March 2004 - Boost-users - lists.preview.boost.org

RE: [Boost-users] Re: Using shared_ptr across shared library boun dary
by Jim.Hyslop 12 Mar '04

12 Mar '04

David Abrahams wrote: > "Jim.Hyslop" <Jim.Hyslop(a)Leitch.com> writes: > > It's been my experience > > (with Visual Studio, at least) that allocating memory in one DLL and > > deallocating it in another leads to problems. Sometimes it doesn't > > go noticed, but when it does, it's really noticeable. > > The library cleverly avoids that, provided the first shared_ptr to > take ownership of the memory does so in the same DLL where the object > was allocated. Huh - whaddya know; I just tried it with this program: int main() { boost::shared_ptr< T > t = CreateT(); t.reset(); T * tptr = CreateRawT(); delete tptr; } where CreateT and CreateRawT are exported from a DLL (and do the obvious thing). 't.reset()' works OK, but 'delete tptr' asserts on an invalid heap pointer. I'll have to study the shared_ptr implementation to see how it avoids this problem. -- Jim Hyslop Senior Software Designer Leitch Technology International Inc. (http://www.leitch.com) Columnist, C/C++ Users Journal (http://www.cuj.com/experts)

1 0

RE: [Boost-users] Using shared_ptr across shared library boundary
by Jim.Hyslop 11 Mar '04

11 Mar '04

Edward Diener wrote: > Is there any problem with using shared_ptr and passing an > instance of it to > a function in another shared library, such as a DLL under > Windows ? Does the > reference count get updated correctly when the copy is passed to a DLL > function, so that when the last shared_ptr in either the > original module or > the DLL goers out of scope, the actual pointer is correctly deleted ? I've been using it successfully across DLL boundaries with no problems. The biggest caveat is to ensure you are using the DLL version of the C Runtime library. If you use the static library version, then each DLL will have its own heap, and a pointer *must* be deallocated in the same DLL in which it was allocated. -- Jim Hyslop Senior Software Designer Leitch Technology International Inc. (http://www.leitch.com) Columnist, C/C++ Users Journal (http://www.cuj.com/experts)

2 1

RE: [Boost-users] Re: Using shared_ptr across shared library boun dary
by Jim.Hyslop 11 Mar '04

11 Mar '04

David Abrahams wrote: > "Edward Diener" <eddielee(a)tropicsoft.com> writes: > > > Is there any problem with using shared_ptr and passing an > instance of it to > > a function in another shared library, such as a DLL under Windows ? > > Works beautifully, even if the two DLLs are using different heaps. Are you sure about the different heaps? It's been my experience (with Visual Studio, at least) that allocating memory in one DLL and deallocating it in another leads to problems. Sometimes it doesn't go noticed, but when it does, it's really noticeable. -- Jim Hyslop Senior Software Designer Leitch Technology International Inc. (http://www.leitch.com) Columnist, C/C++ Users Journal (http://www.cuj.com/experts)

2 1

RE: [Boost-users] Re: Using shared_ptr across shared library boun dary
by Mark Storer 11 Mar '04

11 Mar '04

> Works beautifully, even if the two DLLs are using different heaps. How does that work? Do they both end up with the same allocator... so all the allocations end up on the same heap? --Mark Storer Software Engineer Cardiff Software #include <disclaimer> typdef std::disclaimer<Cardiff> Discard;

1 0

Using shared_ptr across shared library boundary
by Edward Diener 11 Mar '04

11 Mar '04

Is there any problem with using shared_ptr and passing an instance of it to a function in another shared library, such as a DLL under Windows ? Does the reference count get updated correctly when the copy is passed to a DLL function, so that when the last shared_ptr in either the original module or the DLL goers out of scope, the actual pointer is correctly deleted ? If this is not the case, that shared_ptr works correctly across DLL boundaries in the manner described, does instantiating and exporting shared_ptr template instances from the DLL to the module using the DLL solve the problem ?

2 2

boost::thread no parameter?
by GoochRules! 11 Mar '04

11 Mar '04

Greetings, I'm wondering why there is no option to pass a parameter to a thread function with boost's threads. With pthreads and Windows CreateThread I'm used to casting an object to a void*, and retrieving the object within the thread function. The only thing similar with boost::thread seems to be to create a struct with an operator()() and a constructor that takes a void*. operator()() then calls an external function with the passed void*. Just curious about this design decision. --M Peltzer

4 3

Re: [Boost-users] Compiler Generator Tools
by Roel.Sergeant＠pandora.be 11 Mar '04

