)advancedsl.com.ar AMD4400 - ASUS48N-E GeForce7300GT Xubuntu Gutsy

1 0

Re: [Boost-users] All orderings of boost::graph::topological_sort()"
by Matan Nassau 22 Feb '08

22 Feb '08

Thank you very much for your prompt reply, Stephen. On Fri, Feb 22, 2008 at 12:51, Stephen Nuchia wrote: > For any significant-sized input the number of orderings will be too > large to do anything useful with. At the moment my input consists of about 32 vertices, although I'd like not to count on it to remain this way. I can probably assume the number of nodes will remain in the order of magnitude of hundreds or less. I'm talking about a schedule generating program for an undergraduate student. It generates a schedule for a semester for a student, given some constraints (time of day, courses done already and so on) and the course dependency (this course must not be taken before that one and so on). I would very much like to support other complexities such as some courses may be taken while a certain dependent is being taken, while others must be taken before any of their dependents may be taken. But I leave such complexities as worries for future versions. For the time being, all I want is to be able to generate all the different schedules (i.e., all various topo sorts). Indeed, something like std::next_permutation() would make a lot of sense. It would be fast and flexible. My next task would then be to find an algorithm to quickly estimate the number of different orderings. :) > For an input small enough that the > set of feasible orderings will fit in memory I would simply generate all > permutations and screen for feasibility rather than attempting to create > an enumeration algorithm. I don't follow. Can I change the ordering of vertices without ruining the graph's integrity? How? Would this affect the toposort output? > Create a comparability matrix and use it to > prune the permutation tree, perhaps? N^2 bits doesn't sound > prohibitive. I don't follow... > This has probably been studied but I'm at work so I'll speculate. For > larger inputs with many constraints it may be worthwhile to create an > enumeration algorithm. > > For purposes of empirical testing of build-ordering systems it might be > sufficient to go into the algorithm and randomize the selection > processes whenever it makes arbitrary choices. This may or may not be > easily achieved by randomly pre-sorting the input, depending on the > internals of the algorithm. If the sort is stable that will work. That is the reason for my query on my OP in the first place. I don't quite understand how manipulate the input vertices of the depth_first_search(). For what I understand, depth_first_search() works on a set of start vertices (which have no incoming edges), and for each of these the algorith dives in to the depth of the vertex's tree. Permuting the order of these vertices would give me advance me a lot, but I don't know how to do that. > If there is enough interest to justify adding an efficient deterministic > enumerator for poset orders I'll write one. Graph-theoretical > algorithms are a fun way to spend the weekend :-) I personally think the topological_sort() algorithm begs for an algorithm that would allow iterating over all unique orderings. As I stated above, I think something of the sort of std::next_permutation() would make a lot of sense, especially because graph algorithms, and finding all different toposorts in particular, tend to be expensive operations. Simply asking for one ordering at a time gives a lot of flexibility to the user with the added benefit of efficiency and integration with the STL's way of doing things. -- Matan

1 0

All orderings of boost::graph::topological_sort()
by Matan Nassau 22 Feb '08

22 Feb '08

Hello, In implementing a dependency graph subsystem, one of the tasks is figuring out all the possible ways of traversing the graph with respect to the dependencies. boost::graph::topological_sort() looks like a reasonable place to start; however, topological_sort() only gives /one/ linear ordering, not all possible orderings. I digged into the code and found that the algorithm is in fact a call to depth_first_search with a custom visitor. The 0-in-degree vertices are not mentioned in the algorithm, and I suspect manipulating the order of these vertices would be a good start in generating different orderings. I also digged into the depth_first_search() code, but there too it doesn't seem to be straightforward to change the order in which those start-nodes are visited. I am also aware that even if I would find a way to change the order in which the 0-in-degree vertices are visited it will still not give me all the various orderings; I think that repeating the procedure with a bredth_first_search() will give me yet more different orderings, and with I'll have all the different orderings. I may have missed something obvious but I really don't know how to tackle the problem of generating various orderings. Any advice, please? Thank you! -- Matan

