- Boost-users - lists.boost.org

2 bind questions
by Angus Leeming 07 Nov '03

07 Nov '03

I'm having some problems with bind and wonder if someone might help. The attached file compiles and runs fine when compiled with boost-1.30.2 (g++ 3.2 on a PC and g++ 3.3 on a Dec Alpha, both running linux). However, compilation fails if I uncomment the overloaded 'node' member function below. struct FEM { typedef std::vector<Node>::size_type size_type; FEM(std::vector<Node> const & nodes) : nodes_(nodes) {} size_type nnodes() const { return nodes_.size(); } Node & node(size_type i) { return nodes_[i]; } // Node const & node(size_type i) const { return nodes_[i]; } private: std::vector<Node> nodes_; }; boost/boost-1.30.2/boost/function/function_template.hpp:117: could not convert `f->boost::_bi::bind_t<R, F, L>::operator()(A1&) [with A1 = size_t, R = const Node&, F = boost::_mfi::cmf1<const Node&, FEM, unsigned int>, L = boost::_bi::list2<boost::reference_wrapper<FEM>, boost::arg<1> >]((&a0))' to `Node&' Is there a fix for this or have I hit a brick wall? Second question. I'd like to combine these two boost::functions into a single boost::function2<void, size_t, double> all_in_one = ...; all_in_one(i, val); but I'm getting horribly lost trying to determine the '...'. Is it possible? Any clues? Kind regards, Angus ps. What is the prefered way of posting code? Inline or as an attachment? A std::vector<Node> nodes(10); FEM fem(nodes); size_t i = 5; double const val = 14; boost::function1<Node &, size_t> get_node = boost::bind(&FEM::node, boost::ref(fem), _1); boost::function1<void, double> set_value = boost::bind(&Node::set_conc, boost::ref(get_node(i)), _1); set_value(val);

2 2

RE: [Boost-users] Re: shared_ptr compiler error with gcc
by Ben Hutchings 06 Nov '03

06 Nov '03

amerio <amerio(a)hotmail.com> wrote: > > Does Nintendo not supply a newer version? Do they even support the > > standard library (which Boost is somewhat dependent on)? > > Actually, it is gcc 2.95.3 > The stl is supported via the latest stlport. <snip> > > This error message seems to be incomplete. > > oops. here it is : > boost/detail/shared_count.hpp: > In method `void * > boost::detail::sp_counted_base_impl<P,D>::get_deleter(const type_info > &)': Internal compiler error. > > If anyone was successful compiling this on the gamecube, > please share your knowledge ! > (boost 1.30.2, gcc 2.95.3 with prodg for gamecube) I think you have disabled RTTI, so the call to typeid() does not work. There is a bug in 2.95.3 that causes it to fail rather than reporting this error in a helpful way: <http://gcc.gnu.org/bugzilla/show_bug.cgi?id=3689>. If you remove the option '-fno-rtti' from the compiler command line it should work.

1 0

RE: [Boost-users] Problems with call_traits in GCC 3.2 20020927(prerelease)
by Ben Hutchings 06 Nov '03

06 Nov '03

Joaquín Mª López Muñoz <joaquin(a)tid.es> wrote: > The following programs does not compile with error > > foo.cpp: In function `int main()': > foo.cpp:11: no matching function for call to `foo(std::string)' > > Program code: > > #include <boost/call_traits.hpp> > #include <string> > > template<typename Type> > void foo(typename boost::call_traits<Type>::param_type x) > { > } > > int main() > { > foo(std::string("bar")); > > return 0; > } > > > Environment is GCC 3.2 20020927 (prerelease) under Cygwin. > Anyvody run into a similar problem? Workarounds, ideas? Thanx > in advance. This is standard behaviour. See my explanation in <http://groups.google.com/groups?selm=slrnbisehj.1pc.do-not-spam-benh%40tin.…>. Without knowing what it is you are trying to do, I cannot offer a workaround.

3 2

RE: [Boost-users] Problems with call_traits in GCC3.220020927(prerelease)
by Ben Hutchings 06 Nov '03

06 Nov '03

Joaquín Mª López Muñoz <joaquin(a)tid.es> wrote: > Ben Hutchings ha escrito: <snip> > > Without knowing what it is you are trying to do, I cannot offer > > a workaround. > > With pleasure. My original design was > > template<typename Type> > void foo(const Type& x) > { > } > > (of course the actual code does something useful) > and I wanted to take advantage of the call_traits<>::param_type > optimization for basic types. > > So, no way? Should I revert to const Type&? Some clever workaround > to help the compiler swallow the construct? Thx in advance. I would leave it as const Type &. If you think your compiler is not smart enough to make a local copy of what x refers to when Type when that's likely to improve performance, you could use something like this: template<typename Type> void foo(const Type & arg_x) { typename boost::call_traits<Type>::param_type local_x = arg_x; ... }

