- Boost-users - lists.boost.org

[asio] __sync_lock_test_and_set problem, cross-compiling for ARM
by Igor R 09 Nov '10

09 Nov '10

Hello, I'm trying to cross-compile my project for Android (Win, cygwin, crystax-ndk). The linker emits the following error: ../asio/detail/gcc_fenced_block.hpp:41: undefined reference to `__sync_lock_test_and_set_4' I understand that in 1.44 I can disable this "fenced-block" by defining BOOST_ASIO_DISABLE_FENCED_BLOCK, but I'm not sure it's safe to do this. IUUC, __sync_lock_test_and_set gcc function is unavailable for some older arm-based platforms, but should be available for the newer ones, shouldn't it? So maybe I have to define some compiler flag to enable it? I'd appreciate any idea. Thanks.

2 1

[fusion][Functional] LValue arguments limitation
by alfC 09 Nov '10

09 Nov '10

Fusion manual says: Only LValue arguments are accepted. To overcome this limitation, apply Boost.Functional/Forward. I understand that this has to do with not being able to pass literal constants to the function, e.g. unfused<f> uf; uf(1., 2.); //can't First: 1) can anybody explain why is this? 2) uf( (double const&) 1., (double const&) 2. ); seems to work, but it is probably a hack. (not portable maybe?) 3) the manual says also to use Boost.Functional/Forward to overcome this limitation. What does it mean by this? Thank you, Alfredo

3 2

unable to load the module from simple example
by Cuma Tyszkiewicz 09 Nov '10

09 Nov '10

Hello I have followed the simple example from the python boost wiki page: my source file py_moddef.cc: ___________________________________________ #include <string> #include <boost/python.hpp> char const* greet() { return "hello, world"; } BOOST_PYTHON_MODULE(hello_ext) { using namespace boost::python; def("greet", greet); } ____________________________________________ In the next step I have built the shared library 1. g++ -c -fpic -I/usr/include/python2.6/ py_moddef.cc 2. ld -G py_moddef.o -o libpy_moddef.so The libpy_moddef.so file I have copied to the directory which next was included to PYTHONPATH >>> import sys >>> sys.path.append("the_path") I am not able to import the module "hello_ext" defined in py_moddef.cc, but I am able to import each other python module from this path, so something wrong is with the shared library. >>> import hello_ext ImportError: No module named hello_ext Please explain to me what is wrong. I will be gratefull for any help. Best regards Cuma

2 1

[xpressive] Using it with input iterators or how far the backtracing will go back
by Alex Dubov 09 Nov '10

09 Nov '10

Greetings. There exists a fairly common need to apply regexp matching to streams, rather than to blocks of data. Streams are represented by input iterators, which have no ability to decrement their value. On the other hand, xpressive requires a bidirectional iterator, to go back and forth as required by backtracking (if I understand correctly). One approach to work around this limitation is to create an iterator adapter, which will store a copy of every character delivered by input iterator and provide expressive with bidi iterator to work with. So the question is, how can one establish the size for this iterator adapter's internal storage so as to avoid match failures due to incomplete storage and avoid storing too much unnecessary data? I suppose, the storage can be reset when the match is reported. This, however, still leaves an open problem for situations where matches are separated by, possibly, gigabytes of non-matching data. Alternatively, may there be a way to relax the iterator requirements of the xpressive library itself? If xpressive could be made to store partial match data within its data structures, bidi iterators would not be strictly necessary. Thanks.

6 12

[GIL] any_image difference
by Olivier Tournaire 08 Nov '10

08 Nov '10

Hi all, What is the best way to compute the difference between 2 any_image? Of course, the input any_image may not be of the same type... So, if pixels are unsigned int for the 1st one and double for the 2nd one, the result must be stored as double. I (accidently) found pixel_minus_t in pixel_numeric_operations.hpp but cannot figure how to use it. Hope you could help. Regards, Olivier