2 1

void_cast patch
by VennGrayDog 22 Feb '08

22 Feb '08

I found #pragma warning(push) without #pragma warning(pop) difs attached Index: void_cast.hpp =================================================================== --- void_cast.hpp (revision 43370) +++ void_cast.hpp (working copy) @@ -221,6 +221,10 @@ } // namespace serialization } // namespace boost +#ifdef BOOST_MSVC +# pragma warning(pop) +#endif + #include <boost/config/abi_suffix.hpp> // pops abi_suffix.hpp pragmas #endif // BOOST_SERIALIZATION_VOID_CAST_HPP

1 0

Re: [Boost-users] Problem using boost::bind with VC2005 (compiler crashes)
by Daniel Lidström 22 Feb '08

22 Feb '08

> > The compiler seems to crash when trying to compile the > > RailCaterpillarControllerGUI constructor. > > Any ideas why? > > I'm using 8.0.50727.867 (vsvista.050727-8600), boost-1.34.1 > > Try adding the argument placeholders to your bind call: > > // cbToggleDeadZone should be called by widget's callback > CBF_Callback callback( > boost::bind( > &RailCaterpillarControllerGUI::cbToggleDeadZone, > this, > _1, > _2 > )); > > > HTH / Johan Of course, forgot the placeholders... Thanks! -- Daniel

1 0

boost::filesystem::detail::not_found_error with Boost installer on VC2005
by Rainer Thaden 22 Feb '08

22 Feb '08

Hi all, yes, I know, there has been a thread some days ago, but I just registered and don't know how to reply to a mail that I don't have. Anyway, there seems to be a bug in boost 1.34.1 with VC 2005 when using directory_iterator. I used the filesystem before and everything worked fine. Now (since I use the iterator), I get 1>HD2.obj : error LNK2019: Verweis auf nicht aufgelöstes externes Symbol ""unsigned int boost::filesystem::detail::not_found_error" (?not_found_error@detail@filesystem@boost@@3IA)" in Funktion ""private: unsigned int __thiscall boost::filesystem::basic_directory_iterator<class boost::filesystem::basic_path<class std::basic_string<char,struct std::char_traits<char>,class std::allocator<char> >,struct boost::filesystem::path_traits> >::m_init(class boost::filesystem::basic_path<class std::basic_string<char,struct std::char_traits<char>,class std::allocator<char> >,struct boost::filesystem::path_traits> const &)" (?m_init@?$basic_directory_iterator@V?$basic_path@V?$basic_string@DU?$char_traits@D@std@@V?$allocator@D@2@@std@@Upath_traits@filesystem@boost@@@filesystem@boost@@@filesystem@boost@@AAEIABV?$basic_path@V?$basic_string@DU?$char_traits@D@std@@V?$allocator@D@2@@std@@Upath_traits@filesystem@boost@@@23@@Z)". I looked into the filesystem dll with depends and found the function there with ordinal 48 (0x30) (in both the debug and the release Version for multithreading). Also in the lib, I found it. Rather strange. The signatures of the linker and the ones found in the DLL also match. I used the installer from Boost-Consulting. In the thread some time ago, a responder claimed, he could compile successfully. Maybe he used a self-compiled boost version? I am going to test that, but maybe somebody has an answer right now? Best regards, Rainer Thaden

1 1

Problem using boost::bind with VC2005 (compiler crashes)
by Daniel Lidström 22 Feb '08

