- Boost-users - lists.boost.org

[BGL] Problem with color maps for undirected_dfs_visit
by Tarjei Knapstad 14 Nov '03

14 Nov '03

I'm trying to implement a DFS visitor and using it with undirected_dfs_visit, but I'm getting a segfault which relates to the edge color map. I've made the following test app for illustration: **************** START CODE **************** #include <vector> #include <boost/graph/adjacency_list.hpp> #include <boost/graph/undirected_dfs.hpp> using namespace boost; struct Types { typedef adjacency_list<vecS, vecS, undirectedS, property<vertex_color_t, default_color_type>, property<edge_color_t, default_color_type, property<edge_index_t, int> > > GraphType; typedef std::vector<default_color_type> ColorVec; typedef color_traits<default_color_type> Color; }; int main() { Types::GraphType g(6); add_edge(0,1,g); add_edge(1,2,g); add_edge(2,3,g); add_edge(2,4,g); add_edge(4,5,g); Types::ColorVec v_colors(num_vertices(g), Types::Color::white()); Types::ColorVec e_colors(num_edges(g), Types::Color::white()); default_dfs_visitor visitor; undirected_depth_first_visit(g, 0, visitor, make_iterator_property_map(&v_colors[0], get(vertex_index, g)), make_iterator_property_map(&e_colors[0], get(edge_index, g))); }; **************** END CODE **************** The code crashes in undirected_dfs.hpp line 73 the first time it's trying to access a color from the edge color map: EColorValue uv_color = get(edge_color, *ei); Valgrind turns up a lot of illegal reads/writes to the edge_color_map as well, so I'm guessing that I've screwed up on the colormap creation. Could anyone point out my error? Also, is this an error that could be caught compile time by adding some sort of concept check? Thanks in advance for any help! Regards, -- Tarjei

2 2

Boost.Thread: inlined constructor in thread_specific_ptr block application start up
by Roland 14 Nov '03

14 Nov '03

Dear Boosters, I am trying to understand how to use the TSS Thread local storage, and came upon a strange behaviour. I do not know if I am doing something wrong here. Maybe someone can help. I am running on Win2K, VC6.0. (GUI Interface programm type) In my main thread I have: struct oops { // inline oops(); // commented out, won't run oops(); // this one starts up ~oops(); private: int i; }; oops::oops() { i = 0; } oops::~oops() { } boost::thread_specific_ptr<oops> value; I am not assigning to value anywhere. Now when I try to start my programm (I tried only the debug version) my GUI never comes up. When I break into with the debugger I can see, that the programm is sitting somewhere in the main message loop (PumpMessage), but never receiving any messages. When not having the inlined version, behaviour is ok. Any help appreciated. Thank you, Roland

1 0

RE: [Boost-users] boost::shared_ptr inside std::list
by Darryl Green 13 Nov '03

13 Nov '03

> -----Original Message----- > From: James Mastro [mailto:jmastro@rochester.rr.com] > Sent: Thursday, 13 November 2003 11:59 AM > To: Boost Users mailing list > Subject: [Boost-users] boost::shared_ptr inside std::list > > > This one has been driving me nuts all day. In a function I have the > following: > > std::list< boost::shared_ptr<MyObject> > * objectList; An uninitialized pointer to a list of shared pointers to MyObject is probably not what you wanted, but its what you defined. I don't think you bug has anything to do with shared pointers... > > > When debugging my code, the debugger (gdb) dies upon entering the > function with this in it. It says "The Debugger has exited > due to signal > 11 (SIGSEGV).The Debugger has exited due to signal 11 > (SIGSEGV)." My dev > tools felt the need to say it twice for some reason. I pulled this out > and put it in a new project all by itself with the same result. > > If I make it not be a pointer, like this: > > std::list< boost::shared_ptr<MyObject> > objectList; > > then there's no problem. Regards Darryl. ########################################################################## This e-mail is for the use of the intended recipient(s) only. If you have received this e-mail in error, please notify the sender immediately and then delete it. If you are not the intended recipient, you must not use, disclose or distribute this e-mail without the author's prior permission. We have taken precautions to minimise the risk of transmitting software viruses, but we advise you to carry out your own virus checks on any attachment to this message. We cannot accept liability for any loss or damage caused by software viruses. ##########################################################################

2 1

RE: [Boost-users] Re: travelling salesman for BGL ?
by Ben Hutchings 13 Nov '03

13 Nov '03

Dave <better_cs_now(a)yahoo.com> wrote: > You'll be a millionaire (literally!) if you figure out the exact > solution to TSP in an efficient manner! Only if you patent it as a "business method".

1 0

Building Regex++ lib on Pocket PC using Embedded VC++4.0
by Reuti＠sdc.canon.co.in 13 Nov '03

13 Nov '03

Has anyone tried building regex++ lib for PPC using EVC++? I can build it using VC++6.0 using the .mak files supplied with it.But I am getting errors and warnings. Error is regarding the use of typename keyword in lib source and warning is about C++ Exceptions unwinding capability. I replaced C++ try, catch by Win32 __try , __except but don't know what to do with template error? Can anybody help in this? Also can it be confirmed that its possible to port it on Win CE? Thanx in advance. Reuti

2 1

travelling salesman for BGL ?
by Philippe Guglielmetti 13 Nov '03

13 Nov '03

I'm looking for a free "travelling salesman problem" solver. Does anyone know if something exists for the Boost Graph Library ? I found nothing in the book... Is anybody interested in porting existing code, for example http://www.math.princeton.edu/tsp/concorde.html on BGL ? -- Philippe Guglielmetti - www.dynabits.com

3 2

Re: [Boost-users] graph: Min Spanning Tree (advanced start)
by Jeremy Siek 13 Nov '03

