November 2010 - Boost-users

Re: [Boost-users] Boost-users Digest, Vol 2531, Issue 1
by Dhirendra Pal Singh 03 Nov '10

03 Nov '10

> > On Tue, Nov 2, 2010 at 10:06 AM, Dhirendra Pal Singh <s9_rose(a)yahoo.com> > wrote: > > Hi Folks, > > If this question has been asked already, I apologise, > but I have to go back to my group and present the answer. So > please bear with me. > > > > In my current project we are using boost extensively, > with exceptions disabled, until last week, when I started > using boost threads. I get the following error when > compiling... > > > > *************** > > > deepsin@deepsin-lnx01:~/code/boost_thread/boost_threads_eg1$ > g++ -fno-exceptions -g -I /usr/local/include/boost/ t1.cpp > -o t1 -lboost_thread > > In file included from > /usr/local/include/boost/exception_ptr.hpp:9, > > ? ? ? ? ? ? ? ? from > /usr/local/include/boost/thread/future.hpp:14, > > ? ? ? ? ? ? ? ? from > /usr/local/include/boost/thread.hpp:24, > > ? ? ? ? ? ? ? ? from t1.cpp:3: > > > /usr/local/include/boost/exception/detail/exception_ptr.hpp:17:2: > error: #error This header requires exception handling to be > enabled. > > ************** > > > > This means that any header file that includes, > thread.hpp can't be compiled with -fno-exceptions. I have > couple of questions if someone could help me with, I would > appreciate. > > It appears that this is due to a dependency of Boost Thread > on Boost > Exception's exception_ptr. Somebody should look into this > and either > remove that dependency or add #ifndef BOOST_NO_EXCEPTIONS > around the > relevant code (unfortunately I don't have time right now to > look into > this myself.) > > Emil Dotchevski > Reverge Studios, Inc. > http://www.revergestudios.com/reblog/index.php?n=ReCode > Thanks for a response Emil. Do you think you can spare some time for hand holding, so that I can make this change... :) Thanks PS: I love this link on your webpage. http://www.revergestudios.com/reblog/uploads/ReCode/languages1.jpg. Everybody in my group got a kick out of it.

1 0

Polymorphic serialization with polymorphic archives problems... [part 2]
by Bogdan 03 Nov '10

03 Nov '10

First I would like to confirm I tested Robert Ramey's fix (@60273) for multiple registrations and it looks good in my runs. Thank you Robert! On the other hand, I am under the impression there is a bug in void_caster::recursive_unregister(), although I am not 100% certain about the reason: that function member is called in the destructors of the void caster template classes and makes use of the void_caster_registry singleton (more specifically the type_set). I believe it searches the set for all the void caster entries that are descendants of the current one (i.e., the one being destroyed) and performs the following ops: - 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 traversal repeats for whatever process-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 sure about the reason or whether a simple unregistration is enough. I believe this is (again) a question for Robert, but I thank for any feedback I might get. Bogdan

2 1

[Graph] how to find all shortest paths (not just one of them)?
by Erik Sjölund 03 Nov '10

03 Nov '10

Hi, I managed to record a shortest path between two vertices by doing adjacency_list < vecS, vecS, undirectedS > g; std::vector<Vertex> p(boost::num_vertices(g)); Vertex vertex = *(boost::vertices(g).first); boost::breadth_first_search(g, vertex, boost::visitor( boost::make_bfs_visitor( boost::record_predecessors(&p[0], boost::on_tree_edge()) ) ) ); The shortest path can be found by inspecting the vector p. But my goal is slightly different. Does anyone know how to "record" all shortest paths between two vertices? (i.e. all paths having the minimal path length) cheers, Erik Sjölund

3 2

bjam and msvc PDB files destination directory?
by Martin B. 03 Nov '10

03 Nov '10

Hi! Is it possible to tell bjam to put the generated PDB files (MSVC) alongside the dll/lib files into the stage directory? Here's what I'm using so far: (Boost 1.44.0) bjam.exe --build-type=complete runtime-link=shared variant=release toolset=msvc-8.0 threading=multi debug-symbols=on debug-store=database define=_SECURE_SCL=0 --with-regex Is there an additional option that I can pass on the command line to have the (correctly generated) PDB files also copied into the stage directory? thanks! - Martin

