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

Filesystem recursive iterator issues
by mlimber 24 Oct '16

24 Oct '16

I have this code, which does a simple recursive directory iteration and prints each directory entry and then does it again ( http://coliru.stacked-crooked.com/a/0bfab95a513ebed8 ): #include <boost/filesystem.hpp> #include <boost/range/iterator_range.hpp> #include <iostream> namespace fs = boost::filesystem; int main() { try { const auto iter = fs::recursive_directory_iterator{ "." }; for( const auto& entry : boost::make_iterator_range( iter, {} ) ) { std::cout << entry << "\n"; } // Rinse and repeat std::cout << "\n"; for( const auto& entry : boost::make_iterator_range( iter, {} ) ) // CRASH -- Replace 'iter' with 'fs::recursive_directory_iterator{"."}' and it will work. //for( const auto& entry : boost::make_iterator_range( fs::recursive_directory_iterator{"."}, {} ) ) { std::cout << entry << "\n"; } } catch( const std::exception& e ) { std::cerr << e.what() << "\n"; } catch( ... ) { std::cerr << "Unknown exception.\n"; } } As you can see at Coliru, it crashes. (It also crashes on Visual Studio 2012 and with Boost 1.55.0 and on Coliru with <experimental/filesystem> instead of Boost.) Similarly, using a filesystem iterator with std::distance() fails when the iterator passed in is used again later (with boost: http://coliru.stacked-crooked.com/a/1b06dca057872762 and with experimental/filesystem: http://coliru.stacked-crooked.com/a/7bd5e73f1444619d ). It seems that something in the state of the *constant* iterator gets confused or corrupted after it is advanced to the end. I have a workaround for it commented out in the code. I found this possibly related issue: https://svn.boost.org/trac/boost/ticket/11910 Am I missing something? Should I file a new bug report? Cheers! M

1 0

When to use BOOST_ALL_DYN_LINK?
by Florian Lindner 24 Oct '16

24 Oct '16

Hello, http://www.boost.org/doc/libs/1_62_0/libs/log/doc/html/log/installation/con… says: BOOST_LOG_DYN_LINK If defined in user code, the library will assume the binary is built as a dynamically loaded library ("dll" or "so"). Otherwise it is assumed that the library is built in static mode. This macro must be either defined or not defined for all translation units of user application that uses logging. This macro can help with auto-linking on platforms that support it. BOOST_ALL_DYN_LINK Same as BOOST_LOG_DYN_LINK but also affects other Boost libraries the same way. We build our library in a dynamic and a static version using SCons: staticlib = env.StaticLibrary ( target = buildpath + '/libprecice', source = [sourcesPreCICE] ) env.Alias("staticlib", staticlib) solib = env.SharedLibrary ( target = buildpath + '/libprecice', source = [sourcesPreCICE] ) env.Alias("solib", solib) Both have BOOST_LOG_DYN_LINK set: env.Append(CPPDEFINES=["BOOST_LOG_DYN_LINK"]) However, I wonder if this is correct. Should DYN_LINK only be set when building the dynamically linked lib? Thanks, Florian

2 1

Update the index of a hashed_non_unique entry in boost::multiindex
by Ram 24 Oct '16

24 Oct '16

I have a boost::multiindex container which has an index as follows : boost::multi_index::hashed_non_unique< boost::multi_index::tag<IndexByNonUniqueId>, boost::multi_index::const_mem_fun<StoredObj, std::string, &StoredObj::getNonUniqueIdString> > This ID comes in as ""(empty string) when I initially get the object. I add it to the multiindex at this point. But later in the course of the program the ID gets populated using its "setNonUniqueId" method. At this point of time inside "setNonUniqueId" I would like to update the index to the object(in the multiindex) as well(while the ID is being updated from "" to something concrete). How do I do this? How do I locate the specific object and update the index? -R

2 4

upgrade from boost 1.39 to boost 1.50. Issue with regex_match
by MAURICE Jean 21 Oct '16

