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

shared_ptr interactions
by eric lindvall 08 Oct '04

08 Oct '04

I have created a simple template to make instantiation of shared_ptrs a little prettier. Here is an abridged version of the template: template <class T> class ManagedPointer : public boost::enable_shared_from_this<T> { public: typedef boost::shared_ptr<T> shared_ptr; public: static shared_ptr create () { return (shared_ptr (new T ())); } template <class A1> static shared_ptr create (A1 a1) { return (shared_ptr (new T (a1))); } }; This allows for me to create a class: class TestClass : public ManagedPointer<TestClass> { public: TestClass(); }; and create an instance by: TestClass::shared_ptr test = TestClass::create(); I have noticed in certain cases I get very strange behaviour. Most recently I've seen things hanging in the shared_ptr destructor. Changing the code to: TestClass::shared_ptr test (new TestClass()); does "solve" the problem. Is there something subtle in how the templates and subclassing are interacting that I am not understanding? thanks. e.

2 2

boost implementation
by Gast 128 08 Oct '04

08 Oct '04

Dear all, Browsing through source files of boost, I see that some libraries reimplement common idioms found in other libraries. For example the tuple library reimplements its own compile time if and it's own typelist. I guess it would be better if high level boost libraries(e.g. regex, threads, to some extend also the tuples) would use the low level boost libraries (e.g. preprocessor, type_traits, mpl). Besides bugfixes and platform/compiler workarounds, it has also a documentary advantage. If you are familiar with mpl for example, one knows how to interpret the construct with if_. Also one can better interact with libraries if they are using the same base constructs. There almost no disadvantages, because all the low level boost libraries are template based. But probably this is a matter of time, because at first look at the variant library, I see that it uses the mpl. And people are talking about the fusion library for the tuples. Are there boost developers who can comment on this? Wkr, me.

2 1

RE: [Boost-users] Problems with compiling Filesystem library on Solaris
by Puverle, Tomas (IT) 08 Oct '04

08 Oct '04

> Hi, > I get the following error from the Filesystem library when > building boost 1_31 on Solaris with sunpro compiler: > > "/db/pub/eq/build/src/boost_1_31_0/boost/iterator/iterator_fac > ade.hpp", line 622: Error: Could not find a match for > boost::iterator_core_access::equal<boost::iterator_core_access > ::Facade1, boost::iterator_core_access::Facade2>(const > boost::filesystem::directory_iterator, const > boost::filesystem::directory_iterator, > boost::is_convertible<boost::filesystem::directory_iterator, > boost::filesystem::directory_iterator>). > > "/db/pub/eq/build/src/boost_1_31_0/libs/filesystem/build/../sr c/operations_posix_windows.cpp", line 176: Where: While instantiating > "boost::operator==<boost::filesystem::directory_iterator, > boost::filesystem::path, boost::single_pass_traversal_tag, > boost::filesystem::path&, int, > boost::filesystem::directory_iterator, > boost::filesystem::path, boost::single_pass_traversal_tag, > boost::filesystem::path&, int>(const > boost::iterator_facade<boost::filesystem::directory_iterator, > boost::filesystem::path, boost::single_pass_traversal_tag, > boost::filesystem::path&, int>&, const > boost::iterator_facade<boost::filesystem::directory_iterator, > boost::filesystem::path, boost::single_pass_traversal_tag, > boost::filesystem::path&, int>&)". > > "/db/pub/eq/build/src/boost_1_31_0/libs/filesystem/build/../sr c/operations_posix_windows.cpp", line 176: Where: Instantiated from non-template code. > > > > It looks suspiciously similar to bug "[ 976241 ] Compiler > error for tokenizer on Solaris" from 2004-06-20, with the > difference that I am using version 5.3 of sunpro rather than 5.5. I have submitted a similar problem and the proposed fix recently. I am just waiting on our developers to confirm that it works before submitting it to boost, but it seems to fix the tokenizer problem, too. For a quick workaround (before the patch is applied) do this: At the top of boost/type_traits/detail/bool_trait_def.hpp you will see something like this: #if defined(__SUNPRO_CC) # define BOOST_TT_AUX_BOOL_TRAIT_VALUE_DECL(C) \ typedef mpl::bool_< C > type; \ enum { value = type::value }; \ /**/ # define BOOST_TT_AUX_BOOL_C_BASE(C) //1 Comment out the line marked with //1. That seems to have fixed my problems with both 5.3 and 5.5. HTH, Tom -------------------------------------------------------- NOTICE: If received in error, please destroy and notify sender. Sender does not waive confidentiality or privilege, and use is prohibited.

