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

Boost serialisation: base ptr regression in 1.42
by Avi Bahra 27 Mar '10

27 Mar '10

I have a serious regression in boost 1.42 serialisation. Now getting a unregistered class, exception, on Suse linux using gcc 4.2.1. This test passes on boost 1.39/1.40. There are 3 files involved to recreate the crash: included below: Main.cpp BaseHolder.hpp BaseHolder.cpp The file BaseHolder.hpp/cpp defines 3 classes Base Derived BaseHolder Here is what I found: 1/ If all the functionality is put in one file. i.e Main.cpp then _no_ crash 2/ If all 3 files are compiled together then _no_ crash 3/ However if BaseHolder is added to static library and linked with Main.cpp then we get the _crash_ ( only in 1.42 ) Also getting a spurious warning message: Again this warning only appears in 1.42 /var/tmp/ma0/boost/boost_1_42_0/boost/mpl/print.hpp: In instantiation of ?boost::mpl::print<boost::serialization::BOOST_SERIALIZATION_STATIC_WARNING_LINE<98> >?: /var/tmp/ma0/boost/boost_1_42_0/boost/serialization/static_warning.hpp:92: instantiated from ?boost::serialization::static_warning_test<false, 98>? /var/tmp/ma0/boost/boost_1_42_0/boost/archive/detail/check.hpp:98: instantiated from ?void boost::archive::detail::check_object_tracking() [with T = Derived]? /var/tmp/ma0/boost/boost_1_42_0/boost/archive/detail/oserializer.hpp:313: instantiated from ?static void boost::archive::detail::save_non_pointer_type<Archive>::invoke(Archive&, T&) [with T = Derived, Archive = boost::archive::text_oarchive]? /var/tmp/ma0/boost/boost_1_42_0/boost/archive/detail/oserializer.hpp:525: instantiated from ?void boost::archive::save(Archive&, T&) [with Archive = boost::archive::text_oarchive, T = Derived]? /var/tmp/ma0/boost/boost_1_42_0/boost/archive/detail/common_oarchive.hpp:69: instantiated from ?void boost::archive::detail::common_oarchive<Archive>::save_override(T&, int) [with T = Derived, Archive = boost::archive::text_oarchive]? /var/tmp/ma0/boost/boost_1_42_0/boost/archive/basic_text_oarchive.hpp:80: instantiated from ?void boost::archive::basic_text_oarchive<Archive>::save_override(T&, int) [with T = Derived, Archive = boost::archive::text_oarchive]? /var/tmp/ma0/boost/boost_1_42_0/boost/archive/detail/interface_oarchive.hpp:64: instantiated from ?Archive& boost::archive::detail::interface_oarchive<Archive>::operator<<(T&) [with T = Derived, Archive = boost::archive::text_oarchive]? /var/tmp/ma0/clientRoot/workspace/MyProject/SCRATCH/src/Main.cpp:16: instantiated from here /var/tmp/ma0/boost/boost_1_42_0/boost/mpl/print.hpp:55: warning: comparison between signed and unsigned integer expressions gcc.archive bin/gcc-4.2.1/debug/link-static/libnodeattr.a gcc.link bin/gcc-4.2.1/debug/test Here are the 3 files: // =========== file Main.cpp ============================ #include <iostream> #include <fstream> #include "BaseHolder.hpp" using namespace std; int main() { std::string fileName = "test.txt"; { // If this scope is commented out, crash goes away ??? Derived saved("varname"); { std::ofstream ofs( fileName.c_str() ); boost::archive::text_oarchive oa( ofs ); oa << saved; } Derived restored; { std::ifstream ifs( fileName.c_str() ); boost::archive::text_iarchive ia( ifs ); ia >> restored; } assert(saved == restored ); std::remove(fileName.c_str()); } { BaseHolder saved(Derived("varname")); { std::ofstream ofs( fileName.c_str() ); boost::archive::text_oarchive oa( ofs ); oa << saved; } BaseHolder restored; { std::ifstream ifs( fileName.c_str() ); boost::archive::text_iarchive ia( ifs ); ia >> restored; } assert(saved == restored); std::remove(fileName.c_str()); } cout << "Test passed \n"; } //=================== file BaseHolder.hpp ============= #ifndef BASEHOLDER_HPP_ #define BASEHOLDER_HPP_ // defines classes: // Base // Derived // BaseHolder #include <boost/archive/text_iarchive.hpp> #include <boost/archive/text_oarchive.hpp> #include <boost/serialization/base_object.hpp> #include <boost/serialization/serialization.hpp> #include <boost/serialization/vector.hpp> #include <boost/serialization/string.hpp> #include <boost/serialization/assume_abstract.hpp> class Base { public: Base(); virtual ~Base(); virtual bool compare(Base*) const = 0; virtual std::string name() const = 0; private: friend class boost::serialization::access; template<class Archive> void serialize(Archive &, const unsigned int){} }; class Derived : public Base { public: Derived(const std::string& n) : name_(n) {} Derived() {} bool operator==(const Derived& rhs) const; virtual std::string name() const { return name_;} virtual bool compare(Base*) const; private: std::string name_; friend class boost::serialization::access; template<class Archive> void serialize(Archive & ar, const unsigned int) { ar & boost::serialization::base_object<Base>(*this); ar & name_; } }; class BaseHolder { public: BaseHolder( const Derived& ); BaseHolder(); ~BaseHolder(); std::string name() const { return basePtr_->name();} bool operator==(const BaseHolder& rhs) const { return basePtr_->compare(rhs.basePtr_); } private: Base* basePtr_; friend class boost::serialization::access; template<class Archive> void serialize(Archive & ar, const unsigned){ ar & basePtr_; } }; #endif //============= File BaseHolder.cpp ========= // implements class: // Base // Derived // BaseHolder #include <assert.h> #include <sstream> #include <boost/serialization/export.hpp> #include "BaseHolder.hpp" using namespace std; // class Base Base::Base() {} Base::~Base() {} // class derived bool Derived::compare(Base* rb) const { Derived* rhs = dynamic_cast<Derived*>(rb); if(!rhs) return false; return operator==(*rhs); } bool Derived::operator==(const Derived& rhs) const { if (name_ != rhs.name_) { return false; } return true; } BOOST_CLASS_EXPORT(Derived); // class BaseHolder BaseHolder::BaseHolder() : basePtr_(NULL) {} BaseHolder::BaseHolder( const Derived& r) : basePtr_(new Derived(r)) {} BaseHolder::~BaseHolder() { delete basePtr_;} Best regards, Ta, Avi