21 Oct '16

Hi, I am 'upgrading' a huge C++ app from VS2005 (32bits) to VS2010 (64 bits) and boost 1.39 to boost 1.50. This app runs both under Windows and Linux. In a function that searchs subdirectories and files of a given directory, we get the list of all the elements and then this list is filtered by : if(true==boost::regex_match(sAdaptation, what, childNode.infoNode.rFullPathFilePattern)) sAdaptation is ".\TEST\un_noeud" (this subdirectory exists) childNode.infoNode.rFullPathFilePattern is something like {m_pimpl={...}} And regex_match always returns false. If you can give me information about what is happening ... Jean

4 10

build boost 1.58 failed on centos 7.2
by Glenn WANG 20 Oct '16

20 Oct '16

hi, I try to build boost 1.58 on CentOS, but it cannot be done, there're many errors. could anybody help? thanks. ============================== ./b2 --with-log -q gcc.compile.c++ bin.v2/libs/system/build/gcc-4.8.5/release/threading-multi/error_code.o In file included from libs/system/src/error_code.cpp:19:0: ./boost/system/detail/error_code.ipp:48:27: error: expected ‘;’ at end of member declaration const char * name() const BOOST_SYSTEM_NOEXCEPT; ^ ./boost/system/detail/error_code.ipp:48:33: error: ‘BOOST_SYSTEM_NOEXCEPT’ does not name a type const char * name() const BOOST_SYSTEM_NOEXCEPT; ^ ./boost/system/detail/error_code.ipp:56:32: error: expected ‘;’ at end of member declaration const char * name() const BOOST_SYSTEM_NOEXCEPT; ^ ./boost/system/detail/error_code.ipp:56:38: error: ‘BOOST_SYSTEM_NOEXCEPT’ does not name a type const char * name() const BOOST_SYSTEM_NOEXCEPT;

1 0

Boost Coroutine
by Miguel Portilla 19 Oct '16

19 Oct '16

Hello, Using Boost 1.62, I am seeing the following warning when I build my project. warning : Boost.Coroutine is now deprecated. Please switch to Boost.Coroutine2. To disable this warning message, define BOOST_COROUTINE_NO_DEPRECATION_WARNING. I haven't found any documentation on what exactly this means or how to 'switch'. Can someone please point me to some docs or explain what 'switching' entails? Thanks, -M

2 3

Re: [Boost-users] [Boost-build] [System] Link issues using BOOST_ERROR_CODE_HEADER_ONLY
by Matthias Seemann 17 Oct '16

17 Oct '16

Hi, > We decided to fix these by defining the macro BOOST_ERROR_CODE_HEADER_ONLY > according to this link: > > http://www.boost.org/doc/libs/1_56_0/libs/system/doc/reference.html#Header-… > > However, when building Boost itself with this macro defined, we're > encountering several problems: > ... > 2) We get multiple definitions of "boost::system::throws" > Defining the macro BOOST_SYSTEM_NO_DEPRECATED (see that doc page above) resolved the issue of multiple definitions of "boost::system::throws". Matthias

