March 2009 - Boost-users - lists.preview.boost.org

BGL Graph support for named vertices in adjacency_list documentation?
by ljw＠onet.eu 10 Mar '09

10 Mar '09

Is there any documentation available for the support for named vertices in adjacency_list introduced with version 1.38 and announced in the release notes? Regards Łukasz Wojakowski

2 1

[asio] Synchronous/Asynchronous pattern
by Rutger ter Borg 10 Mar '09

10 Mar '09

In an application we would like to support synchronous operations in the application layer, whereas operations below it are done with asynchronous operations. Callback handlers are not appropriate in this case, To make the example more clear, suppose you have a remote_vector which you would like to iterate through, remote_vector a(10); for( int i=0; i<10; ++i ) { std::cout << a[i] << std::endl; } here, a[i] will cause messages to be sent to retrieve the actual vector data before printing it. The network layer is hidden from the user. What is the best way to "emulate" synchronous behaviour with asio? The thread running this iteration should not block, but do other work (at the io_service?) until the requested results are received. Many thanks, Kind regards, Rutger ter Borg

3 7

condition_variable::timed_wait doesn't wait
by Daniele Barzotti 10 Mar '09

10 Mar '09

Hi! I'm trying to use condition_variable::timed_wait without success. Using the following code, the timed_wait returns immediatly also if Receive() is never called. Where I wrong? void Receive() { boost::lock_guard<boost::mutex> lock(_mut); //... code ...// _data_ready = true; // Notify and unlock the mutex _cond.notify_one(); } bool Send() { boost::unique_lock<boost::mutex> lock(_mut); boost::xtime xt; xtime_get(&xt, boost::TIME_UTC); xt.sec += 5; _data_ready = false; //... send data ...// //wait for an answer.. bool ready = _cond.timed_wait(lock, xt, boost::lambda::var(_data_ready)); return ready; } I've also tried to use: ptime t = second_clock::universal_time(); ptime timeout = t + milliseconds(cmd_timeout); bool ready = _cond.timed_wait(lock, timeout, boost::lambda::var(_data_ready)); but with the same effect. Thanks, Daniele.

3 2

Re: [Boost-users] Pointer serialization and deserialization
by niranjan bangera 10 Mar '09

10 Mar '09

Robert, Yes the pointer is been initialized by NULL in the constructor of the class of that pointer. Thanks, Niranjan Do you know for a fact that the *pointer* has been initialized? Robert Ramey "niranjan bangera" <niranjannina@...<http://www.nabble.com/user/SendEmail.jtp?type=post&post=22416132&i=0>> wrote in message news:50a2f6f20903090457n37c0bfb2n3eec9d716a56c84c@...<news://50a2f6f20903090457n37c0bfb2n3eec9d716a56c84c@mail.gmail.com/> ... Hi all, Boost:1.38.0 compiler:gcc4.1.1 I am trying to serialize *and* deserialize a *pointer* to the class, but ended with the segmentation fault, Is that the null *pointer* *serialization * gives the segmentation fault?. I read in document that, library will maintain the Null *pointer*. if not how can i serialize? ,any docs or related refference is a;so acceptable. Thanks in advance.... one more thing: what -1 in archive(text_archive) indicates? Example ------------ #include <fstream> #include "boost/archive/text_iarchive.hpp" #include "boost/archive/text_oarchive.hpp" #include "boost/*serialization*/access.hpp" #include "boost/*serialization*/assume_abstract.hpp" class Dkpt { public: friend class boost::*serialization*::access; template<class Archive> void serialize(Archive & ar, const unsigned int file_version) { (void) file_version; ar & data_; ar & b_i; ar & f_er; ar & pa_er; ar & ct_c; } private: uint8_t data_; bool b_i; bool f_er; bool pa_er; bool ct_c; }; ---- class RB : public RM { public: virtual void checkpoint_save(boost::archive::text_oarchive & oa) { oa & dr; } virtual void checkpoint_retrive(boost::archive::text_iarchive & ia) { ia & dr; } private: Dkpt* dr; ;} Thanks. Niranjan

2 1

Re: [Boost-users] [wave] assertion failed - multithread issue?
by Ein Held 10 Mar '09

10 Mar '09

>> I use seperate wave instances for each thread. Anyway, I'll try to get >> gdb running and debug the problem. > >Thanks! Any hint is highly appreciated. > >> Nontheless there is another problem(?) in wave. Wave still uses the old >> filesystem interface, which makes it incompatible with code compiled >> with BOOST_FILESYSTEM_NO_DEPRECATED. That is a bug, right? > >Ohh, I was not aware of that. How can I reproduce this? > > > >Regards Hartmut Simply define BOOST_FILESYSTEM_NO_DEPRECATED before including boost.wave/boost.filesystem ;)