2 1

[proto] Comma Strangeness
by David Greene 27 Mar '10

27 Mar '10

Well, my little lambda-like experimental language is getting rather large. I'm really having fun with it and am close to actually doing something almost useful with it. :) Unfortunately I've run into a problem I can't seem to isolate to a small testcase. I'm doing something like this: proto::display_expr( list [ obj["name1"].attr["a"], obj["name2"].attr["a"], obj["name3"].attr["b"] ] ); Presume I have all of the terminals, members, etc. declared/defined correctly (at least "correctly" as I understand it). display_expr dumps out an expression without any comma operators. It just shows a subscript off "list" with one "index," obj["name3"].attr["b"]. It's the behavior I would expect if the operands of the commas were not protoized expressions, but they _should_ be protoized. The funny thing is, if I write a small testcase just isolating the terminals, members, etc. needed to construct the expression, it works as expected. But in the context of the (much) larger language, it fails. However, the grammar specification shouldn't have any bearing on what display_expr dumps out. So any ideas as to why commas might be gobbled up prematurely? -Dave

3 4

Re: [Boost-users] [MPI] Code example leaks memory
by Brian Wood 27 Mar '10

27 Mar '10

> Hi all, > > the code example given at: http://www.boost.org/doc/libs/1_41_0/doc/html/mpi/tutorial.html > leaks memory as reported by Valgrind 3.5.0 There are several examples/main()'s on that page. I'm not sure which one you're talking about. I think the following paragraph from that page under the "User-defined data types" heading could be improved. "The redundant copy elimination optimization can only be applied when the shape of the data type is completely fixed. Variable-length types (e.g., strings, linked lists) and types that store pointers cannot use the optimiation, but Boost.MPI will be unable to detect this error at compile time. Attempting to perform this optimization when it is not correct will likely result in segmentation faults and other strange program behavior." I suggest the following: The redundant copy elimination optimization should only be applied when the shape of the data type is completely fixed. When a type has variable-length members (e.g., strings, linked lists) or pointers this optimization shouldn't be used as Boost.MPI will be unable to detect this error at compile time. Attempting to perform this optimization when it is not correct will likely result in segmentation faults and other strange program behavior. I changed "can" to "should" in the first sentence and reworked the second sentence, including correcting the spelling of optimization. Brian Wood Ebenezer Enterprises http://webEbenezer.net (651) 251-9384 "Then Samuel took a stone, and set it between Mizpeh and Shen, and called the name of it Ebenezer, saying, Till now has the L-RD helped us."