1 0

boost graph pruning vertices (nodes)
by Brian Davis 14 Oct '16

14 Oct '16

In reading http://www.boost.org/doc/libs/1_62_0/libs/graph/doc/adjacency_list.html regarding Iterator and Descriptor Stability/Invalidation it leaves the impression that when using vecS there is no real good way to remove nodes. as example: typedef adjacency_list<listS, vecS> Graph; // VertexList=vecS all cases are wrong and typedef adjacency_list<listS, listS> Graph; // VertexList=listS where only one case is correct which leave the programmer walking the mine field. Of course same can be said for stl. My goal to remove graph vertices based on the value of the connected components. I currently have a method which uses connected components to generate a connected components list. I then generate a histogram of the connected components values thus giving me how many vertices are connected and what the connection property value is... assuming I understand connected components however I have thought I understood portions of boost only to find through some brain warp and folding my thought process back on itself like an origami master that I came to understand how certain parts work. vertid, component 0 0 1 0 2 0 3 0 4 1 5 1 6 2 7 2 8 2 9 2 10 2 11 2 12 3 13 3 so in this simple example vert ids [0-3] are connected in subnet 0 with a total of 4, [4-5] are connected in sub net 1 with a total of 2, [6-11] are connected in subnet 2 with a total of 6, and [12-13] is connected in subnet 3 with a total of 2. Histogram yields connected component value and number of vertices that have that value in the bin component : bin counts 0: #### 1: ## 2: ###### 3: ## etc for say subnet/subgraph less than 4 would yield the filtered histogram: 1: ## 3: ## using values 1, and 3 from filtered histogram I can then find the vertex ids to filter. Yielding vertices 4,5,12, and 15 to be removed which would remove subnets 1 and 3. 4 1 5 1 12 3 13 3 I tried to use boost::accumulators::density and maybe I have a different idea of what a histogram really is, but for a typical histogram density is just plane broken due to the seemingly use of count for number of bins (seems to ignore tag::density::num_bins) and associated logic. accumulator_t acc( tag::density::num_bins = n_bins, tag::density::cache_size = n_bins/2 ); Sure wish there was a boost::accumulators::histogram that just did what we need as programmers. I am unsure as to how much boost coolaid to drink before finding any use for density. I did find this relating to the seemingly not as advertised implementation of density http://lists.boost.org/Archives/boost/2008/01/132801.php. From http://www.boost.org/doc/libs/1_62_0/doc/html/accumulators/user_s_guide.html "The tag::density feature returns a histogram of the sample distribution. For more implementation details, see density_impl." err huh no it does not... at least in my testing. It would also be great if the abcicissas (intervals) were some how obtainable to compare to the ordinates (bin counts) . Note the histogram.hpp attached there. So I write by own histogram code using stl. I digress on that. So as I said I then filter the vertex vs connected components data based on a threshold say 4 to remove all vertices that are not connected to 4 or more vertices in a subnet/subgraph So I then have a list of vertices (subnets/subgraphs) that I wish to prune from the graph thus these questions: Question 1: if finding a list of vertices to erase, sorting them, and then using, a reverse iterator to remove(*itr, g) would this also not work? If this works why is it now shown as an example of vecS vertex removal working? Question 2: Should there not be a boost graph-ish way of doing this such as coloring vertices to be pruned by possibly adding property prune = true. then maybe a future call to prune(g) Question 3: Possibly there is a way to get filtered_graph to do the trick? Statement: it is unfortunate the use of the term vertex/vertices in stead of node(s) in boost graph. As some graphs have a vertex (euclidean geometry) such as a graph of the center-lines of the vascular network of a brain while other graphs do not. It is very confusing in this instance to keep the two strait in code. Ugh!

3 2

Re: [Boost-users] [Serialization] Free-standing serialize function for "enum class"
by Bjorn Reese 13 Oct '16

13 Oct '16

On 10/13/2016 06:56 PM, Rüdiger Berlich wrote: > I have for a while suspected a problem with the serialization of an „enum class“ item (currently with Boost 1.61 on Ubuntu 16.04), and have written a small test program to further examine any potential problems. While the generated XML looks fine and I cannot see any problems with the enum class, my custom, free-standing „serialize“ function does not get called. Boost.Serialization automatically converts an enum into an int before it is serialized.

2 1

[Serialization] Free-standing serialize function for "enum class"
by Rüdiger Berlich 13 Oct '16

