October 2011 - Boost-users - lists.preview.boost.org

boost::exception crash when throw from library
by Alexey Badyaev 07 Oct '11

07 Oct '11

Hi all. My little test project (minimal testcase attached to this mail) crashed on ms windows7 when compiled by ms visual studio 2008 express edition. May be, my c++ code is wrong - any help appropriated. -- With best regards, Alexey Badyaev Russian Federation, Kemerovo

2 6

[property_tree] Writing comments in JSON format
by Ruediger Berlich 07 Oct '11

07 Oct '11

Hi there, by trying it out with a modified debug_settings example, I can see that property_tree accepts Javascript ("//") comments inside of the JSON files it parses. I want to find out whether it makes sense to use JSON as an alternative to boost.program_options for configuration files. The idea would be to use a JavaScript-based GUI to modify or create the configuration files in a comfortable way and to use the same files as input for my C++-based programs. As my configuration files are auto-generated by my objects and not everybody loves GUIs, I need to be able to add comments to a property_tree object, so that they are emitted during a "write_json" call. The easiest way to add comments to the config files would be to just add a "comment" data member. This would indeed be very useful for the JavaScript usecase, as comments can be integrated into the GUI / Browser, and would still give guidance to those wanting to modify the JSON files with a standard editor. However, this does not help me to write things like a header, or to add information targeted only at those who modify the files by hand. For these I still need to be able to generate "//"-style comments. From the documentation I could not find out whether this can be done in a standard and portable way. Any pointer would be very welcome! Thanks and Best Regard, Ruediger

2 1

[filesystem] path iterator appends / instead of \ on windows
by Sachin Garg 07 Oct '11

07 Oct '11

In boost v1.47, filesystem\v3\src\path.cpp, inside function m_path_iterator_increment, on line 654, It should be: it.m_element.m_pathname = preferred_separator; Instead of it.m_element.m_pathname = separator; Reason: In below sample code boost::filesystem::path new_path; for (fs::path::iterator i=path.begin();i!=path.end();i++) new_path/=(*i); On windows when iterating a path using above code, line 654 incorrectly appends / instead of \ after drive letter. For example, D:\path\file Becomes, D:/path\file I have created a trac ticket for the same: https://svn.boost.org/trac/boost/ticket/5989 SG

2 2

[boost-users] [serialization] working example of (intrusive) serialization of class with no default constructor? and handling of nested classes with no default constructor
by alfC 07 Oct '11

07 Oct '11

Hi, this question has been asked thousand of times and I still can not see a detailed answer. How can I deserialize a class with no default constructor. The usual response is to point to this documentation http://www.boost.org/doc/libs/1_35_0/libs/serialization/doc/serialization.h… . I read the documentation many time and I can not figure out what does it mean and how it can be used. Does any body has a working small example where this is effectively used? Ideally, I am looking for something like the following, I am willing to accept other options, like making a2 a pointer, etc. int main(){ { A a( ... ); // proper construction syntax std::ofstream ofs("A.xml"); boost::archive::xml_oarchive oa(ofs); a & BOOST_SERIALIZATION_NVP(a); } { std::ifstream ifs("A.xml"); boost::archive::xml_iarchive ia(ifs); A a2(ia); // construct from iarchive or some other intermediate class { a2 & BOOST_SERIALIZATION_NVP(a2); //data can still be loaded in existing object (as usual) } } return 0; } Even if I want to implement my own reading of the xml_iarchive inside the constructor of A, it looks like I need to access the nested elements of the class before accessing to the level of the class (I am thinking in xml terms here). This is where I am stuck, the syntax above forces me to define A::A(boost::archive::xml_iarchive& ia){ ... } If all members are default constructible I can do: A::A(boost::archive::xml_iarchive& ia){ ia & BOOST_SERIALIZATION_NVP(*this); } (yes, the solution in intrusive but the at least I avoid default constructor at all cost) but if one of the member is not default constructible (and I can modify that class) I should be able to do something like, A::A(boost::archive::xml_iarchive& ia) : member1( ...ia?... ){ // ia & BOOST_SERIALIZATION_NVP(member2); // not possible because the archive is still waiting for class of type A to be read } which has two problems, 1) I need to access the elements of the archive (e.g. member2 before A is really handled by boost.serialization) 2.a) for member1 it regresses the problem to the member1 type, which is fine because all this objects will at some point be composed only from build-in "default constructible" types 2.b) if I have no way to modify member1 and it is not default constructible then I can't continue but that is also fine because that means that member1 was not serializable anyway. The real problem is 1). because if I have access to all the classes I will reach a point in which I have to read a built-in class from the iarchive, but I have have access to those nested elements directly, at least not with the documented interface. Thanks, Alfredo