1 0

issues with cost_map is a_star algo
by Mangal 27 Mar '10

27 Mar '10

I have a working modified Dijkstras algo using boost. My graph has an edge_property called my_edge_property. The type of distance_map is time_duration. I provide a distance_combine class which has a function with signature time_duration my_distance_combine::operator()(const time_duration &timeAtVertex, const my_edge_property &weight) { //My Implementation } distance_infinity has a value pos_infin. This modified version of Disjkstra algorithm works fine and runs correctly. Now I want to use a_star algorithm instead of dijisktra. So I provided a hueristic class with a function time_duration myhuersticClass::operator() (vertex_descriptor u) { //My Implementation } and pass that as the hueristic function to a_star algorithm. I expected this should be enough for everything to work but I get compile errors. the error is in file astar_search.hpp (line 289) put(cost, *ui, inf);// compile error at this line error C2679: binary '=' : no operator found which takes a right-hand operand of type 'const boost::date_time::special_values' (or there is no acceptable conversion) Now in the same file one line above that statement put(distance, *ui, inf); compiles fine. The documentation says that the type of cost_map and distance_map are be the same. If that is the case then why does assiging pos_inifn to cost_map does not work?? Looking at it closely I see that CostInf=boost::date_time::special_values, (which is correct) CostZero=my_edge_property (???? do not understand) Also why is the type of CostZero= my_edge_property (if I try to pass distance_zero as minutes(0) compiler does not like it and complains that time_duration cannot be converted to my_edge_property) Looks like cost_map is vector<my_edge_property> and not vector<time_duration> which is what the distance_map is. I don't see any way to passthe cost_map to a_star function. Any ideas?? -- View this message in context: http://old.nabble.com/issues-with-cost_map-is-a_star-algo-tp28054643p280546… Sent from the Boost - Users mailing list archive at Nabble.com.

1 0

boost filesystem::ifstream for files larger than 2GB on OS X
by lucasvickers 27 Mar '10

27 Mar '10

Hello, I have been using the std::ifstream w/o issue except my data files have grown > 2GB, and as is well documented it is failing. I am on OS X 10.6 and I have my executable compiling/running as a x86_64 (64bit) executable. I read that boost supports a 64bit streamposition in boost::filesystem::ifstream. I have compiled boost using darwin ports (I should state I never specifically said anything about x86_64 during this compilation). I have re-run my program and I have the same problem (I use tellg to confirm the issue is the streamposition). My questions are: 1 - Does the boost ifstream in fact support large files? 2 - If 1) is true, then what specifically needs to be done to get this support? 3 - Unrelated, but what do the "-mt" and "-d" added to the dylib file name mean? (libboost_filesystem-mt-d.dylib) thanks! Lucas -- View this message in context: http://old.nabble.com/boost-filesystem%3A%3Aifstream-for-files-larger-than-… Sent from the Boost - Users mailing list archive at Nabble.com.

2 1

[Bind] How do you write it?
by Robert Jones 27 Mar '10

27 Mar '10

Hi All Ok, now I've found out what is really stopping my real code working. In this example my problem is that method has defaulted arguements, which I think mucks up bind's signature matching. Is there an easy way round this, or must I just add appropriate overloads of double_it to A? Thanks, Rob. #include <algorithm> #include <vector> #include "boost/bind.hpp" struct A { int double_it( int i, int j = 2 ) { return i * j; } } a; std::vector< int > v; int main( ) { for ( std::vector< int >::iterator i = v.begin( ); i != v.end( ); ++ i ) { a.double_it( * i ); } // Write it as a for_each/bind std::for_each( v.begin(), v.end(), boost::bind( & A::double_it, boost::ref( a ), _1 ) ); } -- ACCU - Professionalism in programming - http://www.accu.org

2 2

does gather blocks?
by alev mutlu 27 Mar '10

27 Mar '10

Hi all, I am scattering some data, each client works it share, and finally the master collects all the data and proceeds in its way. The problem is that clients finish their work in different time and I wonder if gather blocks until all workers finish. I am getting segmentation faults, and I am not sure if it is due to synchronization or something else. Any idea?