22 Feb '08

Hello! I'm trying to use boost::bind and boost::function to perform a callback mechanism. However, the VC2005 compiler crashes for some reason. Let me show the code right away: #include <boost/bind.hpp> #include <boost/function.hpp> #include <boost/any.hpp> #include <boost/noncopyable.hpp> #include <iostream> #include <tchar.h> struct Fl_Widget; typedef boost::function<void (Fl_Widget&, const boost::any&)> CBF_Callback; struct Fl_Widget { CBF_Callback cbf_callback_; boost::any user_data_; //Three calls to set or probe that functor callback variable. CBF_Callback cbf_callback() const {return cbf_callback_;} void cbf_callback(CBF_Callback c, const boost::any& p) {cbf_callback_=c; user_data_=p;} void cbf_callback(CBF_Callback c) {cbf_callback_=c;} //Calls to do the functor callback void do_cbf_callback() {cbf_callback_(boost::ref(*this),user_data_);} // override user_data_ void do_cbf_callback(const boost::any& user_data) {cbf_callback_(boost::ref(*this),user_data);} }; struct GUI : private boost::noncopyable { static GUI& Instance() { static GUI theGUI; return theGUI; } void print() { std::cout << "Inside GUI" << std::endl; } private: GUI(){} }; struct RailCaterpillarControllerGUI { Fl_Widget widget; RailCaterpillarControllerGUI() { // cbToggleDeadZone should be called by widget's callback CBF_Callback callback(boost::bind(&RailCaterpillarControllerGUI::cbToggleDeadZone, this)); widget.cbf_callback(callback); } void cbToggleDeadZone(Fl_Widget& widget, const boost::any& data) { // assume data is reference to the GUI instance boost::any_cast<boost::reference_wrapper<GUI> >(data).get().print(); } }; int _tmain(int argc, TCHAR** argv) { RailCaterpillarControllerGUI controller; controller.widget.do_cbf_callback(boost::ref(GUI::Instance())); } Compiling gives this: 1>------ Build started: Project: test, Configuration: Debug Win32 ------ 1>Compiling... 1>test.cpp 1>c:\users\daniel.sbg\programming\boost_1_34_1\boost\bind.hpp(1575) : warning C4180: qualifier applied to function type has no meaning; ignored 1> c:\users\daniel.sbg\programming\boost_1_34_1\boost\bind.hpp(1609) : see reference to class template instantiation 'boost::_bi::add_cref<Pm,I>' being compiled 1> with 1> [ 1> Pm=void (__thiscall RailCaterpillarControllerGUI::* )(Fl_Widget &,const boost::any &), 1> I=1 1> ] 1> c:\users\daniel.sbg\programming\test\test.cpp(48) : see reference to class template instantiation 'boost::_bi::dm_result<Pm,A1>' being compiled 1> with 1> [ 1> Pm=void (__thiscall RailCaterpillarControllerGUI::* )(Fl_Widget &,const boost::any &), 1> A1=RailCaterpillarControllerGUI * 1> ] 1>c:\users\daniel.sbg\programming\boost_1_34_1\boost\mem_fn.hpp(318) : warning C4180: qualifier applied to function type has no meaning; ignored 1> c:\users\daniel.sbg\programming\boost_1_34_1\boost\bind\bind_template.hpp(22 5) : see reference to class template instantiation 'boost::_mfi::dm<R,T>' being compiled 1> with 1> [ 1> R=void (Fl_Widget &,const boost::any &), 1> T=RailCaterpillarControllerGUI 1> ] 1>Project : error PRJ0002 : Error result 255 returned from 'C:\Program Files\Microsoft Visual Studio 8\VC\bin\cl.exe'. The compiler seems to crash when trying to compile the RailCaterpillarControllerGUI constructor. Any ideas why? I'm using 8.0.50727.867 (vsvista.050727-8600), boost-1.34.1 Thanks in advance! -- Daniel

2 1

regexp - how to do an anti-match on a word
by David Ward 22 Feb '08

22 Feb '08

Boosters: I'm having trouble determining a regular expression that will do the following: 1. Match the following strings: TermGeMstreamIPTriplePlay_11.0.IpTxMstream_0.DefaultGateway TermGeMstreamIPTriplePlay_11.0.VplsTxMstream_0.CurMbps TermGeMstreamIPTriplePlay_11.0.MplsTxMstream_0.AvgMbps TermGeMstreamIPTriplePlay_11.0.MacTxMstream_0.Address 2. But not match the string: TermGeMstreamIPTriplePlay_11.0.TrafficTxMstream_0.RampStart So, essentially match the strings with IpTx, VplsTx, MplsTx and MacTx, but not match strings with TrafficTx. Right now, I have: .*\.(IpTx|MacTx|MplsTx|VplsTx)Mstream_0\..* This works, however, it is not very robust because there may be more entries added to the matching words over time. Is there someway to do an anti-match on the word "TrafficTx"? Something like: .*\.(!TrafficTx)Mstream_0\..* Thanks, David

