February 2009 - Boost - lists.preview.boost.org

warnings from gcc-4.3.3 -Wall
by John Salmon 04 Feb '09

04 Feb '09

I recently upgraded to gcc/4.3 and started seeing warnings from gcc-4.3.3 -Wall suggesting parentheses around && within ||. It would be great if boost/1.38 could repair these by adding a few pairs of parentheses before the final release. N.B. I did not do an exhaustive compilation of everything in boost with gcc -Wall. The three cases below are the ones that my code base just happens to hit. Thanks, John Salmon /proj/desres/root/Linux/x86_64/boost/1_37_0-14/Release/include/boost/tuple/tuple_comparison.hpp:73: warning: suggest parentheses around && within || /proj/desres/root/Linux/x86_64/boost/1_37_0-14/Release/include/boost/iostreams/filter/symmetric.hpp:126: warning: suggest parentheses around && within || /proj/desres/root/Linux/x86_64/boost/1_37_0-14/Release/include/boost/iostreams/detail/streambuf/indirect_streambuf.hpp:285: warning: suggest parentheses around && within ||

2 1

[mpl] holder class for types with external linkage (similar to integral contants)
by klaus triendl 04 Feb '09

04 Feb '09

Recently, I've been in need of a holder class for strings with external linkage but I didn't find an mpl datatype that could wrap an externally defined entity. Instead of creating a quick solution working for my needs I thought the mpl library could benefit from a generic holder class similar to the integral constant holders. <code> template<typename T, T N> struct external_c { // make it a nullary metafunction typedef external_c type; typedef T value_type; // correct? requirements? typedef integral_c_tag tag; // can't make a static compile time constant but only a static // function static value_type value() { return N; } // implicitly convertible to its runtime type value operator value_type() const { return this->value(); } }; // e.g. extern const char string1[]; const char string1[] = "string1"; extern const std::string string2; const std::string string2("string2"); int main() { external_c<const char* const, string1> x1; external_c<const std::string&, string2> x2; std::cout << x1 << "\n"; std::cout << x2.value() << "\n"; return 0; } </code> The name 'external_c' is just a quick idea and would have to be changed anyway with c++09's possibility to use types with internal linkage in template programming. Would it be reasonable to integrate this into the mpl? Any comments? Klaus Triendl

1 0

[Serialization] Need help with binary archive deserialization error
by George van Venrooij 04 Feb '09

04 Feb '09

Hi, I just ran into an issue with serialization which is very peculiar. When serializing a complex class structure to an archive all goes well without errors. However, when trying to read back the data it works with all xml and text archive types, but when using binary archives I get an archive exception while loading. The exception occurs while loading an array of 3 doubles. From what I can make up out of the code, the archive says there are 20 bytes to load, while the library expects 24. Apparently I'm doing something that has no effect for text-based serialization, but does for binary. Since I didn't find any posts or tickets for this issue I assume I have made a mistake somewhere, only I have been looking for hours and cannot seem to find it. So any suggestions on where to look would be welcome, even the obvious ones ;-) Regards, George van Venrooij

4 4

[interprocess] Severe bug in sp_counted_impl.hpp!
by Lars Hagström 04 Feb '09

04 Feb '09

Hi, I believe that there is a bug in sp_counted_impl_pd::destroy(). Currently the method looks like this: void destroy() // nothrow { //Self destruction, so get a copy of the allocator //(in the future we could move it) this_allocator a_copy(*this); BOOST_ASSERT(a_copy == *this); this_pointer this_ptr (this); //Do it now! scoped_ptr<this_type, scoped_ptr_dealloc_functor<this_allocator> > (this_ptr, a_copy); typedef typename this_allocator::value_type value_type; detail::get_pointer(this_ptr)->~value_type(); } What I think is wrong is on the deallocator functor line. I believe that it should read scoped_ptr<this_type, scoped_ptr_dealloc_functor<this_allocator> > DEALLOCATOR(this_ptr, a_copy); The missing thing is the name for the scoped_ptr instance (the one I called DEALLOCATOR, to make it stand out). And since the name is missing it will become an anonymous instance, which will be destroyed *before* the destructor is called on this_ptr. The upshot is that with the bug we deallocate the memory and then call the destructor, rather than the opposite. If you're unlucky some other thread/process has allocated that memory for some other use, and you will then be running the destructor on that instead... I'm writing a ticket on this as well, just so it doesn't get overlooked. (If I'm wrong about all this my face will be red for a while...) Cheers Lars