1 0

date_time library problem
by Istvan Buki 06 Nov '03

06 Nov '03

Hello, here is a small piece of code that use the posix_time system that fails to compile. I'm wondering whether it is a bug in the library or my usage of the library that is wrong. Thank you for your help. Istvan #include "boost/date_time/posix_time/posix_time.hpp" #include <iostream> int main() { using namespace boost::posix_time ; ptime now = microsec_clock::local_time(); std::cout << to_simple_string( now ) << std::endl ; return 0 ; } g++ -I/opt/boost brol.cpp /tmp/ccQQjwdW.o(.gnu.linkonce.t._ZN5boost9date_time15month_formatterINS_9gregorian10greg_monthENS0_13simple_formatEE12format_monthERKS3_RSo+0x3a): In function `boost::date_time::month_formatter<boost::gregorian::greg_month, boost::date_time::simple_format>::format_month(boost::gregorian::greg_month const&, std::basic_ostream<char, std::char_traits<char> >&)': : undefined reference to `boost::gregorian::greg_month::as_short_string() const' /tmp/ccQQjwdW.o(.gnu.linkonce.t._ZN5boost9date_time15month_formatterINS_9gregorian10greg_monthENS0_13simple_formatEE12format_monthERKS3_RSo+0x5c): In function `boost::date_time::month_formatter<boost::gregorian::greg_month, boost::date_time::simple_format>::format_month(boost::gregorian::greg_month const&, std::basic_ostream<char, std::char_traits<char> >&)': : undefined reference to `boost::gregorian::greg_month::as_long_string() const' collect2: ld returned 1 exit status The system is running SuSE linux 9.0 compiler: gcc 3.3.1 boost version 1.30.2

2 1

RE: [Boost-users] date_time library problem
by Ben Hutchings 06 Nov '03

06 Nov '03

Istvan Buki <istvan.buki(a)euronet.be> wrote: > here is a small piece of code that use the posix_time system that > fails to compile. > I'm wondering whether it is a bug in the library or my usage of the > library that is wrong. <snip> Library code normally needs to be compiled separately into a library file that is then statically or dynamically linked with the client code. Boost is unusual in that many of its libraries contain only inline and template functions that are compiled along with the client code. However, the date_time, filesystem, graph, python, regex, signals, test, and thread libraries have some code that must be separately compiled. See the file tools/build/index.html in the Boost distribution or on www.boost.org.

1 0

RE: [Boost-users] shared_ptr compiler error with gcc
by Ben Hutchings 06 Nov '03

06 Nov '03

amerio <amerio(a)hotmail.com> wrote: > I'm trying to compile a gamecube sample with boost_1_30_2 and gcc > version egcs-2.91.57 (nintendo package): That's a very old version of GCC. Boost isn't tested with anything older than version 2.95.3, and that only on Linux, with a 41% failure rate. (See <http://boost.sourceforge.net/regression-logs/> for the details.) Does Nintendo not supply a newer version? Do they even support the standard library (which Boost is somewhat dependent on)? There are unofficial builds of GCC 3.x (which Boost works well with) for the Gamecube though I don't know whether they include libstdc++. Google for "gcc gamecube" and you'll find them. > #include <boost/shared_ptr.hpp> > int main() > { > boost::shared_ptr<int> i ( new int ); > } > > I got: > boost/detail/shared_count.hpp: > In method `void * > boost::detail::sp_counted_base_impl<P,D>::get_deleter(const This error message seems to be incomplete.

2 1

libraries based on Boost
by Mohamed Iqbal 06 Nov '03

06 Nov '03

Hi... Is there any library that is based on Boost but not a part of it? Mohammed

2 1

Problems with call_traits in GCC 3.2 20020927 (prerelease)
by Joaquín Mª López Muñoz 06 Nov '03

06 Nov '03

The following programs does not compile with error foo.cpp: In function `int main()': foo.cpp:11: no matching function for call to `foo(std::string)' Program code: #include <boost/call_traits.hpp> #include <string> template<typename Type> void foo(typename boost::call_traits<Type>::param_type x) { } int main() { foo(std::string("bar")); return 0; } Environment is GCC 3.2 20020927 (prerelease) under Cygwin. Anyvody run into a similar problem? Workarounds, ideas? Thanx in advance. Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

1 0

MultiArray - issue with views and striding.
by Nasatyr, Ben (London) 06 Nov '03

06 Nov '03

Can anyone help shed light on the following? Suppose I have a 1d-array of size 10 (index base is 0). If I create a view on [0,10), strided by 3, shouldn't the size of the result be 4, corresponding to indices 0, 3, 6 and 9? Instead I get 3. Please forgive me if I'm off the mark here, as I'm new to boost in general and MultiArray in particular. The relevant line (assuming len => size) in boost/multi_array/base.hpp seems to be index len = (finish - start) / index_factor; (boost::detail::multi_array::multi_array_impl_base<>::generate_array_view() line 427). I would have written index len = 1 + (finish - 1 - start) / index_factor; The following code sample should make all clear. Any clarification greatly appreciated. Cheers, Ben typedef boost::multi_array<double, 1> array_type; typedef array_type::index index; typedef array_type::index_range range; index i; int values = 0; // Setup our base array and data. // A is a 1-d array with size 10, indices 0,1,...,9 and values 0,1,...,9. array_type A(boost::extents[10]); for(i = 0; i != 10; ++i) A[i] = values++; for(i = 0; i != 10; ++i) assert(A[i]==i); assert(A.size()==10); // The following views of A should all be effectively the same, // corresponding to indices 0,3,6,9 of A, so having // size 4, indices 0,1,2,3 and values 0,3,6,9? array_type::array_view<1>::type view1 = A[boost::indices[range(0,12,3)] ]; array_type::array_view<1>::type view2 = A[boost::indices[range(0,11,3)] ]; array_type::array_view<1>::type view3 = A[boost::indices[range(0,10,3)] ]; for(i = 0; i != 4; ++i) { assert(A[i*3]==view1[i]); assert(A[i*3]==view2[i]); assert(A[i*3]==view3[i]); } assert(view1.size()==4); assert(view2.size()==4); // Assert fails!? assert(view3.size()==4); // Assert fails!?

1 0