1 0

Problems with compiling Filesystem library on Solaris
by Barak Simon 08 Oct '04

08 Oct '04

Hi, I get the following error from the Filesystem library when building boost 1_31 on Solaris with sunpro compiler: "/db/pub/eq/build/src/boost_1_31_0/boost/iterator/iterator_facade.hpp", line 622: Error: Could not find a match for boost::iterator_core_access::equal<boost::iterator_core_access::Facade1, boost::iterator_core_access::Facade2>(const boost::filesystem::directory_iterator, const boost::filesystem::directory_iterator, boost::is_convertible<boost::filesystem::directory_iterator, boost::filesystem::directory_iterator>). "/db/pub/eq/build/src/boost_1_31_0/libs/filesystem/build/../src/operations_posix_windows.cpp", line 176: Where: While instantiating "boost::operator==<boost::filesystem::directory_iterator, boost::filesystem::path, boost::single_pass_traversal_tag, boost::filesystem::path&, int, boost::filesystem::directory_iterator, boost::filesystem::path, boost::single_pass_traversal_tag, boost::filesystem::path&, int>(const boost::iterator_facade<boost::filesystem::directory_iterator, boost::filesystem::path, boost::single_pass_traversal_tag, boost::filesystem::path&, int>&, const boost::iterator_facade<boost::filesystem::directory_iterator, boost::filesystem::path, boost::single_pass_traversal_tag, boost::filesystem::path&, int>&)". "/db/pub/eq/build/src/boost_1_31_0/libs/filesystem/build/../src/operations_posix_windows.cpp", line 176: Where: Instantiated from non-template code. It looks suspiciously similar to bug "[ 976241 ] Compiler error for tokenizer on Solaris" from 2004-06-20, with the difference that I am using version 5.3 of sunpro rather than 5.5. My guess is there must be some flag/macro/workaround I am missing (as this specific library is supposed to be portable to many OSs and compilers) , but have no idea what that may be. Any suggestions would be a blessing. Thanks, Barak ________________________ Barak Simon GED IT Core Platform Deutsche Bank 190 George St. Sydney, 2000 Phone +61 2 925 85070 Fax +61 2 925 95050 -- This e-mail may contain confidential and/or privileged information. If you are not the intended recipient (or have received this e-mail in error) please notify the sender immediately and destroy this e-mail. Any unauthorized copying, disclosure or distribution of the material in this e-mail is strictly forbidden.

1 0

Typelists
by Puverle, Tomas (IT) 07 Oct '04

07 Oct '04

What is the status of typelist support in boost? I read through a backlog of email in the archives but the last ones are from ~2002. Grepping through the source of 1.31.0 for (typelist | type_list) reveals some in the detail namespace for python, but that's all. Thanks, Tom -------------------------------------------------------- NOTICE: If received in error, please destroy and notify sender. Sender does not waive confidentiality or privilege, and use is prohibited.

2 1

uBLAS lapack bindings
by Daniel L Elliott 07 Oct '04

07 Oct '04

Hello, I have a question about using uBLAS/LAPACK bindings on AIX 5.1 and VA6. I am doing something like the following: matrix<double> matrixTwo(200,200); identity_matrix<double> identMatrix = identity_matrix<double>(200); int inverseInfo = gesv(matrixTwo,identMatrix); I am getting the following errors: line 102.36: 1540-0130 (S) "matrix_traits<boost::numeric::ublas::matrix<double,boost::numeric::ublas::row_major,boost::numeric::ublas::unbounded_array<double,std::a\ llocator<double> > > >::size1" is not declared "/npd/delliott/boost/numeric/bindings/traits/matrix_traits.hpp", line 102.7: 1540-0700 (I) The previous message was produced while processing "boost::numeric::bindings::traits::matrix_size1<boost::numeric::ublas::matrix<double,boost::numeric::ublas::row_major,boost::numeric::ublas::unbounded_array<double,std::allocator<double> > > >(matrix<double,boost::numeric::ublas::row_major,boost::numeric::ublas::unbounded_array<double,std::allocator<double> > > &)". "/npd/delliott/boost/numeric/bindings/lapack/gesv.hpp", line 135.40: 1540-0700 (I) The previous message was produced while processing "boost::numeric::bindings::lapack::gesv<boost::numeric::ublas::matrix<double,boost::numeric::ublas::row_major,boost::numeric::ublas::unbounded_array<double,std::allocator<double> > >,boost::numeric::ublas::identity_matrix<double> >(matrix<double,boost::numeric::ublas::row_major,boost::numeric::ublas::unbounded_array<double, std::allocator<double> > > &, identity_matrix<double> &)". "test.cpp", line 120.21: 1540-0700 (I) The previous message was produced while processing "main()". "/npd/delliott/boost/numeric/bindings/lapack/gesv.hpp", line 99.9: 1540-0038 (S) The name "typename traits::matrix_traits<boost::numeric::ublas::matrix<double,boost::numeric::ublas::row_major,boost::numeric::ublas::unbounded_array<double,std::allocator<double> > >>::matrix_structure" is not a type. "/npd/delliott/boost/numeric/bindings/lapack/gesv.hpp", line 95.9: 1540-0700 (I) The previous message was produced while processing "boost::numeric::bindings::lapack::gesv<boost::numeric::ublas::matrix<double,boost::numeric::ublas::row_major,boost::numeric::ublas::unbounded_array<double,std::allocator<double> > >,boost::numeric::ublas::identity_matrix<double>,boost::numeric::bindings::traits::detail::array<int> >(matrix<double,boost::numeric::ublas::row_major,boost::numeric::ublas::unbounded_array<double,std::allocator<double> > > &, array<int> &, identity_matrix<do...". "/npd/delliott/boost/numeric/bindings/lapack/gesv.hpp", line 137.16: 1540-0700 (I) The previous message was produced while processing "boost::numeric::bindings::lapack::gesv<boost::numeric::ublas::matrix<double,boost::numeric::ublas::row_major,boost::numeric::ublas::unbounded_array<double,std::allocator<double> > >,boost::numeric::ublas::identity_matrix<double> >(matrix<double,boost::numeric::ublas::row_major,boost::numeric::ublas::unbounded_array<double\ ,std::allocator<double> > > &, identity_matrix<double> &)". As I understand it, I can simply pass the uBLAS matrix types and the traits/lapack .hpp files will take care of the rest. Any light that could be shed upon this subject would be of great assistance. By the way, I would have happily send this to uBLAS-dev, but have not been admitted yet :( Thank you, Dan