1 0

[proto] : Using proto with NVIDIA CUDA nvcc (redux)
by Manjunath Kudlur 03 Nov '10

03 Nov '10

I remember there were postings in this mailing list before about trying to using proto with NVIDIA's CUDA compiler, and Joel Falcou has contributed some code to make proto work with nvcc. Well, I was trying few experiments with the most recent CUDA compiler release (3.2). I believe one of the exciting things about using proto with CUDA is the ability to write mini programs in different DSELs using proto and ship the mini programs over to the "device" side and execute it there. This idiom will look as follows : template<typename F> __global__ void kernel(F f) { f(...); } kernel<<<blocksize, gridsize>>>(a+b /* a big proto expression */); I tried to write the simplest form of the above use case : #include <boost/proto/proto.hpp> namespace proto=boost::proto; template<typename> struct expr; struct domain : proto::domain<proto::pod_generator<expr> > {}; template<typename E> struct expr { BOOST_PROTO_BASIC_EXTENDS(E, expr<E>, domain) }; template<typename F> __global__ void kernel(F f) { } struct term {}; expr<proto::terminal<term>::type> a = {{{}}}; expr<proto::terminal<term>::type> b = {{{}}}; expr<proto::terminal<term>::type> c = {{{}}}; int main() { kernel<<<1,1>>>(proto::deep_copy(b+c)); return 0; } Note the deep_copy when passing proto expression to kernel functions. Since proto captures terminals by reference, the host side pointers that are held in proto expressions won't make sense on the device side, so at the very least, we have to make copies of terminals before passing to the device. Now, compiling the above program with nvcc (CUDA 3.2, host compiler gcc 4.4.3, Ubuntu 10.04) results in the following error (besides spewing a bunch of warnings) : $ nvcc -v -keep t.cu -I $HOME/boost/boost-trunk/ ... t.cudafe1.stub.c: In function ‘void __device_stub__Z6kernelI4exprIN5boost5proto7exprns_10basic_exprINS2_3tag4plusENS2_7argsns_5list2IS0_INS4_IS6_NS8_IS0_INS4_INS5_8terminalENS7_4termI4termEELl0EEEESE_EELl2EEEES0_INS4_IS9_NSA_IiEELl0EEEEEELl2EEEEEvT_(_Z4exprIN5boost5proto7exprns_10basic_exprINS1_3tag4plusENS1_7argsns_5list2IS_INS3_IS5_NS7_IS_INS3_INS4_8terminalENS6_4termI4termEELl0EEEESD_EELl2EEEES_INS3_IS8_NS9_IiEELl0EEEEEELl2EEEE&)’: t.cudafe1.stub.c:8: error: invalid cast from type ‘const expr<boost::proto::exprns_::basic_expr<boost::proto::tag::address_of, boost::proto::argsns_::list1<expr<boost::proto::exprns_::basic_expr<boost::proto::tag::plus, boost::proto::argsns_::list2<expr<boost::proto::exprns_::basic_expr<boost::proto::tag::plus, boost::proto::argsns_::list2<expr<boost::proto::exprns_::basic_expr<boost::proto::tag::terminal, boost::proto::argsns_::term<term>, 0l> >, expr<boost::proto::exprns_::basic_expr<boost::proto::tag::terminal, boost::proto::argsns_::term<term>, 0l> > >, 2l> >, expr<boost::proto::exprns_::basic_expr<boost::proto::tag::terminal, boost::proto::argsns_::term<int>, 0l> > >, 2l> >&>, 1l> >’ to type ‘char*’ t.cudafe1.stub.c:8: error: invalid cast from type ‘const