June 2007 - Boost-users - lists.preview.boost.org

[Preprocessor] Expand to nothin - BOOST_PP_SEQ_PUSH_BACK
by Manuel Jung 24 Jun '07

24 Jun '07

Hi The BOOST_PP_SEQ_PUSH_BACK Macro expands to the seq with the extra element. But i don't want it to expand. I want to use the sequence later instead. How can i supress the expanding? I looked through the docs, but didn't find the right clue. I'm new too Preprocessorprogramming, so please excuse, if my question is stupid. Not the easiest thing, starting Preprocessor and Metaprogramming from nothing but the docs. Kind Regards Manuel Jung

2 2

Boost.graph : 3D graphics primitives hierarchy (scenegraph)
by Nicholas Yue 23 Jun '07

23 Jun '07

Hi, I am considering using boost::graph to represent a DAG of 3D graphics primitive hierarchy, much like a scenegraph in games with the exception that no drawing code is embedded with this structure. The DAG is important in that I will be injecting changes to certain nodes which than affects/updates all it's child nodes. The intention is to allow me to write tools to work with this hierarchy. Eventually, these geometries may be turned into renderable entities (via some visitor iterator) for visualization/verification but that is not the main purpose. I am very new to boost::graph and hope someone can point me at information or open source project that is relevant to my exploration. Regards

4 4

Dijkstra Visitor - Accessing the priority queue
by Jonas Gabriel 23 Jun '07

23 Jun '07

Good morning, I'm trying to make a single pair shortest path function using the goal-directed search technique based on dijkstra_shortest_paths and Visitors. In some point when the target vertex is processed, I would like to exit dijkstra_shortest_paths.To do that I 'm thinking trying to empty the priority queue. Can I access the priority queue inside a visitor using something like an extern declaration, since none of the Dijkstra Visitor's methods gets the priority queue as an argument ? With regards, Jonas Gabriel

2 1

How to "splice" function arguments?
by Zeljko Vrba 23 Jun '07

23 Jun '07

I have the following situation with parallel overloaded functions: void f(X *x) | void f(int g, X *x) { { | { Impl<false>::f(-1, x); | Impl<true>::f(g, x); } | } template<bool F> struct Impl { typedef Detail<F> detail; void f(int g, X *x) { // CODE X *t = detail::smth(g, x); // CODE } }; In the "false" case, the argument g of Impl::f is not used and I want to get rid of it in compile-time as an optimization. The rest of the CODE is completely identical. In other words, I want to transform the "false" case, at compile-time, to the following: void f(X *x) { Impl<false>::f(x); } template<> struct Impl<false> { typedef Detail<false> detail; void f(X *x) { // CODE X *t = detail::smth(x); // CODE } }; Now, I could manually duplicate all of the code in Impl::f(int, X*) into Impl::f(X*) and remove the extra argument in the call to detail::smth. Doing such specialization manually seems _very_ inelegant and error-prone (I'm not really a fan of copy-pasting the code :)) Is there a better way; ie. doing it automatically? If so, how? It is acceptable to manually overload the top-level f() function, as it has no other code except the dispatch to Impl::f. I want full compile-time solution; ie. I do not want the stack space for the unneeded integer parameter ever to be allocated. [The code for the "false" case will be called on every function invocation, so YES, performance DOES matter even without apriori profiling.] Signatures in the implementation of Impl and Detail are flexible and I can change them if it helps solving the problem. The only thing that I do not want to change are the signatures of top-level f() function overloads. Thanks, Zeljko.

2 2

[boost.test] Provoking a core-dump
by Zeljko Vrba 23 Jun '07

23 Jun '07

I'm using the boost minimal testing facility - mostly because I do not want to build full boost. [I spent one day about an hour trying to figure out how to build a 64-bit version on Solaris with SunCC and failed. So I gave up. Besides, minimal testing facility is enough for me.] Is it possible to configure BOOST_CHECK or BOOST_REQUIRE to behave as assert? BOOST_REQUIRE throws an exception which eventually gets caught and the test just prints "failed". That's not what I want. I want that the message is printed and abort() immediately called so that I can inspect the core-dump and debug the problem. A quick&dirty fix is to add something like the following on top of each test: #ifdef BTEST_OVERRIDE #define BOOST_REQUIRE assert #define BOOST_CHECK assert #endif Is there another possibility? Ideally, there would be a command-line switch to toggle this behaviour at runtime. So the first run of tests in "normal" mode just gives a quick overview of everything, and another "diagnostic" mode (running with a hypothetical "-abort" switch) calls abort() as soon as any CHECK or REQUIRE fails.

2 1

Boost test thread safety
by Chris 23 Jun '07

23 Jun '07

I have been working for a while on trying to diagnose whether my library or the boost test library is causing many of my thread issues. I hear that the boost test is not thread safe, is the correct? How incomplete is the thread safety and is there a time span on when it will be achieved? -Thanks for the current thread Chris

2 1

Boost.Test internal framework error: unknown reason
by Ole Christian Weidner 23 Jun '07

23 Jun '07

Aloha, I have a strange problem when running the Boost Unit Test Framework on MacOS X (Intel). I'm using Boost 1.33.1 built from source with gcc 4.0.1 (Apple Computer, Inc. build 5367). Compiling and linking of my tests works fine with: /usr/bin/g++-4.0 -arch i386 -Wl,-all_load -L$BOOST_LOCATION - lboost_unit_test_framework ... but whenever I try to run my test application I get a not very meaningful error message (the same code works fine on Linux and Windows): [Session started at 2007-06-11 13:06:27 -0500.] SAGA C++ Conformance Tests Boost.Test internal framework error: unknown reason saga_test_suite_v2 has exited with status 200. The error seems to occur somewhere within the framework initialization - the entry point (test_suite * init_unit_test_suite (int argc, char* argv[])) is never reached. Any ideas how to track down the error? Are there any command line arguments to make the framework's output more verbose? Thanks, Ole

