August 2010 - Boost-users - lists.preview.boost.org

cannot find namespace boost?
by Andreas Ronnquist 07 Aug '10

07 Aug '10

Hi! When compiling my project using Boost 1.43 and Visual C++ 2010 Express I get the following: error C2653: 'boost' : is not a class or namespace name error C2143: syntax error : missing ';' before '<' error C4430: missing type specifier - int assumed. Note: C++ does not support default-int error C2238: unexpected token(s) preceding ';' where I am using boost::shared_ptr<Some_Class> Is this some simple newb mistake I am making? This is on a cmake-based project, which works just fine with boost 1.37, but there on the other hand I get a whole lot of warnings like this: "Unknown compiler version - please run the configure tests and report the results" which I interpret as Visual C++ 2010 being too new to be recognized by the older version of boost. Is this some major change in boost that I have missed? I am just using this shared_ptr, and I believe that it is only header-based, and I don't need to compile boost to use it? Is that wrong? best regards -- Andreas Ronnquist <gusnan(a)gusnan.se>

2 2

Trouble building 32bit version
by Noah Roberts 06 Aug '10

06 Aug '10

I have a 64bit Linux system and I need to build boost for 32bit in order to link to this library I have to use. Trouble is that I keep getting this error when I try to link to any boost library: nroberts@Adad:~$ g++ test.cpp -m32 -L/home/nroberts/boost-x86/lib -lboost_thread-gcc44-mt /home/nroberts/boost-x86/lib/libboost_thread-gcc44-mt.so: undefined reference to `__sync_fetch_and_add_4' collect2: ld returned 1 exit status test.cpp just contains an empty main(). I've tried google and arrived at various places: http://stackoverflow.com/questions/130740/link-error-when-compiling-gcc-ato… http://old.nabble.com/compiling-32-bit-boost-libraries-on-a-64-bit-system--… Among others. I'm trying to build version 1.38. Here's my last attempt at supplying bjam with the correct parameters: bjam architecture=x86 address-model=32 toolset=gcc instruction-set=i686 stage It's funny though, even at the link in the reply held at nabble.com the documentation doesn't seem to have anything on any of these parameters. They don't appear in the bjam --help output either. So I seem to be kind of screwed in figuring this out myself. I'm just trying to guess until I get it but I'm out of ideas. What do I need to do? Thanks.

2 2

[assign] Default list_of() behavior
by Ryan McConnehey 06 Aug '10

06 Aug '10

None of the examples, for list_of, show the use of the default list_of constructor. When I tried to use list_of() for a vector I expected a blank vector without any elements. What I found was a vector that had one element of the vector type. This element had a default value. Is this the correct behavior? t_ucvector temp( list_of() ); //The temp vector has one element of type unsigned char with a default value of zero. There is a structure with a couple of constant std::vectors that are being initializing in the constructor's preamble. I've included the structure with the working version of the code. I've included the desired creation of the std::vector, but because of list_of() is not working. Thank you for your help. Ryan struct SplitterFixture { public: SplitterFixture(void) : header(list_of(2)(2)) , messageA(list_of('a').repeat(2,'a')) , input(t_boostVector(new t_ucvector(list_of(2)(2).range(messageA)))) //Desired implementation. //: header(list_of().repeat(2,'2')) //, messageA(list_of().repeat(3,'a')) //, input(t_boostVector(new t_ucvector(list_of().range(header).range(messageA)))) {} public: typedef std::vector<unsigned char> t_ucvector; typedef boost::shared_ptr<t_ucvector> t_boostVector; public: t_ucvector const header; t_ucvector const messageA; t_boostVector const input; };

2 6

[preprocessor] removing parentheses
by Edward Diener 06 Aug '10

06 Aug '10

I have a single macro parameter which allows its passed value to contain commas. Therefore the user of the macro needs parentheses around the parameter if it contains one or more commas ( to does not have to ). When processing this parameter in a macro I would like to remove the outer parentheses if they are there before passing the value on for further processing. Is there a way to do this using some preprocessor library macro ?

4 12

x86_64 needs -fPIC issue, anyone successful?
by Bill Freeman 06 Aug '10

06 Aug '10

I'm trying to build a library (librets) that links against boost libraries in order to build a shared library (as part of a SWIG based python extension). The target machine, as well as all of my development machines at the moment, are the x86_64 architecture. I've done most of my work on an ubuntu 9.04 box. I've tried with the boost packages available from the ubuntu repositores, and I have also, in case those packages were miss-built, successfully built (and tried the examples, successfully) boost 1.43.0 from source, getting no errors from ./bjam install . In either case, I wind up with the following error when trying to build librets: /usr/bin/ld: /home/f/Desktop/rets/lib/libboost_system.a(error_code.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a shared object; recompile with -fPIC Googling around, this seems to be a problem only seen on x86_64 machines. (Why one doesn't need to make position independent code for the 32 bit machines, if that is in fact true, is beyond me.) So, I'm wondering if someone else with this issue has already figured out how to get around it. Or, if not, can someone tell me how to tell the bjam stuff to apply -fPIC so that I can try that. (I'm an old timer. If it were make based, I could hack it myself, but I'm hesitant to learn all about the boost build system for what is for me a side task.) Or if someone knows that the -fPIC won't help, or has alternate suggestions, please let me know. TIA, Bill