expr<boost::proto::exprns_::basic_expr<boost::proto::tag::address_of, boost::proto::argsns_::list1<expr<boost::proto::exprns_::basic_expr<boost::proto::tag::plus, boost::proto::argsns_::list2<expr<boost::proto::exprns_::basic_expr<boost::proto::tag::plus, boost::proto::argsns_::list2<expr<boost::proto::exprns_::basic_expr<boost::proto::tag::terminal, boost::proto::argsns_::term<term>, 0l> >, expr<boost::proto::exprns_::basic_expr<boost::proto::tag::terminal, boost::proto::argsns_::term<term>, 0l> > >, 2l> >, expr<boost::proto::exprns_::basic_expr<boost::proto::tag::terminal, boost::proto::argsns_::term<int>, 0l> > >, 2l> >&>, 1l> >’ to type ‘size_t’ # --error 0x1 -- The CUDA compiler works by splitting the original C++ code into "host" code and "device" code. The device code is completely lowered to C. So it seems like the above error is when passing a proto expression to the C-lowered version of device code. The proto expression is a simple POD type, so I am perplexed by the above error. (BTW, when the host compiler is gcc 4.1.3, the above program causes a segmentation fault in gcc, but that's for another forum). This might well be CUDA compiler's problem, but I dug a little deeper. I transformed the proto expression into a simpler POD type as follows : #include <boost/proto/proto.hpp> namespace proto=boost::proto; struct filler {}; template<typename tag, typename C0, typename C1=filler> struct myexpr { C0 c0; C1 c1; }; struct myadd {}; struct myterm {}; struct xform_callable : proto::callable { template<typename> struct result; template<typename This, typename LHS> struct result<This(LHS&)> { typedef myexpr<myterm, LHS> type; }; template<typename LHS> typename result<xform_callable(LHS&)>::type operator()(LHS &l) { myexpr<myterm, LHS> e = {l}; return e; } template<typename This, typename tag, typename LHS, typename RHS> struct result<This(tag, LHS, RHS)> { typedef myexpr<tag, LHS, RHS> type; }; template<typename tag, typename LHS, typename RHS> typename result<xform_callable(tag, LHS, RHS)>::type operator()(tag, LHS l, RHS r) { myexpr<tag, LHS, RHS> e = {l, r}; return e; } }; struct xform : proto::or_< proto::when<proto::plus<proto::_, proto::_>, xform_callable(myadd(), xform(proto::_left), xform(proto::_right))> , proto::when<proto::terminal<proto::_>, xform_callable(proto::_value)> > {}; template<typename> struct expr; struct domain : proto::domain<proto::pod_generator<expr> > {}; template<typename E> struct expr { BOOST_PROTO_BASIC_EXTENDS(E, expr<E>, domain) }; template<typename F> __global__ void kernel(F f) { } struct term {}; expr<proto::terminal<term>::type> a = {{{}}}; expr<proto::terminal<term>::type> b = {{{}}}; expr<proto::terminal<term>::type> c = {{{}}}; int main() { kernel<<<1,1>>>(xform()(b+c)); return 0; } The above program compiles fine with CUDA 3.2. That leaves me with the question : what does a proto expression contain more than myexpr above, that causes nvcc to fail? Again, this might be CUDA compiler's problem and they are the ones that should fix it. However, if someone in this forum can think of a simple fix to proto (or some other boost library that is actually causing the problem), that would be great too. The uniform template expansion across device and host code in CUDA programs is a powerful feature, and I envision some cool libraries that can be written using proto with CUDA. So making proto work with CUDA is a big win. Manjunath