13 Oct '16

Hi there, I have for a while suspected a problem with the serialization of an „enum class“ item (currently with Boost 1.61 on Ubuntu 16.04), and have written a small test program to further examine any potential problems. While the generated XML looks fine and I cannot see any problems with the enum class, my custom, free-standing „serialize“ function does not get called. It is modelled after the examples for non-intrusive serialization, so I do not understand the reason behind this. Any idea ? /*************************************/ // Compile with g++ --std=c++14 -o enum_ser -I/opt/boost/include/ -L/opt/boost/lib enum_ser.cpp -lboost_serialization #include <boost/serialization/nvp.hpp> #include <boost/archive/xml_oarchive.hpp> #include <boost/archive/xml_iarchive.hpp> #include <iostream> #include <sstream> enum class test_enum : int {a=5,b=6}; namespace boost { namespace serialization { //----------------------------------------- template<typename Archive> void serialize(Archive &ar , test_enum &x, const unsigned int){ std::cout << "Serializing via custom serialize function" << std::endl; int t = 0; if (Archive::is_saving::value) { t = static_cast<int>(x); ar & t; } else { ar & t; x = static_cast<test_enum>(t); } } //----------------------------------------- } } int main(int argc, char **argv){ std::ostringstream oss; test_enum x = test_enum::a; { boost::archive::xml_oarchive oa(oss); oa << BOOST_SERIALIZATION_NVP(x); } std::cout << oss.str() << std::endl; } /*************************************/ Output of the program (note: the „Serializing via custom serialize function“ does not appear) : /*************************************/ <?xml version="1.0" encoding="UTF-8" standalone="yes" ?> <!DOCTYPE boost_serialization> <boost_serialization signature="serialization::archive" version="14"> <x>5</x> </boost_serialization> /*************************************/ Thanks and Kind Regards, Beet

1 0

Boost test failures on arm platform
by Carlos Ferreira 13 Oct '16

13 Oct '16

Hello! I'm cross-compiling Boost to an arm platform and because of that, boost fails to execute the cxx11_* tests and then it outputs the information that the compiler does not support a series of cxx11_* capabilities. Is there a way to disable these library tests, when cross-compiling? -- Carlos Miguel Ferreira Researcher at Telecommunications Institute Aveiro - Portugal Work E-mail - cmf(a)av.it.pt Skype & GTalk -> carlosmf.pt(a)gmail.com LinkedIn -> http://www.linkedin.com/in/carlosmferreira

2 16

Announcing version 3 of the cxx_dual library
by Edward Diener 12 Oct '16

12 Oct '16

I have created version 3 of the cxx_dual library. version 3 is a large rewrite of the documentation to simplify the use of the library, as well as a change in the directory structure so that the basic usage of the library is through a single header file and the use of a namespace alias, and not through the macro interface. As in: #include <boost/cxx_dual/regex.hpp> void SomeFunction() { cxxd_regex_ns::regex re("A regular expression etc."); bool result(cxxd_regex_ns::regex_match("Some string...",re)); // etc. } What ! No macros ! Actually a few macros still exist for testing which dual library has been chosen and overriding the dual library chosen, but the basic functionality means almost no macros in user code. The library is available at https://github.com/eldiener/cxx_dual.git. What is it ? The Cxx Dual library, or CXXD for short, is a C++ macro header-only library which chooses between using a Boost library or its C++ standard equivalent library for a number of different C++ implementations, while allowing the end-user to use the same code to program either choice. An implementation is a Boost library which has a C++ standard library equivalent whose public interfaces are nearly the same in both cases. An 'implementation' is called a 'mod' for short and each of the possible Boost or C++ standard implementations for that mod are called a 'dual library'. The Cxx Dual library does this by automatically choosing either the Boost library or the C++ standard library for a particular mod, including the correct header file(s) needed for the dual library chosen, and creating a namespace alias for the namespace of the dual library chosen. CXXD also provides a macro-based solution for distinguishing between the Boost version and the C++ standard version of a dual library so that specific code for a particular dual library choice may be written in those cases where the public interfaces diverge.