2 1

"Re: Boost, HP-UX and gcc"
by stacie＠dcindigos.com 22 Feb '08

22 Feb '08

I'm on 11.23 itanium using gcc 4.1.1 and the latest build of 1.35 beta (feb 21) With my setup, c++config,h indicates that the header wchar.h does exist, however gcc does *not* set _GLIBCPP_USE_WCHAR_T nor_GLIBCXX_USE_WCHAR_T. My build keeps throwing No WCHAR_MIN and WCHAR_MAX errors, and suggests that I adjust integer_traits<>. I was going to just let it slide but unfortunately I received this error while trying to build using the Unit Test framework. Since unit testing is my friend, I would really like to get this working. I tried a work-around from the boost mailing list, but the code that I found in my files was not quite like what he modified: ***** Added code at top of bad header file: #ifdef hpux #include <wchar.h> #undef WCHAR_MIN /* (wchar_t)0 */ #define WCHAR_MIN 0 #endif * error simply changed to: /home/mlinck/develop/api/tp-sources-trunk/boost_1_33_1/boost/regex/v4/regex_traits_defaults.hpp:94:28: error: missing binary operator before token "(" * tried a different workaround at the code where the error is thrown. Noticed that broken code gets skipped for HP_aCC compiler. Figure we can maybe get away with same for gcc compiler: => #if !defined HP_aCC && !defined hpux * seems to compile. Had to apply to boost/regex/v4/perl_matcher.hpp as well ****** How would I modify integer_traits<> ? Since that is what is suggested...... Thanks!

3 2

Re: [Boost-users] "Re: Boost, HP-UX and gcc"
by stacie＠dcindigos.com 21 Feb '08

21 Feb '08

Hi Boris, No help there. I'm still getting the same errors. I'm willing to look at it some more if you have any more suggestions. How can I turn off the building of the python and serialization libraries? There are other libraries that may be broken with HP, too and I'd like to exclude those from the build. Stacie > -------- Original Message -------- > Subject: Re: [Boost-users] "Re: Boost, HP-UX and gcc" > From: "Boris Gubenko" <Boris.Gubenko(a)hp.com> > Date: Thu, February 21, 2008 4:19 pm > To: <boost-users(a)lists.boost.org> > Cc: Boris Gubenko <Boris.Gubenko(a)hp.com> > > > I don't think John's fix in http://svn.boost.org/trac/boost/changeset/42890 made > it to 1.35. You migh want to apply it locally and see if it fixes your problem. > > Thanks, > Boris > > ----- Original Message ----- > From: <stacie(a)dcindigos.com> > To: <boost-users(a)lists.boost.org> > Sent: Thursday, February 21, 2008 3:36 PM > Subject: [Boost-users] "Re: Boost, HP-UX and gcc" > > > > I'm on 11.23 itanium using gcc 4.1.1 and the latest build of 1.35 beta > > (feb 21) > > > > With my setup, c++config,h indicates that the header wchar.h does exist, > > however gcc does *not* set _GLIBCPP_USE_WCHAR_T nor_GLIBCXX_USE_WCHAR_T. > > > > My build keeps throwing No WCHAR_MIN and WCHAR_MAX errors, and suggests > > that I adjust integer_traits<>. > > I was going to just let it slide but unfortunately I received this error > > while trying to build using the Unit Test framework. Since unit testing > > is my friend, I would really like to get this working. > > > > I tried a work-around from the boost mailing list, but the code that I > > found in my files was not quite like what he modified: > > > > ***** > > Added code at top of > > bad header file: #ifdef hpux #include <wchar.h> #undef > > WCHAR_MIN /* (wchar_t)0 */ #define WCHAR_MIN 0 #endif > > * error simply changed > > to: > > /home/mlinck/develop/api/tp-sources-trunk/boost_1_33_1/boost/regex/v4/regex_traits_defaults.hpp:94:28: > > error: missing binary operator before token "(" > > * tried a different workaround at the code where the error > > is thrown. Noticed that broken code gets skipped for > > HP_aCC compiler. Figure we can maybe get away with same > > for gcc compiler: > > => #if !defined HP_aCC && !defined hpux > > > > * seems to compile. Had to apply to > > boost/regex/v4/perl_matcher.hpp as well > > ****** > > > > > > How would I modify integer_traits<> ? Since that is what is > > suggested...... > > > > > > Thanks! > > > > > > > > _______________________________________________ > > Boost-users mailing list > > Boost-users(a)lists.boost.org > > http://lists.boost.org/mailman/listinfo.cgi/boost-users > > > > _______________________________________________ > Boost-users mailing list > Boost-users(a)lists.boost.org > http://lists.boost.org/mailman/listinfo.cgi/boost-users

2 1