1 0

[serialization] Potential problems with void_caster
by Bogdan 27 Mar '10

27 Mar '10

I am trying this again, because if it needs a fix it would be nice to make it into the next boost release. I get an access violation after explicitly unloading a DLL that contains serialization code for a series of classes. I believe there might be a bug in void_caster::recursive_unregister(): that function member is called in the destructors of the void caster template classes. It searches the set inside void_caster_registry singleton for all the entries that are descendants of the current one (i.e., the void_caster being destroyed) and performs the following operations: - delete the shortcuts. - remove the corresponding entries from the set. However, void_caster::recursive_unregister() doesn't remove from the registrar the entry corresponding to the current void_caster(). This creates the problem below: Consider a scenario where a dll containing some serialization functionality is loaded explicitly at runtime (i.e., using LoadLibrary() on Windows or dlopen on Unix). There are no parent-child relationships among the serialized classes registered by that DLL: all serialized classes are derived from one abstract base. While unloading the library, its void_caster singletons are destroyed but there are no shortcuts to remove from the global registrar since there are no parent-child relationships. All entries corresponding to those are still in the global registrar and point to the void_caster singletons ALREADY DESTROYED AND DISCARDED. When the main process exits, the registrar repeats the traversal for whatever executable-bound void_caster singletons are about to be destroyed. During these subsequent traversals, void_caster::recursive_unregister() visits the entries left after the previous explicit dll download and tries to identify the parent-child relationship. In doing so, it accesses the dangling vc pointers generating access violations: void_cast.cpp(Line 286): const void_caster * vc = *it; void_cast.cpp(Line 287): if(vc->m_parent == this){ //<--------------------- here is the access violation I can reproduce the crash systematically, but I am not 100% sure about the reason or whether a simple unregistration is enough. I believe this is (again) a question for Robert Ramey, but I thank for any feedback I might get. Bogdan PS On a different tack I would like to confirm I tested Robert fix (@60273) for multiple registrations and it looks good in my runs. Thank you!