1 0

Issues trying to use boost::unordered_map
by Ram 11 Oct '16

11 Oct '16

Hi, I am trying to use a boost::unordered_map as a member of my class. The unordered_map is defined as, boost::unordered_map<int, int> a; as a private member in the class. I have included #include "boost/unordered_map.hpp" using namespace boost; This is the compilation error I get, x_objs.cpp c:\codeworks\common\lib\boost\boost\unordered\detail\buckets.hpp(767) : error C2059: syntax error : '(' c:\codeworks\common\lib\boost\boost\unordered\detail\buckets.hpp(766) : while compiling class template member function 'void boost::unordered::detail::functions<H,P>::construct(bool,const H &,const P &)' with [ H=boost::hash<int>, P=std::equal_to<int> ] c:\codeworks\common\lib\boost\boost\unordered\detail\table.hpp(152) : see reference to class template instantiation 'boost::unordered::detail::functions<H,P>' being compiled with [ H=boost::hash<int>, P=std::equal_to<int> ] c:\codeworks\common\lib\boost\boost\unordered\detail\unique.hpp(179) : see reference to class template instantiation 'boost::unordered::detail::table<Types>' being compiled with [ Types=boost::unordered::detail::map<std::allocator<std::pair<const int,int>>,int,int,boost::hash<int>,std::equal_to<int>> ] c:\codeworks\common\lib\boost\boost\unordered\unordered_map.hpp(74) : see reference to class template instantiation 'boost::unordered::detail::table_impl<Types>' being compiled with [ Types=boost::unordered::detail::map<std::allocator<std::pair<const int,int>>,int,int,boost::hash<int>,std::equal_to<int>> ] c:\codeworks\sample_issue.h(276) : see reference to class template instantiation 'boost::unordered::unordered_map<K,T>' being compiled with [ K=int, T=int ] li_pip.cpp ld_pip.cpp pip.cpp . . . Can someone please help me? Thanks, Ram

6 39

[fiber] Asio autoecho example crashes (VS 2015 u3)
by DePizzottri 11 Oct '16

11 Oct '16

Hi all, I'm found that asio autoecho example crashes after main's return statement. Debugger shows write access from the fiber scheduler to the already destroed io_service. Substitution reference to io_service: boost::asio::io_service & io_svc_; in round_robin algo to the shared_ptr: boost::shared_ptr<boost::asio::io_service> io_svc_; with approprieate correction of the other code solves the problem. Maybe I misunderstood something in the descrition from documentation? Or its some kind of feature? -- View this message in context: http://boost.2283326.n4.nabble.com/fiber-Asio-autoecho-example-crashes-VS-2… Sent from the Boost - Users mailing list archive at Nabble.com.

2 4

Re: [Boost-users] [ASIO] Do coroutines allow serving multiple clients at once?
by mmarkd9＠libero.it 10 Oct '16

10 Oct '16

Thanks, does spawning a coroutine also spawn a new thread? I'd be inclined to think so from what you told me

2 1

R: Re: [ASIO] Do coroutines allow serving multiple clients at once?
by mmarkd9＠libero.it 10 Oct '16

10 Oct '16

Thanks, does spawning a coroutine also spawn a new thread? I'd be inclined to think so from what you told me ----Messaggio originale---- Da: "Oliver Kowalke" <oliver.kowalke(a)gmail.com> Data: 10-ott-2016 15.19 A: "boost-users"<boost-users(a)lists.boost.org> Ogg: Re: [Boost-users] [ASIO] Do coroutines allow serving multiple clients at once? 2016-10-10 9:44 GMT+02:00 mmarkd9(a)libero.it <mmarkd9(a)libero.it>: I'd like to know if this code is able to serve multiple clients connecting at the same time to the server. I would say no. You spawn only one coroutine that calls in the for-loop async. operations of accept(), read() and write() in a sequence. No other client is able to connect to your server, e.g. connect() on the client side returns, until you enter accept() again. In order to serve multiple clients at the same time you need to spawn a coroutine that accepts new connections and for each client request (read()/write() operations) you need to spawn a new coroutine.

1 0

Re: [Boost-users] [ASIO] Do coroutines allow serving multiple clients at once?
by mmarkd9＠libero.it 10 Oct '16

10 Oct '16

Thanks, does spawning a coroutine also spawn a new thread? I'd be inclined to think so from what you told me

1 0

[Boost.Filesystem] crashes in static destruction of global string object
by Nguyen Huu Hai 10 Oct '16

10 Oct '16

Hello, libs/filesytem/src/portability.cpp defines valid_posix as a std::string object instead of a const char * pointer. May I know if there are reasons for that? I'm using Boost.Filesystem as a shared library and my processes crash on exit with stack traces pointing to the static destruction of valid_posix. Making it const char* fixes the crashes. Regards, Hai