13 Nov '03

Hi Matthew, Yes, that is, as long as that initial list of edges forms a tree. Here's my guess as how to do it for prim's minimum spanning tree. You'll have to directly call breadth_first_visit, redoing all the initialization work done in dijkstra and prim (prim calls dijkstra, which then calls bfs). You'll need to initialize the predecessor map to reflect the starting tree, and also you'll want to set the color of the nodes in the starting tree to black. Then you'll have to color the nodes adjacent to nodes in the tree (that aren't in the tree) gray and stick them in the priority queue... that is, all except for one of the nodes adjacent to the tree (pick an arbitrary one). That node will be your starting node for the bfs. In you're using Kruskal's it is a different story... I'll look at that later if you need it. Cheers, Jeremy On Wednesday, November 12, 2003, at 11:24 AM, Matthew Galati wrote: > Is there a way to give an advanced start to solve Min Spanning Tree > using boost/graph? That is, I have a list of edges in my graph that > must be in the Min Spanning Tree (i.e., their cost is -M, for very big > M). > > Since algorithms for MST are greedy, I assume this can be done - but > is their an easy way to interface to boost/graph's MST algorithms? > > Thanks for any help. > Matt > > -- > Matthew Galati > ISE Lehigh University > IBM Service Parts Solutions > 610.758.4042 (Office) > 610.882.0779 (Home) > magh(a)lehigh.edu, magal11(a)us.ibm.com > http://sagan.ie.lehigh.edu/mgalati/ > > _______________________________________________ > Boost-users mailing list > Boost-users(a)lists.boost.org > http://lists.boost.org/mailman/listinfo.cgi/boost-users

1 0

boost::shared_ptr inside std::list
by James Mastro 13 Nov '03

13 Nov '03

This one has been driving me nuts all day. In a function I have the following: std::list< boost::shared_ptr<MyObject> > * objectList; When debugging my code, the debugger (gdb) dies upon entering the function with this in it. It says "The Debugger has exited due to signal 11 (SIGSEGV).The Debugger has exited due to signal 11 (SIGSEGV)." My dev tools felt the need to say it twice for some reason. I pulled this out and put it in a new project all by itself with the same result. If I make it not be a pointer, like this: std::list< boost::shared_ptr<MyObject> > objectList; then there's no problem. Oddly, both versions of the code work fine when run normally (not being debugged.) I don't see what's wrong with the first version. Is it OK to put boost shared ptrs in a std::list? Is there something else I'm missing? Or is my dev environment just having problems? -jim

1 0

Re:[Boost-users] graph: Min Spanning Tree (advanced start)
by rlyon＠comgroup.net.au 12 Nov '03

12 Nov '03

It's no longer necessarily a minimum spanning tree. The exact algorithm will depend upon whether your starting set of edges always is a tree. Regards ... ____________________Reply Separator____________________ Subject: [Boost-users] graph: Min Spanning Tree (advanced start) Author: Matthew Galati <magh(a)lehigh.edu> Date: 12/11/2003 11:24 AM Is there a way to give an advanced start to solve Min Spanning Tree using boost/graph? That is, I have a list of edges in my graph that must be in the Min Spanning Tree (i.e., their cost is -M, for very big M). Since algorithms for MST are greedy, I assume this can be done - but is their an easy way to interface to boost/graph's MST algorithms?

1 0

Re: [Boost-users] int64_t error
by Gilad Suberri 12 Nov '03

12 Nov '03

Thanks. It seems that I do not have the file "stdint.h", so in cstdint.hpp, int64_t is defined in the last #else section (that goes with #ifdef BOOST_HAS_STDINT_H). When operators.hpp is included before this file (cstdint.hpp,) int64_t does not get defined at all because this next statement from cstdint.hpp becomes false: # if defined(BOOST_HAS_LONG_LONG) && \ !defined(BOOST_MSVC) && !defined(__BORLANDC__) && \ (!defined(__GLIBCPP__) || defined(_GLIBCPP_USE_LONG_LONG)) && \ (defined(ULLONG_MAX) || defined(ULONG_LONG_MAX) || defined(ULONGLONG_MAX)) When the include statements are switched around int64_t gets defined. This is connected to posix_time_config.h where "boost/date_time/time_duration.hpp" is included before "boost/date_time/time_resolution_traits.hpp", which causes the above problem. I switched the two include statements and I am not running into that compiler error anymore, but I am not sure if that will cause other compiler errrors. Hope some of this made sense, let me know what you think or if I am missing anything. Thanks. Gilad ----- Original Message ----- From: "Jeff Garland" <jeff(a)crystalclearsoftware.com> To: "Boost Users mailing list" <boost-users(a)lists.boost.org> Sent: Wednesday, November 12, 2003 10:10 AM Subject: Re: [Boost-users] int64_t error > On Wed, 12 Nov 2003 08:42:08 -0500, Gilad Suberri wrote > > Hi, I am compiling my projects on Solaris, using the latest version > > of the boost date_time library. All the gregorian examples compile > > and build fine, but when I try to compile posix_time ones I get this > > error: > > > > ...error details snipped... > > The version of gcc is 3.0.2. > > I do not get this compiler error when I use gcc 2.95.3, but I needed > > to use the newer version. > > > > Any ideas? Thanks. > > I suspect that somehow you are not getting boost::int64_t defined. Although > looking at the logic in boost/config I'm not seeing why at the moment -- you > might experiment with trying to make sure boost::int64_t is defined. I won't > have a chance to look until later as I'm running out the door. Bottom line is > it should work with this compiler... > > Jeff > > > > _______________________________________________ > Boost-users mailing list > Boost-users(a)lists.boost.org > http://lists.boost.org/mailman/listinfo.cgi/boost-users

2 2