2 1

Boost Graph: vertex class has no copy constructor
by Trevor Harmon 26 Mar '10

26 Mar '10

Hi, I'm using the Boost Graph library to construct a directed graph: typedef boost::directed_graph<BasicBlock> Graph; The vertex class, "BasicBlock", is not my own; it comes from the LLVM project. Unfortunately, the declaration above causes a slew of compiler errors (see below) that appear to be related to BasicBlock's copy constructor. Sure enough, I checked BasicBlock.h, and its copy constructor is declared private and is not even implemented. There appears to be some sort of design issue that prevents this class from having a copy constructor. Is there some workaround for this on the Boost Graph side, or must vertex classes always have a copy constructor? Thanks, Trevor /Users/twharmon/Development/boost_1_42_0/boost/pending/property.hpp: In constructor ‘boost::property<Tag, T, Base>::property(const T&) [with Tag = boost::vertex_bundle_t, T = llvm::BasicBlock, Base = boost::no_property]’: Clepsydra.cpp:46: instantiated from here /Users/twharmon/Development/LLVM/llvm-2.7/llvm-2.7-src/include/llvm/ BasicBlock.h:81: error: ‘llvm::BasicBlock::BasicBlock(const llvm::BasicBlock&)’ is private /Users/twharmon/Development/boost_1_42_0/boost/pending/property.hpp: 31: error: within this context /Users/twharmon/Development/boost_1_42_0/boost/pending/property.hpp: In copy constructor ‘boost::property<boost::vertex_bundle_t, llvm::BasicBlock, boost::no_property>::property(const boost::property<boost::vertex_bundle_t, llvm::BasicBlock, boost::no_property>&)’: /Users/twharmon/Development/boost_1_42_0/boost/pending/property.hpp: 22: instantiated from ‘typename boost::adjacency_list<boost::listS, boost::listS, boost::bidirectionalS, boost::property<boost::vertex_index_t, unsigned int, typename boost::graph_detail::vertex_prop<Prop>::type>, boost::property<boost::edge_index_t, unsigned int, typename boost::graph_detail::edge_prop<EdgeProp>::type>, GraphProp, boost::listS>::vertex_descriptor boost::directed_graph<VertexProp, EdgeProp, GraphProp>::add_vertex(const typename boost::graph_detail::vertex_prop<Prop>::type&) [with VertexProp = llvm::BasicBlock, EdgeProp = boost::no_property, GraphProp = boost::no_property]’ /Users/twharmon/Development/boost_1_42_0/boost/graph/ directed_graph.hpp:454: instantiated from ‘typename boost::directed_graph<VertexProp, EdgeProp, GraphProp>::vertex_descriptor boost::add_vertex(const typename boost::directed_graph<VertexProp, EdgeProp, GraphProp>::vertex_property_type&, boost::directed_graph<VP, EP, GP>&) [with VP = llvm::BasicBlock, EP = boost::no_property, GP = boost::no_property]’ Clepsydra.cpp:46: instantiated from here /Users/twharmon/Development/LLVM/llvm-2.7/llvm-2.7-src/include/llvm/ BasicBlock.h:81: error: ‘llvm::BasicBlock::BasicBlock(const llvm::BasicBlock&)’ is private /Users/twharmon/Development/boost_1_42_0/boost/pending/property.hpp: 22: error: within this context cc1plus: warnings being treated as errors /Users/twharmon/Development/boost_1_42_0/boost/pending/property.hpp: In copy constructor ‘boost::property<boost::vertex_index_t, unsigned int, boost::property<boost::vertex_bundle_t, llvm::BasicBlock, boost::no_property> >::property(const boost::property<boost::vertex_index_t, unsigned int, boost::property<boost::vertex_bundle_t, llvm::BasicBlock, boost::no_property> >&)’: /Users/twharmon/Development/boost_1_42_0/boost/pending/property.hpp: 22: warning: synthesized method ‘boost::property<boost::vertex_bundle_t, llvm::BasicBlock, boost::no_property>::property(const boost::property<boost::vertex_bundle_t, llvm::BasicBlock, boost::no_property>&)’ first required here /Users/twharmon/Development/boost_1_42_0/boost/graph/ directed_graph.hpp: In member function ‘typename boost::adjacency_list<boost::listS, boost::listS, boost::bidirectionalS, boost::property<boost::vertex_index_t, unsigned int, typename boost::graph_detail::vertex_prop<Prop>::type>, boost::property<boost::edge_index_t, unsigned int, typename boost::graph_detail::edge_prop<EdgeProp>::type>, GraphProp, boost::listS>::vertex_descriptor boost::directed_graph<VertexProp, EdgeProp, GraphProp>::add_vertex(const typename boost::graph_detail::vertex_prop<Prop>::type&) [with VertexProp = llvm::BasicBlock, EdgeProp = boost::no_property, GraphProp = boost::no_property]’: /Users/twharmon/Development/boost_1_42_0/boost/graph/ directed_graph.hpp:154: warning: synthesized method ‘boost::property<boost::vertex_index_t, unsigned int, boost::property<boost::vertex_bundle_t, llvm::BasicBlock, boost::no_property> >::property(const boost::property<boost::vertex_index_t, unsigned int, boost::property<boost::vertex_bundle_t, llvm::BasicBlock, boost::no_property> >&)’ first required here

3 4

Linking boost in Visual Studio VC++ project
by Federico Golfrè Andreasi 26 Mar '10

26 Mar '10

Hi ! I'm trying to find out how to link boost libraries under windows to a visual studio VC++ project (that will be a .dll library). Where can I find some information ?! I've compiled the boost libraries and followed the instruction found at: http://www.boost.org/doc/libs/1_35_0/more/getting_started/windows.html but when I build the project it gives me an error trying to find a libboost_filesystem-vc80-mt-gd-1_41.lib, but no one libraries build starts with the 'lib' prefix ? Why he search for them ? Thank you, Federico

4 3