2 2

[ASIO] Do coroutines allow serving multiple clients at once?
by mmarkd9＠libero.it 10 Oct '16

10 Oct '16

I'm using ASIO with coroutines2 to create a ssl echo server. The code I've been experimenting is the following: http://nopaste.linux-dev.org/?1120051 While this is highly readable compared to the asynchronous callbacks version (thanks coroutines), I'd like to know if this code is able to serve multiple clients connecting at the same time to the server. In theory, every time I return control to the io_service object (i.e. I yield in the spawn coroutine) it should take care of whatever connection is pending / buffered / scheduled to be accepted next, right? Is this a reliable way to handle multiple requests? I don't plan on supporting massive workloads (e.g. 500k connections at once) but having some resilience would be nice indeed. Any other thoughts/tips are also warmly welcome.

2 1

Epoch to date
by Philippe de Rochambeau 09 Oct '16

09 Oct '16

Hello, how do you convert an Unix epoch (e.g., 1476015285) to a formatted date string (e.g., 1984-01-01) with Boost? Many thanks. Philippe

2 1

BoostLog and BoostASIO don't seem to play nice together
by james 08 Oct '16

08 Oct '16

Hi, I'm facing pretty much the same issue as was reported here in 2014: http://boost.2283326.n4.nabble.com/Boostlog-and-ASIO-crash-in-shared-librar… In short, I have a .so that wraps BoostASIO to create a HTTP client. Loading this .so into a program that also uses BoostLog results in either a segfault or no callbacks being fired (even async_resolve which I believe should eventually time-out). Removing all uses of and linking to BoostLog in the main program makes it all work again. Before I try to produce a minimum example, is this a known issue? Code is built with BOOST_LOG_DYN_LINK, asio is header only. We're using boost 1.55, and g++ (Ubuntu 4.8.4-2ubuntu1~14.04.3) 4.8.4 Thanks James

1 0

graph strongly connected components
by Mika S 07 Oct '16

07 Oct '16

I am trying to use the implementation for strongly connected components http://www.boost.org/doc/libs/1_61_0/libs/graph/doc/strong_components.html Is this the iterative or the recursive implementation of tarjan's algorithm ?

2 2

GCC 5.3.0 failing Features Check when building Boost
by Carlos Ferreira 06 Oct '16

06 Oct '16

Hello I'm having some issues regarding the features check, when build Boost using OpenWRT GCC 5.3.0 cross-compiler for armv7. I'm building against musl. I have passed -std=c++14 to the cxxflags, but it seems to have no effect. This is the features output > - 32-bit : yes > - arm : yes > - symlinks supported : yes > - C++11 mutex : yes > - lockfree boost::atomic_flag : no > - Boost.Config Feature Check: cxx11_auto_declarations : no > - Boost.Config Feature Check: cxx11_constexpr : no > - Boost.Config Feature Check: cxx11_defaulted_functions : no > - Boost.Config Feature Check: cxx11_final : no > - Boost.Config Feature Check: cxx11_hdr_tuple : no > - Boost.Config Feature Check: cxx11_lambdas : no > - Boost.Config Feature Check: cxx11_noexcept : no > - Boost.Config Feature Check: cxx11_nullptr : no > - Boost.Config Feature Check: cxx11_rvalue_references : no > - Boost.Config Feature Check: cxx11_template_aliases : no > - Boost.Config Feature Check: cxx11_thread_local : no > - Boost.Config Feature Check: cxx11_variadic_templates : no > - has_icu builds : no > warning: Graph library does not contain MPI-based parallel components. > note: to enable them, add "using mpi ;" to your user-config.jam > - zlib : yes > - iconv (libc) : no > - iconv (separate) : yes > - icu : no > - icu (lib64) : no > - native-atomic-int32-supported : yes > - pthread-supports-robust-mutexes : yes > - compiler-supports-visibility : yes > - gcc visibility : yes > - long double support : yes > - zlib : yes What am I missing, so that all those cxx_* features can be checked? -- Carlos Miguel Ferreira Researcher at Telecommunications Institute Aveiro - Portugal Work E-mail - cmf(a)av.it.pt Skype & GTalk -> carlosmf.pt(a)gmail.com LinkedIn -> http://www.linkedin.com/in/carlosmferreira

