November 2005 - Boost-users

Documentation Typo
by Merrill Cornish 20 Nov '05

20 Nov '05

FYI: The documentation page for the ptr_list<> class, http://www.boost.org/libs/ptr_container/doc/ptr_list.html, has a heading the says "Class ptr_vector". The headings are correct for ptr_vector<>, ptr_array<>, and ptr_deque. Merrill

2 1

Thread sheduling priority
by Leon Mergen 19 Nov '05

19 Nov '05

Hello, If been lookup up possible planned functonality for adjusting a thread's priority in the boost thread library. In my search, i found this email: http://aspn.activestate.com/ASPN/Mail/Message/boost/1690190 In that message, sent 2 years ago, William E. Kempf states that they are implemented but are still undergoing design changes. At the moment, I still don't see that functionality in the main boost library - is there any idea when this is going to be part of the main boost library ? Thanks in advance. -- Leon Mergen

1 0

[Iostreams] file_descriptor::seek() bug
by Takeshi Mouri 19 Nov '05

19 Nov '05

Hello, I try file_descriptor::seek() for a big file over 4GB on Win32, but the result was not correct. file_descriptor.cpp(185) > LONG lDistanceToMoveHigh = > off < 0xffffffff ? > static_cast<LONG>(off >> 32) : > 0; I think that lDistanceToMoveHigh always becomes 0. The condition above should be "off > 0xffffffff". Regards, Takeshi Mouri

2 1

build times
by gast128 19 Nov '05

19 Nov '05

Dear all, we use boost a lot in our code, but unfortunately this slows down the build times enormously. Although we use the precompiled header in Visual Studio, tracing the includes (with the /showincludes option), it seems that even relatively basic libraries (e.g. tuples, boost\utility.hpp, boost function and boost bind) end up in a massive include of mpl and type traits header files. So I begin to wonder if this wonderful template and preprocessor stuff is really such a good idea for those basic library. Besides that it doesn't improve the readability of the source, my guess is that it has a negative effect on short build times. Can a boost guru comment on this? Note that this is not a criticism, only an observation. I always like to write a boost bind construction, especailly with the == addition in 1.33. Wkr, me

14 91

[iostreams] Assertion in detail/optional.hpp, line 54
by Daniel Lidström 18 Nov '05

18 Nov '05

Hello! when compiled in debug mode, the following program causes an exception in the file p:\boost_1_33_0\include\boost-1_33\boost\iostreams\detail\ optional.hpp, line 54 (I'm using VC 7.1.3088, Windows XP): #include <iostream> #include <boost/iostreams/concepts.hpp> // sink #include <boost/iostreams/stream.hpp> // stream using namespace std; class IOutput { public: virtual void print(int, const char*) const = 0; }; class Output : public IOutput { public: void print(int id, const char* a) const { cout << a; } }; static IOutput* gpOutput; template<int id> class my_sink : public boost::iostreams::sink { public: std::streamsize write(const char* s, std::streamsize n) { gpOutput->print(id, s); std::streamsize sz = 0; for( const char* p=s; *p; ++p ) sz++; return sz; } }; typedef boost::iostreams::stream<my_sink<5> > output; int main() { gpOutput = new Output; output stream; stream << "Daniel" << endl; return 0; } When I compile in release mode the program prints `Daniel', just as expected. Is there a problem with my program? Thanks in advance! Hälsningar, Daniel

2 1

typo in Boost.Function docs
by Weston Markham 18 Nov '05

18 Nov '05

The following is from the documentation published at www.boost.org. Surely that "!" in the "throws" description should not be there, right? Weston functionN invocation result_type operator()(arg1_type a1, arg2_type a2, ... , argN_type aN) const; Effects: f(a1, a2, ..., aN), where f is the target of *this. Returns: if R is void, nothing is returned; otherwise, the return value of the call to f is returned. Throws: bad_function_call if !this->empty(). Otherwise, may through any exception thrown by the target function f.

2 1

calling on the template gods
by Doug Henry 18 Nov '05

18 Nov '05

I have been attempting to hide the boost interface when using the boost smart pointers. I basically want to get rid of boost declarations in my API's. This basically involves typedef-ing template classes, and the best solution I have been able to come up with is: template <typename T> struct pointer { typedef boost::scoped_ptr<T> obj; typedef boost::scoped_array<T> arr; typedef boost::shared_ptr<T> sobj; typedef boost::shared_array<T> sarr; }; so I can do things like pointer<int>::arr var(new int[10]). This works fine, but if I try to get tricky it falls apart. Another example: template <typename T> class Stuff { private: pointer<T>::arr some_stuff; }; This does not work (I can't even start to tell you why), but I have hit the point where I have to start second guessing my approach to this problem. Are there other (possibly terrible) tricks to get around the typedef templates problem? Any input will probably help. -thanks

5 4

auto_ptr w/ user specified deleter
by christopher baus 18 Nov '05

18 Nov '05

I'm messing around with the pool library and I was wondering if anybody has done any work on a version of auto_ptr with which allows a user specified deleter like scoped_ptr. Thanks, Christopher

1 0

[BGL,Graph] bundle and property maps
by Stephane Grabli 18 Nov '05

18 Nov '05

Hi, I'm (still) trying to use the strong components algorithm with a graph storing vertices as a list (listS). I've learned thanks to this mailing list that I had to explicitely declare and intialize a vertex_index map. The declaration of my graph type should therefore look like that: typedef adjacency_list<listS, listS, directedS,property<vertex_index_t,std::size_t> > Graph; and the component map should be declared this way: typedef graph_traits<Graph>::vertices_size_type v_size_t; //the range of ComponentMap vector_property_map< v_size_t, property_map<Graph, vertex_index_t>::type> componentMap(num_vertices(g), get(vertex_index,g) ); //g is an instance of Graph I would like to know if there is a way to combine this with the use of bundles. In practice, I used to have this declaration for my graph type: typedef adjacency_list<listS, listS, directedS,VertexBundle> Graph; with: struct VertexBundle{ int id; }; How should I merge these two types of declarations? Is there a way to define the required vertex_index as a bundle property and to use it to define my component map afterward? Thanks! Stephane PS: I apologize for the level of the questions, but I couldn't get a good understanding of all the property mechanisms out of the documentation.

3 2

[IOStreams] linking the iostreams libs with my own zlib.lib
by Christian Henning 17 Nov '05

17 Nov '05

Hi there, my project is using its own created zlib.lib, thats why I want to create the iostreams libs using our own zlib.lib . Creating the iostreams libs I use the following build command: bjam -sTOOLS=vc-7_1 --libdir=c:\boost\lib --with-iostreams "-sZLIB_BINARY=C:\CHH\External\Libraries\Release\zlib" "-sZLIB_INCLUDE=C:\CHH\External\Includes\zlib" "-sZLIB_LIBPATH=C:\CHH\External\Libraries\Release" --without-python stage Note that the ZLIB_SOURCE variable is missing. Otherwise the build would create it's own zlib library. Here is the warning I get when bjam is building the debug lib: LINK : warning LNK4098: defaultlib 'MSVCRT' conflicts with use of other libs; use /NODEFAULTLIB:library The problem arises because it's using a release built zlib.lib with the debug built iostream lib, since I can only pass one zlib to the build command. A solution would be to have two variables ZLIB_LIBPATH_DEBUG and ZLIB_LIBPATH_RELEASE. Which is the standard procedure. Or is there a variable in the boost build for creating debug lib only? Greets, Christian

2 4