2 6

signals2::slot_type & g++ anti-aliasing warning
by Ákos Maróy 08 Nov '10

08 Nov '10

Hi, I bumped into an issue, where g++ would give me an anti-aliasing warning for very basic / straightforward signals2 code. consider the following sample code: #include <iostream> #include <boost/bind.hpp> #include <boost/signals2.hpp> typedef boost::signals2::signal<void (void)> sig_t; void ff(void) { std::cout << "ff" << std::endl; } int main(void) { sig_t::slot_type ss = boost::bind(&ff); return 0; } when compiled with the -Wall and -O2 flags, I get the following compiler warning: ... sample.cxx:13: instantiated from here /home/maroy/src/mxp/usr/include/boost/function/function_base.hpp:321: warning: dereferencing type-punned pointer will break strict-aliasing rules /home/maroy/src/mxp/usr/include/boost/function/function_base.hpp:325: warning: dereferencing type-punned pointer will break strict-aliasing rules (I can provide the full compiler output if needed) this is with boost 1.44 on a linux x86_64, g++ version 4.4.3 omitting either the -O2 or the -Wall flags will make the warning go away. what am I doing wrong? Akos

3 4

[polygon] Valgrind errors when normalizing any angle polygons
by Josh Pieper 08 Nov '10

08 Nov '10

Hello, We have been using the 1.44.0 release of polygon for some internal projects and have come across a number of oddities. The first one I've got is a valgrind error when normalizing polygons with numerically large coordinate values. Using the gmp_override.hpp header does resolve these problems, but my understanding is that it is the intent for the library to merely be numerically imprecise, rather than crash, for coordinates which are too large. Our simplified test case is below, and was compiled on an AMD64 Ubuntu 10.4 system. The first call to Normalize has no problems, but the second call causes valgrind to report both uninitialized memory usage and invalid reads. (I can produce the valgrind log on demand, but it is somewhat large for inclusion in an email). While we have deployed 1.44.0 internally, I tested against the most recent svn version and the problem still exists. Is this merely a mis-use of the API? Regards, Josh Pieper ------------------------ #include <stdint.h> #include "boost_polygon/r66294/polygon/polygon.hpp" typedef boost::polygon::point_data<int32_t> BoostPoint; typedef boost::polygon::polygon_data<int32_t> BoostPolygon; typedef boost::polygon::polygon_set_data<int32_t> BoostPolygonSet; typedef boost::polygon::polygon_with_holes_data<int32_t> BoostPolygonWithHoles; typedef std::vector<BoostPolygonWithHoles> BoostPolygonWithHolesVector; template <typename Array> void Normalize(const Array& data, int32_t scale) { std::vector<BoostPoint> points; for (size_t i = 0; i < sizeof(data) / sizeof(*data); i += 2) { points.push_back(BoostPoint(data[i] / scale, data[i + 1] / scale)); } using namespace boost::polygon::operators; BoostPolygonSet poly_set; poly_set |= BoostPolygon(points.begin(), points.end()); BoostPolygonWithHolesVector normal; poly_set.get(normal); } int main() { int32_t data[] = { 0, 0, -1000000, -1000000, 19000000, -1000000, 18292893, 707106, 8292893, -9292893, 8170784, -9440890, 8078584, -9610754, 8000000, -10000000, 8078584, -10389245, 8170784, -10559109, 8292893, -10707106, 18292893, -20707106, 19000000, -19000000, -1000000, -19000000, 0, -20000000, 0, 0, }; Normalize(data, 10); // Fine. Normalize(data, 1); // Crashes. }

