February 2008 - Boost-users

Static build on darwin
by Thomas Hauk 12 Feb '08

12 Feb '08

Hi all, I'm having trouble with using Boost.Thread (1.34.1) on Mac OS X (10.4.11/ppc/Xcode 2.3). I have an Xcode project ("BSD command line tool") that links to what I believe is the static build of Boost.Thread (libboost_thread-mt- d-1_34_1.a). However, when Xcode goes to run the executable, it gives the error: > Running… > dyld: Library not loaded: bin.v2/libs/thread/build/darwin/debug/ > threading-multi/libboost_thread-mt-d-1_34_1.dylib > Referenced from: /path/to/my/executable/executablename > Reason: image not found > Program received signal: "SIGTRAP". If I run the executable inside an X11 shell, I get a similar error. If I add the path to the .dylib to the $DYLD_LIBRARY_PATH environment variable, I can run the executable. Which makes me wonder if somehow this .a file is really a .dylib. What's also strange is that although the build directories for all the other libraries have the same structure (e.g. filesystem/build/ darwin/debug/link-static/libboost_filesystem-d-1_34_1.a), the thread directory has a threading-multi directory under the link-static directory (meaning the file I am trying to link to is in thread/build/ darwin/debug/link-static/threading-multi/libboost_thread-mt-d-1_34_1.a). I'm building all the libraries with bjam 3.1.15 and the command line "bjam --toolset=darwin". FWIW, I'm also using the Boost filesystem library with this executable, without any compile, link or run-time errors. Any help/ideas? Also: is there any plan to distribute the Boost libraries for Mac OS X in a similar way to how they are distributed for Windows (i.e. installer that has pre-built library binaries)? T -- Some people, when confronted with a problem, think "I know, I'll use regular expressions." Now they have two problems. --Jamie Zawinski, in comp.lang.emacs

1 0

[serialization] exceptions are not caught
by Krishanu Debnath 12 Feb '08

12 Feb '08

Hello, This is a platform specific problem, platform is RHEL4. We use boost::serialization code, and linked with their static library. And the problem we are facing that all of our exception handler codes are not working any more i.e. catch handler is not called if an exception is thrown. We are not even calling the boost::serializer code. If we remove serializer code everything works fine. Serializer code works pretty fine in suse platform. Now my question is how come exception handler codes are not executing just because we are linking with serializer library. Any insight will be helpful. Thanks, Krishanu

1 0

[filesystem] bug in wpath with utf8 locales
by Filip Konvička 12 Feb '08

12 Feb '08

Hi, there is a problem with wpath when the system locale is UTF-8. My setup is g++ 4.2.3 --enable-clocale=gnu on a 2.6.18-5-686 GNU/Linux, boost 1.34.1, locale "cs_CZ.UTF-8". The problem seems to be in translating wstring into native strings inside Filesystem. The error I'm getting (and others too, see for example "Problem with exists() and wpath" and "Bug in filesystem library?" on this list) is boost::filesystem::wpath::to_external conversion error I've seen some discussions on the boost-dev list dealing with this issue, and to me this looks like a serious bug in wpath implementation. Until this is resolved, I think that we need to #ifdef this and use char-based paths on Linux. Are there some plans to fix this? I'm willing to help if that is needed. Thanks, Filip

1 0

Boost threads.
by Paul J R 12 Feb '08

12 Feb '08

Hi all, I had a question about boost threads as im using them currently in a project. I've written a very simple implementation of what im trying to do and the result I get, but im having trouble understanding why: mytest.cpp: class A { public: A() {} ~A() {} virtual void worker() { std::cout << "I am A worker" << std::endl; } virtual A *factory() { std::cout << "Factory for A class called" << std::endl; return new A; } virtual void operator()() { std::cout << "Calling A operator()" << std::endl; worker(); } void threadcontroller() { boost::thread_group thre; for(;;) { std::cout << "I am in the A.threadcontroller() method" << std::endl; A *mea = factory(); mea->worker(); thre.create_thread(*mea); exit(0); } } }; class B: public A { public: B() {} ~B() {} virtual void worker() { std::cout << "I am B worker" << std::endl; } virtual void operator()() { std::cout << "Calling B operator()" << std::endl; worker(); } B *factory() { std::cout << "Factory for B class called" << std::endl; return new B; } }; int main() { B meb; meb.worker(); meb.threadcontroller(); } Output: $ ./mytest I am B worker I am in the A.threadcontroller() method Factory for B class called I am B worker Calling A operator() I am A worker The first output is from the meb.worker() call in main(), the second is from meb.threadcontroller. the thrid is the factory call from within threadcontroller, the forth is mea.worker() in thread controller, then the next 2 are from after the thread_group.create_thread() call. Once its in the thread it seems to revert to class A's methods. Now, i know if i overload the threadcontroller class in B it'd solve all my problems, but the real threadcontroller method is alot more complex and class A has about 6 sub classes so replicating the threadcontroller in each of the classes would be a little tedious. Have i stumbled across a bug, or is that how its supposed to work? Thanks in advance

5 7

[python] Integration without a DLL
by Robert Dailey 11 Feb '08

11 Feb '08

Hi, I have a game that currently integrates the python interpreter using the Python C API. I would like to expose various C++ interfaces & functions to python, but I want to do so without having to create a separate DLL, as state needs to be modified in the actual core engine of the game as the user calls C++ functions from Python. For example, say I have a C++ function called CompleteQuest(). In Python, we would script a quest for my game and at some point call CompleteQuest() from python. This should map directly to a function in the executable (game). Is there a way to do this? Thanks.