5 12

[fusion] Are Sequence stream IO operators recursive?
by Hugh Dickinson 03 Nov '10

03 Nov '10

I'm trying to output an adapted fusion sequence (PMT::QE::DataSet<PMT::QE::Measurement>) containing other adapted fusion sequences to std::cout. I had hoped for recursive behaviour, such that the contents of each sub-sequence are printed in turn. Having looked at the code in $BOOST_ROOT/boost/fusion/sequence/io/detail/out.hpp it looks like: ---- PMT::QE::DataSet<PMT::QE::Measurement> output; ... Fill output ... std::cout << output << std::endl; ---- should work. Unfortunately, I'm getting compilation errors like: no match for ‘operator<<’ in ‘os << boost::fusion::operator* [with Iterator = boost::fusion::basic_iterator<boost::fusion::struct_iterator_tag, boost::fusion::assoc_struct_category, const PMT::QE::DataSet<PMT::QE::Measurement>, 0>](((const boost::fusion::iterator_base<boost::fusion::basic_iterator<boost::fusion::struct_iterator_tag, boost::fusion::assoc_struct_category, const PMT::QE::DataSet<PMT::QE::Measurement>, 0> >&)((const boost::fusion::iterator_base<boost::fusion::basic_iterator<boost::fusion::struct_iterator_tag, boost::fusion::assoc_struct_category, const PMT::QE::DataSet<PMT::QE::Measurement>, 0> >*)((const boost::fusion::basic_iterator<boost::fusion::struct_iterator_tag, boost::fusion::assoc_struct_category, const PMT::QE::DataSet<PMT::QE::Measurement>, 0>*)first))))’ The other errors simply replace the index 0 with 1 and 2 (PMT::QE::DataSet has 3 elements). On the other hand, the following does work: ---- PMT::QE::DataSet<PMT::QE::Measurement> output; ... Fill output ... std::cout << at_c<0>(output) << std::endl; std::cout << at_c<1>(output) << std::endl; std::cout << at_c<2>(output) << std::endl; ---- Is this the expected behaviour? What am I missing? Is there a way to achieve the desired recursive behaviour? Many thanks, Hugh

4 5

boost threads without exceptions...
by Dhirendra Pal Singh 03 Nov '10

03 Nov '10

Hi Folks, If this question has been asked already, I apologise, but I have to go back to my group and present the answer. So please bear with me. In my current project we are using boost extensively, with exceptions disabled, until last week, when I started using boost threads. I get the following error when compiling... *************** deepsin@deepsin-lnx01:~/code/boost_thread/boost_threads_eg1$ g++ -fno-exceptions -g -I /usr/local/include/boost/ t1.cpp -o t1 -lboost_thread In file included from /usr/local/include/boost/exception_ptr.hpp:9, from /usr/local/include/boost/thread/future.hpp:14, from /usr/local/include/boost/thread.hpp:24, from t1.cpp:3: /usr/local/include/boost/exception/detail/exception_ptr.hpp:17:2: error: #error This header requires exception handling to be enabled. ************** This means that any header file that includes, thread.hpp can't be compiled with -fno-exceptions. I have couple of questions if someone could help me with, I would appreciate. Q1). https://svn.boost.org/trac/boost/ticket/2100, actually says that threads can be compiled without exceptions. Unfortunately its not the case, as I looked at top-of-trunk codebase and the problem still exists. Is this true? I mean threads require exceptions for a fact that is?. If not then can someone help me how to get it to compile without exceptions, and discard the reset of my email please. Q2). If the answer to Q1 is "yes boost threads require exceptions", then was the decision conscious, given a ticked, was opened and closed specifically for this purpose. Q3). Also any idea what is needed to make boost threads work without exception. I mean in terms of man-hours. If my group is okay perhaps I may be willing to work on that change, submit it back to the community, but would need some sort of understanding of the work required and time needed. Any help would be much appreciated.... Thanks in advance... /dev/deep

3 2

cannot open shared object file
by Mike Douglass 03 Nov '10

03 Nov '10

Downloaded the trunk ( Today Nov 2 ) Did: ./bjam --with-libraries=graph,regex ./bjam compiled OK. mv boost-trunk/stage/lib/* /usr/local/lib mv boost-trunk/boost /usr/include my program plc.cpp compiled OK with the following makefile. g++ -ggdb -O3 plc.cpp -o plc -I./Includes -L/usr/local/lib -lboost_graph -lboost_regex But running plc gave the following error ./plc < graph.xml > out.xml ./plc: error while loading shared libraries: libboost_graph.so.1.45.0: cannot open shared object file: No such file or directory But clearly libboost_graph.so.1.45.0 is there. [george@localhost lib]$ ls -lrt /usr/local/lib total 6768 -rwxr-xr-x 1 root root 1155683 Nov 2 20:57 libboost_regex.so.1.45.0 -rwxr-xr-x 1 root root 1155683 Nov 2 20:57 libboost_regex.so -rw-r--r-- 1 root root 2563346 Nov 2 20:57 libboost_regex.a -rwxr-xr-x 1 root root 534275 Nov 2 20:57 libboost_graph.so.1.45.0 -rwxr-xr-x 1 root root 534275 Nov 2 20:57 libboost_graph.so -rw-r--r-- 1 root root 923492 Nov 2 20:57 libboost_graph.a [george@localhost lib]$ Whats wrong? Running Centos 5.5 [george@localhost Program]$ uname -a Linux localhost.localdomain 2.6.18-194.el5 #1 SMP Fri Apr 2 14:58:14 EDT 2010 x86_64 x86_64 x86_64 GNU/Linux Thanks