3 4

Re: [Boost-users] x86_64 needs -fPIC issue, anyone successful?
by Bill Freeman 06 Aug '10

06 Aug '10

Bill Freeman wrote: > I'm trying to build a library (librets) that links > against boost libraries in order to build a shared > library (as part of a SWIG based python extension). ... > /usr/bin/ld: /home/f/Desktop/rets/lib/libboost_system.a(error_code.o): > relocation R_X86_64_32 against `.rodata.str1.1' can not be used when > making a shared object; recompile with -fPIC > ... > Or, if not, can someone tell me how to tell the bjam > stuff to apply -fPIC so that I can try that. ... Neal Becker wrote: --------------------------------------- You should have a libboost_system.so to link against (which should have, and probably was, built with -fpic). Either: 1) You don't have libboost_system.so where linker is looking or 2) Usually these are setup as symlinks, libboost_system.so -> libboost_system.so.x.y.z You could be missing the symlink, or it could point to the wrong place. ------------------------------------------ The so files are there. But librets explicitly wants the .a, presumably because it wants to bind in the actual code, rather than the stub. While python extension libraries themselves need to be shared, so that they can be loaded dynamicly, at run time, the librets folks apparently don't want to depend on the boost shared libraries being installed on the target system. While the wisdom of that choice is debatable, librets is simply something I want to use (from python). I'm not involved in its development community. I'm just looking for the easy way out. Getting the boost static's built PIC seems worth trying. Bill

1 0

[Boost.Log] Adding indentation to log messages written via BOOST_LOG_TRIVIAL
by Torri, Stephen CIV NSWCDD, W15 06 Aug '10

06 Aug '10

Symptom: Adding indentation to log messages sometimes results in no indentation being added to incoming log messages. Environment: Fedora 13 Boost version: trunk (1.44 prior to beta release) Boost.Log version: 1.0 Compiler: GCC 4.4.4 Research: As I read the documentation I see that the Boost.Log talks about a non-empty log record is considered to be tied to the current thread. Since my program is single threaded I did not consider my initial attempt of using inline functions in a separate namespace with a namespace variable to keep track of indentation level to be a problem. My first attempt looked like: namespace project { namespace log { static boost::uint32_t indent_level = 0; #ifdef DEBUG #include <boost/log/trivial.hpp> #include <boost/format.hpp> /*! * \brief Record the entering of a function * \param function String name of the function being entered */ template <typename Function> inline void record_enter ( Function const& function ) { BOOST_LOG_TRIVIAL(trace) << std::string ( indent_level, ' ' ) << boost::str ( boost::format ( "(Enter) %1%" ) % function ); indent_level += 2; } /*! * \brief Record the exiting from a function * \param function String name of the function being exited */ template <typename Function> inline void record_exit ( Function const& function ) { BOOST_LOG_TRIVIAL(trace) << std::string ( indent_level, ' ' ) << boost::str ( boost::format ( "(Exit) %1%" ) % function ); indent_level -= 2; } /*! * \brief Record a message with one input * \param format_string String written in the Boost.Format style to be used to parse input * \param t1 Input value */ template <typename Format, typename T1> inline void record_debug ( Format const& format_string, T1 const& t1 ) { BOOST_LOG_TRIVIAL(debug) << std::string ( indent_level, ' ' ) << boost::str ( boost::format ( format_string ) % t1 ); } #else template <typename Function> inline void record_enter ( Function const& ) {} template <typename Function> inline void record_exit ( Function const& ) {} template <typename Format, typename T1> inline void record_debug ( Format const&, T1 const& ) {} } // namespace log } // namespace project #endif When I run the code I see that the indentation is tracking by steps of 2 as I go into functions and reduced by 2 as I exit. There are exceptions where the indentation is 0 for a few lines or so then resumes the correct level. The problem I think is that there are more than one thread handling what I send to Boost.Log via BOOST_LOG_TRIVIAL. It as if there is one thread that has the correct level and another thread that has indent_level set to 0. For example: 1 [2010-Aug-06 12:03:15.329757] [0x88bb070] [trace] (Enter) void Test::functionA() 2 [2010-Aug-06 12:03:15.330867] [0x88bb070] [debug] Function A: do something 3 [2010-Aug-06 12:03:15.330989] [0x88bb070] [trace] (Enter) double Value::CalcX() 4 [2010-Aug-06 12:03:15.331124] [0x88bb070] [debug] X: 5 5 [2010-Aug-06 12:03:15.331188] [0x88bb070] [debug] Y: 10 6 [2010-Aug-06 12:03:15.331252] [0x88bb070] [trace] (Enter) double Value::CalcX() My intial guess is that I have Boost.Log compiled multi-threaded and I have a single-threaded solution. Is there a way to track which thread was recording a message in Boost.Log? I would like to confirm Boost.Log is using different threads to handle my messages. Also is there a better way to do what I want? Stephen Stephen Torri, PhD NAVAL SURFACE WARFARE CENTER DAHLGREN 17214 Ave B Suite 121 Dahlgren, Va. 22448 540-653-7055 I am not a contracting officer. I cannot modify or initiate contracts, nor do I have the authority to financially commit the government in any way.