2 1

Re: [Boost-users] [wave] assertion failed - multithread issue?
by Ein Held 09 Mar '09

09 Mar '09

Hello Sohail actually test is never ment to return (check the for(;;) ;) - so this should not do any harm on the running program. This is definitely not a good coding style, but IMO acceptable for that simple non-productive example ;)

1 0

Static linking/GCC 4.3.2 on Linux (-fPIC missing?)
by Anteru 09 Mar '09

09 Mar '09

Hi, I've been using Boost 1.37 so far on Linux, and linked some libraries statically into a shared library. That is, I have a .so (Foo.so), which statically links in Boost.Filesystem (among others). Yesterday I tried to compile Boost 1.38 for Linux, everything worked fine, but as soon as it got to linking my .so, GCC stops as Boost.Filesystem has not been compiled with -fPIC. I don't remember I had to set -fPIC on 1.37 though. Setting cxxflags=-fPIC obviously solves the problem (along with a recompile of Boost), so my question is: Is this a regression? Did I miss something? The environment is Linux with GCC 4.3.2, on x64. Cheers, Anteru

3 6

multidimensional array query
by Sean Farrow 09 Mar '09

09 Mar '09

Hi: I need to define a multi-dimensional array with 4 collumns and two rows, What is the correct way to do this? My current definition is: boost::multi_array<std::wstring, 2> table(boost::extents[4][Lines.size()]); Is this correct, Lines is a vector decared prevously in the code. Any help appreciated Sean.

1 0

[serialization] redefinition of ‘struct boost::archive::detail::init_guid<T>
by Jan Boehme 09 Mar '09

09 Mar '09

Hi, I have something like the following object hierarchy: class Root {}; class A : Root {}; // BOOST_CLASS_EXPORT in header file class B : Root {}; // BOOST_CLASS_EXPORT in header file class U : A {}; // BOOST_CLASS_EXPORT in header file class V : A {}; // BOOST_CLASS_EXPORT in header file class X : B {}; // BOOST_CLASS_EXPORT in header file class Y : B {}; // BOOST_CLASS_EXPORT in header file There are a lot of shared pointers to the base classes Root, A and B which are to serialize.Boost 1.37 works fine while Boost 1.38 brings "redefinition of ‘struct boost::archive::detail::init_guid<A>" "redefinition of ‘struct boost::archive::detail::init_guid<B>" "redefinition of ‘struct boost::archive::detail::init_guid<U>" "redefinition of ‘struct boost::archive::detail::init_guid<V>" "redefinition of ‘struct boost::archive::detail::init_guid<X>" "redefinition of ‘struct boost::archive::detail::init_guid<Y>" after trying to link. It seems clear I have to put the BOOST_CLASS_EXPORT stuff in the belonging implementation files. But this corrupts the input of files which were created with the former binaries using the BOOST_CLASS_EXPORT in the header. I need an advice how and where the macro to place in terms of a long time decision. Below the class definition (which I find a good solution) seems no longer to work. Thank you very much, Jan.

2 1

Re: [Boost-users] [wave] assertion failed - multithread issue?
by Ein Held 09 Mar '09

09 Mar '09

>From: Hartmut Kaiser <hartmut.kaiser <at> gmail.com> >Subject: Re: [wave] assertion failed - multithread issue? >Newsgroups: gmane.comp.lib.boost..user >Date: 2009-03-09 14:55:58 GMT (41 minutes ago) >> when parsing some cpp files via boost.wave using multiple threads, I >> randomly get a failed assertion. >> >> "Assertion failed: static_cast<RefCountType>(*d.begin()) != 0, file >> C:\projekte\libs\boost\include\boost- >> 1_38/boost/wave/util/flex_string.hpp, line 1216" >> >> I am pretty sure that this is a multi-thread issue, as when using only >> one thread to parse the files, everything works fine. >> Unfortunately I cannot give you more specific information as I cannot >> get gdb to run on my system. >> I am using gcc 4.3.3 > >That certainly makes it almost impossible to reproduce the problem for me. > >But one note: if you're using Wave in a multithreaded context (i.e. using the same Wave context/iterators >from different threads) you certainly need to properly lock the Wave objects as these are not thread safe >per se. The only thing guaranteed is that Wave doesn't use any (not protected) global data structures >internally.. So, as long as a certain wave context/iterator instance is used from one thread only (even in >multi threaded applications), you should be fine. > >HTH >Regards Hartmut Hello Hartmut I use seperate wave instances for each thread. Anyway, I'll try to get gdb running and debug the problem. Nontheless there is another problem(?) in wave. Wave still uses the old filesystem interface, which makes it incompatible with code compiled with BOOST_FILESYSTEM_NO_DEPRECATED. That is a bug, right? Thank you

2 1