2 1

Creating list of Graphs
by Yajima 06 Oct '11

06 Oct '11

Hi List I would like a help on figuring out how to manage bunch of Graph object as a vector. (vector could be ublas vector or stl doesn't matter) What I'm looking for is an operation that looks like the following: // (It doesn't work as it is.) typedef boost::adjacency_matrix<boost::directedS> Graph; Graph g0, g1, g2; ublas::vector<Graph> ListOfGraphs; ListOfGraphs.resize(3); ListOfGraphs(0) = g0; ListOfGraphs(1) = g1; ListOfGraphs(2) = g2; I'd appreciate it if you could give me an advice. Thank you. -- Yaji

1 0

[Test] Test cases in static libraries
by Jérémy Coulon 06 Oct '11

06 Oct '11

Hi all, I'm currently converting cppunit tests to Boost.Test. Boost test auto-registration is awesome ! I'm now facing a problem and need help from Boost community. mytest -> libFoo.so -> libBar.a Let's say I have a dynamic library Foo which depends on a static library Bar. I created an executable mytest which uses Foo library. It doesn't know about Bar library. So far, I successfully created test cases in mytest for Foo library using the public library interface (exported symbols). I just figured out how to create test cases inside Foo library (for internal non-exported classes) using console_test_runner utility. Question: how do I test code inside Bar static library ? Regards, Jeremy

3 3

Re: [Boost-users] boost::asio::io_service - cancel posted work
by Marat Abrarov 06 Oct '11

06 Oct '11

Hi, Christopher. > Since the callback never happens, my instance never gets destroyed. > > How do I effectively cancel any outstanding work the io_service has posted to > it, such that the smart pointer in the function object that the bind created > will lose its reference count? You can destroy io_service instance. asio::io_service::~io_service destroys all callbacks. Another solution is to close acceptor (and all other instances of asio::basic_io_object related to io_service instance), call asio::io_service::reset() and clear all queued callbacks by running them (asio::io_service::run/poll etc). Regards, Marat Abrarov.

2 1

boost::bind - What if bound (method, instance) doesn't exist anymore?
by Christopher Pisz 06 Oct '11

06 Oct '11

I've got some problem with the shared_ptr usage in my library. I am suspecting its in combination with boost::bind. What happens when a boost::function is called and has been created with a (method, instance) combination, but the instance no longer exists? What happens when this is in combination with boost::asio::io_service? Here is my suspect code: B::SharedPtr me = boost::dynamic_pointer_cast<B, A>(shared_from_this()); callback = boost::bind(&B::Foo, me, boost::asio::placeholder::error); // this is a boost::asio::ip::tcp::acceptor acceptor.async_accept(socket_, callback); Now suppose I use a raw pointer instead. Now suppose the instance that just got bound was destroyed before the accept happened. What will be the result? I'd try to just type up a test myself, but this requires alot of foundation code.

2 1

shared pointer problem - can't destruct
by Christopher Pisz 06 Oct '11

06 Oct '11

I've gotten myself into a situation. I've got a class A that created class Bs class B needs to register callbacks to the class A that created it. So I made member variables in class B that are boost::function type class A binds the method to be called and hands that to class B on construction So, not class A always has a reference count and cannot be destoryed. I tried passing a weak_ptr to boost::bind instead and it won't compile. what to do?

2 1

boost::thread provides semaphore?
by k m 05 Oct '11

05 Oct '11

Hi All I think boost::thread has semaphore back in the version 37. It seems it disappear from the current version. Is that right? If so, why? Thanks! Gary

2 1