2 5

multi_array and variants
by Alle Meije Wink 08 Nov '10

08 Nov '10

Dear all, This still seems to be a problem that every once in a while someone tries to solve and then stops. It's my turn again... I'm trying to make a wrapper template class for 1,2,...8-dimensional images whose values can be char,short,...double to store brain image statistics. I had a go at this before and followed the suggestions described here http://lists.boost.org/boost-users/2009/08/50732.php What I now have is a class bisArray: ... template<typename T> class bisArray: public bisObject { typedef boost::variant < boost::multi_array_ref<T, 1>, boost::multi_array_ref<T, 2>, boost::multi_array_ref<T, 3>, boost::multi_array_ref<T, 4>, boost::multi_array_ref<T, 5>, boost::multi_array_ref<T, 6>, boost::multi_array_ref<T, 7>, boost::multi_array_ref<T, 8>, boost::multi_array_ref<T, 9> > bisArray_t; bisArray_t _bisArray; ... and a class bisImage, which has a bisArray as well as a pointer to the data as read from file (1-dimensional structure): ... template<typename T> class bisImage: public bisObject { public: /** Constructor using existing array data Reformats array as bisImage */ bisImage ( T* _data, size_t _dimensions, std::vector<size_t> _sizes ): data(_data), dimensions(_dimensions) { this.storage=bisArray<T>( _data, _dimensions, _sizes); this.dummy = T(); } ... When I try to initialise an image read from file using ... bisImage<unsigned char> newimage=bisImage<unsigned char> ((unsigned char *)testimage->data, ndim, dims); ... I get, as the first of many errors, this message bisimage.hpp|33|error: no matching function for call to ‘canabis::bisArray<unsigned char>::bisArray()’| That seems to suggest that I call an empty (default) constructor, which I don't do *anywhere* Has anyone ever tried to do something similar? Or do you have any thought about what I am doing wrong here? Is it multi_array complaining or variants? Many thanks Alle Meije

2 1

boost MPI issue on windows
by Nick Collier 08 Nov '10

08 Nov '10

Hi, I've got an MPI application that uses the boost mpi libraries. It runs fine on OSX and Linux when compiled using both debug and more optimized compiler flags. Unfortunately, on windows, the application only runs when compiled using Visual Studio 2008 debug configuration. When I run the executable compiled in the release configuration it crashes, but not in the same place every time in my code. It does only occur though when I make communicator::send and communicator::recv calls, sending and receiving vectors of ints. In those cases, I get exceptions like: {routine_=0x00d37d38 "MPI_Send" result_code_=805931010 message="MPI_Send: Invalid count, error stack: MPI_Send(176): MPI_Send(buf=0x001BF950, count=-1833296, MPI_PACKED, dest=1, tag=2001, MPI_COMM_WORLD) failed MPI_Send(101): Negative count, value is -1833296" } Sometimes its for a send and sometimes for a receive, but the exception is always for a "negative count" which looks suspiciously like an uninitialized integer. I figured I'm doing something wrong either in how I've compiled the boost libraries or in my own code (although its odd that it works in both linux and osx). Any suggestions obviously appreciated. This is for boost-1.39. We are prevented from using anything newer because the machine we ultimately deploy on has boost-1.39 installed. On windows, I'm using microsoft's mpi implementation (based on MPICH2) under windows 7. The boost mpi libraries are compiled in both static debug and release mt versions. thanks very much, Nick

6 31

[BGL] Unable to sort list of edge_descriptors
by Mike Douglass 08 Nov '10

08 Nov '10

I'm trying to sort a list of edge_descriptors, but I get the compiler error described below. typedef adjacency_list < vecS, vecS, bidirectionalS >::edge_descriptor edgeDescriptor; // defined globally template < typename Graph, typename edgeDescriptor > struct SortByName : public binary_function<edgeDescriptor, edgeDescriptor, bool> { bool operator()(const Graph & g, const edgeDescriptor& a, const edgeDescriptor& b) { return g[a].eName < g[b].eName; } }; template < typename Graph > void Sort_Test(Graph & g) { typedef std::list<edgeDescriptor> edge_list; typename graph_traits<Graph>::edge_iterator edge_iter, edges_end; edge_list A; for (tie(edge_iter, edges_end) = edges(g); edge_iter != edges_end; ++edge_iter) A.push_back(*edge_iter); for (edge_list::iterator i = A.begin(); i != A.end(); ++i) cout << " *i = " << " " << g[*i].eName << "\n"; // If the following line is commented out the program compiles and runs OK. A.sort(SortByName< typename Graph, typename edgeDescriptor >()); //With the above line uncommented the compiler says: /* ./Includes/Utilities.cpp: In function ‘void Sort_Test(Graph&)’: ./Includes/Utilities.cpp:539: error: wrong number of template arguments (1, should be 2) ./Includes/Utilities.cpp:513: error: provided for ‘template<class Graph, class edgeDescriptor> struct SortByName’ make: *** [plc] Error 1 */ } But clearly there are two template arguments - < typename Graph, typename edgeDescriptor > Thanks

4 7