11 Mar '04

Hi Hossein, I know that Spirit can be used for scanning and lexical analysis, at least that where I use it for. I don't know about the rest, but I guess Spirit will help you out there too. In my little project I use it for parsing XML, HTML and other datafiles. For more info on spirit : http://spirit.sourceforge.net/ If you need a C/C++ parser, you might consider Wave, which uses Spirit, and is highly standard compliant. Hope this helps, Roel. >----- Oorspronkelijk bericht ----- >Van : Hossein Haeri [mailto:powerprogman@yahoo.com] >Verzonden : maandag , maart 8, 2004 01:07 PM >Aan : boost-users(a)lists.boost.org >Onderwerp : [Boost-users] Compiler Generator Tools > >Dear all, > >Who knows where I can find the compiler generation >stuff in the Boost repository? I need scaner, lexical >analyser, compiler, code generator, ... Are there any >such tools in the Boost which can build all of the >above goodies for me? > >TIA, >--Hossein > >__________________________________ >Do you Yahoo!? >Yahoo! Search - Find what you?re looking for faster >http://search.yahoo.com >_______________________________________________ >Boost-users mailing list >Boost-users(a)lists.boost.org >http://lists.boost.org/mailman/listinfo.cgi/boost-users > > >

5 5

Compile problem
by fmonkey＠fmonkey.net 11 Mar '04

11 Mar '04

This may be a silly question, but why does the following code give me a compile error (only on Linux BTW, not Windows): template<typename TreeIterT> void ClassName::compile_something(const TreeIterT& iter, compile_context &ctx) { const TreeIterT::value_type& ident = get_first_leaf(*iter); } In member function `void ClassName::compile_something(const TreeIterT&, CompileContext&)': 128: error: syntax error before `&' token Also: template<typename TreeNodeT> void ClassName::compile_something_else(const TreeNodeT &tree, compile_context &ctx) { TreeNodeT::const_iterator iter = tree.begin(); } In member function `void ClassName::compile_something_else(const 138: error: syntax error before `=' token Any idea what is causing this?

2 1

Graph, inv_adjacency_iterator_generator problem
by Peter Aronsson 10 Mar '04

10 Mar '04

Hi, I have a problem with traversing over all parents of a vertex. If I use (the undocumented?) inv_adjacency_iterator_generator class to return iterators for the parents (predecessors) of a vertex I get a compile error. It worked on an earlier version of the library (in debian testing), but does not work for the latest version(1.31.0). I think the problem is that it can no longer find a suitable constructor. Here is my code: // some typedefs removed...... typedef boost::graph_traits<TaskGraph>::in_edge_iterator InEdgeIterator; typedef boost::adjacency_list<boost::listS, boost::listS, boost::bidirectionalS, VertexProperty, EdgeProperty> TaskGraph; typedef boost::inv_adjacency_iterator_generator<TaskGraph, VertexID, InEdgeIterator>::type ParentsIterator; std::pair<ParentsIterator, ParentsIterator> parents(VertexID v, TaskGraph &tg) { InEdgeIterator e,e_end; tie(e,e_end) = in_edges(v,tg); ParentsIterator c(e),c_end(e_end); // problem here. No suitable constructor. return make_pair(e,e_end); } (Traversing through the children/predecessors works fine.) Regards Peter Aronsson -- _________________________________________________________________ / Peter Aronsson, Phd Student at PELAB (Programming Environments \ | Laboratory ) Department for Computer & Information Science | | Linköping University, Sweden | |=================================================================| | petar(a)ida.liu.se , phone +46 (0)13-28 1737 Room 3B:490 | \_________________________________________________________________/

2 3

mpl lambda facilities
by Matt Calabrese 10 Mar '04

10 Mar '04

I'm trying to create a metafunction which takes in a boost::mpl::vector containing spirit parser types, and yields a vector of the original vector's length plus one with each element representing the parser type made of the recursive boost::spirit::alternative of all of the original parser types from the corresponding element in the original vector and an epsilon_parser at the end. in other words, the metafunction should be able to take in: (using boost::mpl and boost::spirit) vector< chlit<>, strlit<>, range<> > and yield a result: vector< alternative< chlit<>, alternative< strlit<>, alternative< range<>, epsilon_parser > > >, alternative< strlit<>, alternative< range<>, epsilon_parser > >, alternative< range<>, epsilon_parser >, epsilon_parser > where empty is my own parser type, similar to epsilon_parser I'm still new to metaprogramming and the mpl lambda facilities and I'm having some trouble creating a metafunction which produces my desired type. This is what I've come up with, though it doesn't work - I'm assuming because of my second use of mpl::_1 in the nested template parameter of mpl::front in mpl::push_front (which I'd like to have represent the current state of the mpl::fold). template< typename vector_t > struct get_alternatives { typename mpl::fold_backward< vector_t, mpl::vector< epsilon_parser >, typename mpl::push_front< mpl::_1, spirit::alternative< mpl::_2, typename mpl::front< mpl::_1 >::type > >::type >::type type; }; Any help is greatly appreciated. Thanks in advance.

2 1