2 1

FW: Compilation problem with boost on SunCC (Forte 6.2 & 8)
by Puverle, Tomas (IT) 06 Oct '04

06 Oct '04

A repost, since no answer to my previous mail. > -----Original Message----- > From: Puverle, Tomas (IT) > Sent: 01 October 2004 15:05 > To: 'boost(a)lists.boost.org' > Subject: Compilation problem with boost on SunCC (Forte 6.2 & 8) > > Please find attached a small program and a the resulting > compilation errors on SunCC. > This occurs in iterator_facade.hpp, operator != (), calling > to interator_core_access::equal. I have tracked this down to > the following problem: > > f(mpl::true_); > f(mpl::false_); > > f(is_convertible<>()); //will not work > > is_convertible<> derives from mpl::bool<>, however it seems > that a bug in the compiler prevents it from converting to > mpl::true_/mpl::false_ in the above scenario (and in the > program attached) > > A workaround for this is to change it to the following: > > f(mpl::true_); > f(mpl::false_); > > f(mpl::bool_<is_convertible<>::value>()); > > Presumably this will be present in other places but this is > the only one our developers have hit so far. > > Thanks, > > Tom -------------------------------------------------------- NOTICE: If received in error, please destroy and notify sender. Sender does not waive confidentiality or privilege, and use is prohibited.

3 3

Re: Problem compiling Windows C++ code wit regex
by arno.schaefer＠sqs.de 06 Oct '04

06 Oct '04

> Yes, that occurs if you have a > > using namespace std; > > statement in your code, and you are using Microsoft Visual C++ 6 (it's > actually a compiler bug). > > Removing (or possibly moving until after all your includes, or just > re-ordering your includes) the using statement will fix the issue. > Thanks very much for that hint John, I found two places in our header which are using the "using namespace std;". After I have switched that to the explicit "using" statements I could also switch back to your default "user.hpp". Perhaps I can ask one more question about this strange behaviour, just because I want to understand why: I try to use my own string class, which is inherited from std::string. Defining "VC6_STUPID_BEHAVIOUR" in the example code below compile and run the code successfuly, using the expanded name leads to following compiler error: error C2614: 'MyCString' : illegal member initialization: 'string' is not a base or member You know WHY??? Also because of the same compiler error in VC6 or is there something else wrong. On Solaris and HP-UX it works fine in every combination. // vvvvvvvvvvvvvvv snipp vvvvvvvvvvvvvvvv #include <string> #include <iostream> #ifdef VC6_STUPID_BEHAVIOUR using std::string; # define std_string string #else # define std_string std::string #endif class MyCString : public std_string { public: MyCString(const char *s_); : std_string(s_) { std::cout << "\"" << *this << "\"" << std::endl; }; }; int main (int argc, char *argv[]) { MyCString s(argv[0]); return 0; } // ^^^^^^^^^^^^ snipp ^^^^^^^^^^^^

2 1

boost install with gentoo issue?
by dave＠superelite.net 06 Oct '04

06 Oct '04

Well, more likely my mistake, ive been up awhile.. I installed the boost libs onto my gentoo machine with emerge. I then added -lboost_regex to my linker flags..thats it. When I run my app, it hangs, I think when waiting on pulling data of off a UDP socket. Though I cant tell, I'm about to crash and its been one of those days. But I do know everything is back to normal when I take -lboost_regex out of the flags again. I know I'm awfully vague due to lazy/tiredness, but does this ring a bell with anyone? I've installed boost the traditional way (bjam) on other linux distros before without any issue, but ive never tried emerge before. I dont know if this has something to do with it, or i am somehow missing something really obvious.. Thanks for bearing with me, Dave

4 4

uBLAS LAPACK bindings
by Daniel L Elliott 06 Oct '04

06 Oct '04

Hello, Are the uBLAS LAPACK bindings for general matrices or for any type of matrix? Do they have to use a particular type of storage? This is how it appears in the code. Would anyone like to talk about how other types of matrices might be used? Thank you, Dan

2 1