2 1

[1.34.1] Release Candidate
by Thomas Witt 23 Jun '07

23 Jun '07

Hi, The 1.34.1 release candidate is now available from http://www.boost.org/release/ please give it a try and report any problems to the developer list or directly to me. I'll be travelling till July 5th. Internet access is unlikely, so far :( Thanks Thomas -- Thomas Witt witt(a)acm.org

2 1

problem... maybe boost could help
by Lukasz 22 Jun '07

22 Jun '07

Hello, I'm fairly new to boost and don't know all of the functionality. I have been using the serialization library, read about mpl, signals and that's about it. Maybe someone will have an idea if there is anything in boost that i could use to solve the following problem (or any other solution). I'm working a an Agent framework, and at this point, specifically on an event (messaging system). The basic layout is the following (what concerns the messaging part): The framework contains Agents and Behaviors. They all need to communicate with each other. Both of these classes derive from an interface EventReceiver, which declares, amongst other things, a pure virtual ProcessEvent(Event * event) function. Specific behaviors and agents are obviously derived from the base ones. All events are derived from the base Event class (to allow users to create their own events with whatever information they need in there). The main messaging hub is a class MessengerService, which contains two maps, the first one relates static UUID's of events to UUID's of receivers , and the second receiver UUID's with a list of actual events of type Event*. The basic use case is: a behavior calls a register function in MessengerService to add itself as a receiver for a certain type of event (register/unregister can happen at anytime). At anytime, a behavior/agent can create an event, and call a send function, which adds it to the list of events that are pending for a given receiver (if a receiver was registered for this event type). As all agents/behaviors are ticked, a "Monitor" is moved to the current agent/behavior. The monitor queries the messenger service for all pending events destined for the current agent/behavior, and calls a processEvent function on the agent/behavior passing it the event as an argument. The problem is here. I want agents/behaviors to add handlers for different event types that they registered for, i.e. class SomeBehavior: public Behavior { ... processEvent(EventTest1 & event); processEvent(EventTest2 & event); ... } Since the events were registered, i know that the correct handlers exist in the most derived class types (there will also be a forwarding scheme for unregistered or broadcast events). The problem is that i only have a pointer to the base class (EventReceiver) and to the base class of the event (Event). I can't cast directly because i don't know the type to cast to. Also, i don't want to add virtual functions for all possible events to the base class, because that would mean that users would have to play around in the base class (framework) every time they add a new event. Lastly, i wanted to avoid having a general processEvent(Event & event) function and then have the users "switch" on the event type. I solved part of the problem (the first half) by adding a "MostDerviedHelper", which looks sort of like this: template<class Derived> class MostDerivedHelper { .... void dispatchEvent(Event * event) { // Do dispatching here m_derivedPointer->processEvent(*(event->This())); } private: Derived * m_derivedPointer; } Each derived type has a helper, which obviously contains the correct most derived type (since it's itself). The problem now, is that i don't know what to cast the event to so that it calls the correct handler. I can't add the same type of helper to event, because then i would have to pass it the receiver derived type through a template, but since it would have to be a template, i can't call it using my base class event pointer (can't create templated virtual functions). I tried adding a virtual function called "This()" to all events, which would return the derived classes' this pointer, but eventhough it's the correct type, i still need to cast it (which i can't) since it has to be resolved at compile time. I read quite a few things about double/multiple dispatch (in Loki for example), but it's not quite the solution i'm looking for because a usual implementation uses a callback, and that would require that i make my handlers static. MPL is not much of a help either because it's based on static polymorphism, and what i actually need is dynamic. I thought of using boost function & boost bind combined with a double dispatch scheme, but this doesn't work either because then i can't serialize it (i could re-perform the binding after a restore, which i already do for boost::signals, but it's a hassle). So, i'm pretty much out of ideas. I think to sum it up, what i would sort of need is: template<class ReceiverType, class EventType> class Dispatcher { ... send(ReceiverType * receiver, eventType * event) { receiver->processEvent(event); } } In this situation, i would already have all the correct types. The problem is that i can't get all this info at the same time. I also thought about using the curiously recurring template pattern as such: template<class Derived> class Event { template<class derivedReceiver> dispatch(derivedReceiver, Derived event) { .... } } This would work, except that since all events now must specialize to some type of base, i can't use a base class pointer to store in my map, i.e. i can't store all of my events in a map. If anyone has some brilliant idea as to how this can be done (if it can), then i would appreciate the input. Thanks, Lukasz -- View this message in context: http://www.nabble.com/problem...-maybe-boost-could-help-tf3959613.html#a112… Sent from the Boost - Users mailing list archive at Nabble.com.

3 3

[wave] preprocessing error
by Ovanes Markarian 22 Jun '07

22 Jun '07

Hello Hartmut, I have a question to you. I got a header file with an error. The content of a file looks like: #ifndef SOME_GUARD #define SOME_GUARD ... some content #endif SOME_GUARD As you see #endif SOME_GUARD is wrong, standard defines that #endif is followed by a newline. Unfortunately wave threw an exception. Calling what member on the std::exception instance caught resulted in the string: boost::wave::preprocess_exception I spent some time to search where that happened and which line was guilty. Is there any way for you to write the file and line which caused the error? I assume smth bad happens inside the wave lib, since I even can not compare curren and end iterator tokens. With Kind Regards, Ovanes Markarian

2 3