1 0

Multi file test module in Boost.Test
by Guillaume Dumont 03 Nov '10

03 Nov '10

Hi, I have been trying to get a multi file test module to execute without success. I have the following setup: // main.cpp #define BOOST_TEST_MODULE example #include <boost/test/unit_test.hpp> // suite1.cpp #include <boost/test/unit_test.hpp> BOOST_AUTO_TEST_SUITE(suite1); BOOST_AUTO_TEST_CASE(test1) { ... } BOOST_TEST_SUITE_END(); // suite2.cpp #include <boost/test/unit_test.hpp> BOOST_AUTO_TEST_SUITE(suite2); BOOST_AUTO_TEST_CASE(test2) { ... } BOOST_TEST_SUITE_END(); But only one of the two test cases runs. If I put the suites into the same .cpp file all the test cases are run. Is it possible to link against the static library variant of boost test and have a multi file test module? From what I understand it should be, but I can just find how. Thank you. -- Guillaume Dumont ========================= dumont.guillaume(a)gmail.com

1 0

[proto] vec3 sample error
by Kamil Zubair 02 Nov '10

02 Nov '10

I found a strange error when I try the vec3 sample from proto, my code is like this: struct Vec3SubscriptCtx : proto::callable_context< Vec3SubscriptCtx const > { typedef int result_type; Vec3SubscriptCtx(int i) : i_(i) {} // Index array terminals with our subscript. Everything // else will be handled by the default evaluation context. int operator ()(proto::tag::terminal, int const (&arr)[3]) const { return arr[this->i_]; } int i_; }; struct Vec3 : proto::extends<proto::terminal<int[3]>::type, Vec3> { explicit Vec3(int i=0, int j=0, int k=0) { (*this)[0] = i; (*this)[1] = j; (*this)[2] = k; } int &operator [](int i) { return proto::value(*this)[i]; } int const &operator [](int i) const { return proto::value(*this)[i]; } // Here we define a operator = for Vec3 terminals that // takes a Vec3 expression. template< typename Expr > Vec3 &operator =(Expr const & expr) { typedef Vec3SubscriptCtx const CVec3SubscriptCtx; (*this)[0] = proto::eval(proto::as_expr(expr), CVec3SubscriptCtx(0)); (*this)[1] = proto::eval(proto::as_expr(expr), CVec3SubscriptCtx(1)); (*this)[2] = proto::eval(proto::as_expr(expr), CVec3SubscriptCtx(2)); return *this; } template< typename Expr > Vec3 (Expr const & expr) { typedef Vec3SubscriptCtx const CVec3SubscriptCtx; (*this)[0] = proto::eval(proto::as_expr(expr), CVec3SubscriptCtx(0)); (*this)[1] = proto::eval(proto::as_expr(expr), CVec3SubscriptCtx(1)); (*this)[2] = proto::eval(proto::as_expr(expr), CVec3SubscriptCtx(2)); } void print() const { std::cout << '{' << (*this)[0] << ", " << (*this)[1] << ", " << (*this)[2] << '}' << std::endl; } }; int main(int argc, const char* argv[]) { Vec3 a(1, 2, 3); Vec3 b(3, 2, 1); Vec3 c; c = b; c.print(); return 0; } I expected that when c is printed it will display 3, 2, 1 ( the value of b ) but instead it print 0,0,0 like b is never assigned to c. I try this using vs 2010 with boost 1.44 in windows 7 64.

2 1