1 1

[regex] link error on upgrade to 1.62.0
by James Walker 05 Oct '16

05 Oct '16

Hi, I'm trying to update from 1.60.0 to 1.62.0 on Mac with Xcode 4.6.3, and I'm getting link errors on Regex. I am definitely compiling all 17 of the .cpp files in libs/regex/src. The errors are shown below. Any idea what I might be missing? Undefined symbols for architecture i386: "boost::re_detail_106200::perl_matcher<char const*, std::allocator<boost::sub_match<char const*> >, boost::regex_traits<char, boost::cpp_regex_traits<char> > >::construct_init(boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::_match_flags)", referenced from: boost::re_detail_106200::perl_matcher<char const*, std::allocator<boost::sub_match<char const*> >, boost::regex_traits<char, boost::cpp_regex_traits<char> > >::perl_matcher(char const*, char const*, boost::match_results<char const*, std::allocator<boost::sub_match<char const*> > >&, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::_match_flags, char const*) in BVHJointMap.o boost::re_detail_106200::perl_matcher<char const*, std::allocator<boost::sub_match<char const*> >, boost::regex_traits<char, boost::cpp_regex_traits<char> > >::perl_matcher(char const*, char const*, boost::match_results<char const*, std::allocator<boost::sub_match<char const*> > >&, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::_match_flags, char const*) in AnimationsManager.o "boost::re_detail_106200::perl_matcher<char const*, std::allocator<boost::sub_match<char const*> >, boost::regex_traits<char, boost::cpp_regex_traits<char> > >::find()", referenced from: bool boost::regex_search<char const*, char, boost::regex_traits<char, boost::cpp_regex_traits<char> > >(char const*, char const*, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::_match_flags) in BVHJointMap.o "boost::re_detail_106200::perl_matcher<char const*, std::allocator<boost::sub_match<char const*> >, boost::regex_traits<char, boost::cpp_regex_traits<char> > >::match()", referenced from: bool boost::regex_match<char const*, std::allocator<boost::sub_match<char const*> >, char, boost::regex_traits<char, boost::cpp_regex_traits<char> > >(char const*, char const*, boost::match_results<char const*, std::allocator<boost::sub_match<char const*> > >&, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::_match_flags) in AnimationsManager.o ld: symbol(s) not found for architecture i386

1 1

boost::asio cross-library link issue
by Stanley Friesen 05 Oct '16

05 Oct '16

We are using Boost 1.54. The issue i have is a strange problem with the ASIO library. We have two libraries build using ASIO from Boost 1.54 (libpcl and one of our own libraries). For some reason, despite being built with the same version of boost, the layout of class task_io_service differs between the two libraries. The member function instantiations are in the libraries as weak symbols, so we do not get a link error, but which version of each function that is called from any given place is unpredictable, and varies when we change out own code. In particular, at least op_queue_ and all following members have offsets that differ by 8 bytes (the size of a 64-bit pointer) between the two builds. The result is that sometimes the op_queue back pointer is corrupted (by the other version setting stopped_), or the stop flag is not seen because it was set by the other build. The result is either that run() returns immediately without doing anything, or out application crashes from trying to dereference an invalid pointer. (Often what is supposed to be a NULL back pointer is set to 1, which does not compare equal to NULL). I have looked for conditionally compiled elements in task_io_service, op_queue, and epoll_reactor with no luck. I can see no reason the two library builds should end up with different class layouts. So, does anybody have any clue what might be happening here? Or how to go about fixing the problem? [I hesitate to try forcing strong symbols in our library, since this would break libpcl, due to many functions being inline]. -- View this message in context: http://boost.2283326.n4.nabble.com/boost-asio-cross-library-link-issue-tp46… Sent from the Boost - Users mailing list archive at Nabble.com.

5 8