1 0

Problem with enable_shared_from_this and shared_ptr
by Ryan McConnehey 06 Aug '10

06 Aug '10

If class "Pool" inherits from enable_shared_from_this the documentation says I'm able to do the following. shared_ptr<Pool> aPool = shared_from_this(); The minimal code I've included shows that this isn't happening. The code compiles but I get this error "tr1::bad_weak_ptr". What should I be doing different? Ryan [CODE] //C++ Includes #include <exception> #include <iostream> #include <vector> #include <algorithm> //Boost Includes #include <boost/shared_ptr.hpp> #include <boost/enable_shared_from_this.hpp> class pool : public boost::enable_shared_from_this< pool > { private: using boost::enable_shared_from_this< pool >::shared_from_this; public: boost::shared_ptr<int> get(void) { boost::shared_ptr<pool> aPool = shared_from_this(); //This is where the tr1::bad_weak_ptr happens. return boost::shared_ptr<int>(new int); } }; int main(int argc, char * argv[]) { try { pool * myPool = new pool; boost::shared_ptr<int> a = myPool->get(); } catch (std::exception & exception) { std::cout << exception.what() << std::endl; } return 0; }

6 13

Re: [Boost-users] x86_64 needs -fPIC issue, anyone successful? (Bill Freeman)
by Vial, Florent 06 Aug '10

06 Aug '10

Hi Bill, > /usr/bin/ld: /home/f/Desktop/rets/lib/libboost_system.a(error_code.o): > relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a shared object; > recompile with -fPIC > Or if someone knows that the -fPIC won't help, or has alternate suggestions, please let me know. -fPIC is usually used when building a shared object, so it should not be relevant as you are linking statically to the library. However, I had the same problem as you lately and it had not _anything_ to do with fPIC whatsoever. It was because i was compiling with gcc 3.4.5 and the flag -fvisibility=hidden and -fvisibility-inlines-hidden. Those 2 flags are correctly implemented from gcc 4.1 onwards i think. Maybe this helped, --Florent Arnold & Richter Cine Technik GmbH & Co. Betriebs KG Sitz: Munchen - Registergericht: Amtsgericht Munchen - Handelsregisternummer: HRA 57918 Personlich haftender Gesellschafter: Arnold & Richter Cine Technik GmbH Sitz: Munchen - Registergericht: Amtsgericht Munchen - Handelsregisternummer: HRB 54477 Geschaftsfuhrer: Franz Kraus; Dr. Martin Prillmann A new star is born - ALEXA the most versatile digital camera system. Get all the latest information from arridigital.com and YouTube.

1 0

[program_options] unknown positional option
by Jeffrey Finkelstein 05 Aug '10

05 Aug '10

I'm having some trouble using the program_options library. Here is some code slightly modified from an example in the documentation for boost::program_options. <code> #include <iostream> #include <vector> #include <string> #include <boost/program_options/positional_options.hpp> #include <boost/program_options/variables_map.hpp> #include <boost/program_options/parsers.hpp> namespace po = boost::program_options; int main(int argc, char ** argv) { po::positional_options_description desc; desc.add("files", -1); std::cout << "Declared positional options." << std::endl; po::variables_map vm; po::store(po::command_line_parser(argc, argv).positional(desc).run(), vm); std::cout << "Stored command line arguments into vm." << std::endl; po::notify(vm); std::cout << "Notified vm that all args have been parsed" << std::endl; std::cout << " num files specified: " << vm.count("files") << std::endl; std::vector<std::string> files = vm["files"].as<std::vector<std::string> >(); for (unsigned int i = 0; i < files.size(); ++i) { std::cout << "file: " << files[i] << std::endl; } return 0; } </code> Linking with version 1.40, the above code, when run with the command line arguments "hello world", produces the following output: <shell> Declared positional options. terminate called after throwing an instance of 'boost::exception_detail::clone_\ impl<boost::exception_detail::error_info_injector<boost::program_options::unkno\ wn_option> >' what(): unknown option files Aborted </shell> Linking with version 1.43, the above code, when run with the command line arguments "hello world", produces the following output: <shell> Declared positional options. Segmentation fault </shell> Am I missing something about how to use positional arguments with this library? I am using Ubuntu 10.04 on x86. The 1.40 version of boost::program_options was installed through the repositories, I compiled the 1.43 version manually. Jeffrey

2 4