4 9

[thread] is this a bug?
by Hector C 11 Feb '08

11 Feb '08

Hello fellow boost fans, Could anyone explain to me why the destructor (in the program listed below) seems to be called 10 times but the constructor only one time? The output of the program is also shown below. Cheers, Hector C. --- BEGIN PROGRAM --- #include <boost/thread/thread.hpp> #include <iostream> using namespace std; struct ThreadBug { ThreadBug() { cout << "Begin sending ... " << endl; } ~ThreadBug() { cout << "End sending ... " << endl; } void operator()() { cout << "Do something ... " << endl; } } w_ThreadBug; int main() { cout << "New Sequence -----------" << endl; boost::thread thrd(w_ThreadBug); thrd.join(); cout << "End New Sequence -----------" << endl; return 0; } ---END PROGRAM --- Begin sending ... New Sequence ----------- End sending ... End sending ... End sending ... End sending ... End sending ... End sending ... End sending ... End sending ... End sending ... Do something ... End sending ... End New Sequence ----------- End sending ...

2 1

[DateTime] posix_time_zone constructor crashes with bad format string
by Richard Buckle 11 Feb '08

11 Feb '08

Hello, I'm using Boost 1.34.1 with Visual Studio 2005 SP1. I've found that if the posix_time_zone constructor is invoked with certain types of invalid time zone string, the application crashes with an attempt to dereference an invalid iterator. Examples of strings that do this are strings with no digits, e.g. "UTC" and strings with a daylight saving time named but no changeover rules, such as "GMT0BST". This is a problem for us as in our application the time zone string is input by the user, and we need to fail gracefully if the user naively supplies an invalid string such as the two above. I think this is easily fixed by testing the iterators against end(), although I'm not quite sure which exception to throw in this case. Any ideas? TIA, Richard.

1 0

Linking to boost_regex library giving unresolved symbols
by shouvik ganguly 11 Feb '08

11 Feb '08

Hi, I am trying to access the function regex_search() by linking to the boost_regex library. However I am getting the following unresolved symols on GNU/Linux OS 1.undefined symbol: _ZN5boost14reg_expressionIcNS_12regex_traitsIcEEN4_STL9allocatorIcEEE14set_expressionEPKcS8_j(../libAspPatternFindReplace.so) 2. undefined symbol: _ZN5boost9re_detail21raise_regex_exceptionERKN4_STL12basic_stringIcNS1_11char_traitsIcEENS1_9allocatorIcEEEE (../libAspPatternFindReplace.so) 3. undefined symbol: _ZN5boost14c_regex_traitsIcE17transform_primaryERN4_STL12basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEERKS8_ (../libAspPatternFindReplace.so) 4. undefined symbol: _ZN5boost14c_regex_traitsIcE9transformERN4_STL12basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEERKS8_ (../libAspPatternFindReplace.so)The first and third unresolved symbol complain about the functions set_expression() and transform_primary(). However I am not able to get a clue as to the 1st and 4rth link error. How can I resolve the link errors. Thanks _________________________________________________________________ Express yourself instantly with MSN Messenger! Download today it's FREE! http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/

2 1

Re: [Boost-users] Review of Logging library is in progress
by David Abrahams 11 Feb '08

11 Feb '08

on Mon Feb 11 2008, "Gennadiy Rozental" <rogeeff-AT-gmail.com> wrote: > Hi, > > To those who managed to miss this very exciting submission, I would like to > encourage to find some time to invest into reviewing it. Admitedly this is > very open-bound problem domain and many people have different opinions on > design requirements and possible interfaces. It makes it even more important > to get large number of reviewers representing wide spectrum of users base. > Please let me know about any time constrains, I might try to arange for the > review period extension. You left out a very important piece of information: when is the review deadline? -- Dave Abrahams Boost Consulting http://boost-consulting.com

1 0

[program_options] exceptions
by Filip Konvička 11 Feb '08

11 Feb '08

Hi, I started using the program_options library and I have a question about using the exceptions thrown by the library. I'd like to localize the errors reported by the library, but I can't see why the members of the exception classes are private, or why are there no getters to retrieve the properties needed to re-construct the messages in another language. Also, I can't make default wstring values work...my attempt is: #include <boost/program_options.hpp> #include <string> using namespace std; using namespace boost::program_options; int main() { wstring lang; options_description lang_opt_desc; variables_map vm; lang_opt_desc.add_options() ("lang", wvalue<wstring>(&lang)->default_value(L"en_US"), "Language ID.") return 0; } fails with both gcc and msvc, saying that boost/lexical_cast.hpp(174) : error C2679: binary '=' : no operator found which takes a right-hand operand of type 'std::basic_string<_Elem,_Traits,_Ax>' (or there is no acceptable conversion) (I'm using Boost 1.34.1 with MSVC 9.0 and g++ 4.2.3). I looked in the code and it looks that I could use ... ("lang", wvalue<wstring>(&lang)->default_value(L"en_US", "en_US"), "Language ID.") ... which works fine...I just wanted to know whether this is the intended approach. Thanks, Filip

1 0