4 6

boost serialization and alignment
by meir shoham 04 Feb '09

04 Feb '09

Hi, When I'm setting struct member alignment as 1 BYTE in visual studio 2005 project environment (C++), I receive alignment error coming from boost serialization library (STATIC_ASSERT) at compilation time. This problem does not happen when I wrap the structures containing the boost serialization template methods using #pragma pack(push,1) and #pragms pack(pop). I'm using boost only to serialize/deserialize strctures and classes to/from XML. Is there a problem for boost serialization library to handle alignment other then the default one? Thanks, Meir

1 0

Re: [boost] Coverity Static Code Analysis
by Michael Fawcett 03 Feb '09

03 Feb '09

On Tue, Feb 3, 2009 at 3:10 PM, Marshall Clow <marshall(a)idio.com> wrote: > > fixed, but I think that someone with a bit of legal training needs to look > at the license that Coverity wants people to agree to before using the scan > results. [ It looks pretty harmless to me, but IANAL ] > > <http://scan.coverity.com/policy.html#license> I agree. Doesn't Boost have access to just that sort of support? --Michael Fawcett

1 0

Interest in a library for building simulations of dynamical systems?
by Jay Graham 03 Feb '09

03 Feb '09

Is there interest in a library for building simulations of dynamical systems represented by time-based differential equations? The proposed library would implement ideas found in the C++ Model Developer (CMD) library developed by Ray Sells, but in a more generic way. For example, the proposed library would support the use of user defined data types when defining states and their derivatives. Other CMD features to be included in the proposed library are: 1. The ability to mix both continuous and discrete events. 2. Support for a simulation structure that makes it easy to go from a mathematical description of a system to code. 3. Support for time-varying system topologies (e.g., multi-stage missiles). 4. User extendable set of numerical integration algorithms. For more information on CMD see: http://www.dtic.mil/srch/doc?collection=t2&id=ADA433836 and http://rocket.itsc.uah.edu/u/education/files/R/IPT_Producer/AIAA/Sells/Sell… Code snippet: // Define states and their derivatives struct x_pos : state< double, // state type double // derivative type > {}; struct y_pos : state< double, double > {}; // Define a simulation object struct Ball: obj< rk4, // fourth-order Runge-Kutta numerical integration algorithm mpl::vector< x_pos, y_pos > // State vector > { Ball(double x, double y, double vx, double vy) { // Initial state s_<x_pos>().value = x; s_<y_pos>().value = y; this->vx = vx; this->vy = vy; // Integrators add_integrator<x_pos>(this->vx); add_integrator<y_pos>(this->vy); } template< typename S, typename C > inline void update(S &sim, C &clock) { // Update the derivatives vx = 0.8*vx; vy = 0.8*vy; // Use a discrete event to stop the simulation // after 10 seconds if (clock.event(10.0)) { sim.halt(); } } template< typename C > void report(C &clock) { // Report current state at 1 second intervals if (clock.sample(1.0)) { std::cout << "Ball" << clock.t << " " << s_<x_pos>().value << " " << s_<y_pos>().value << "\n"; } } double vx; double vy; }; // Define the integration time-step struct dts { typedef double time_type; static time_type value() { return 0.01; } }; // Define a simulation stage struct stage_1 : stage< dts, mpl::vector<Ball> > {}; main() { // Create a simulation object Ball ball(0.0, 0.0, 10.0, 5.0); // Create the simulation execution object sim< mpl::vector< stage_1 > > sim(); // Initialize the simulation stage sim.stage_<stage_1>().obj_<Ball>() = &ball; // Run the simulation run(sim); }

7 7

[parallel_sort] Proposal
by Edouard A. 03 Feb '09

03 Feb '09

Howdy, I just finished the first working version of parallel_sort. It seems there is a lot of interest for this function in the C++ community, and in the boost community specifically, which is the motivation for this post. ;) My implementation is somewhat very different in nature from the one present in Intel's TBB. It's designed to take advantages of a multi-core user machine, it's not designed to run on a grid for massive sorting operation. It only uses the STL and the boost library. It doesn't have any fancy prerequisites, except having several cores to unleash its power! Current benchmarks on my Q6600 vs2008-64-bit default allocator: 2 threads : 160 % faster 4 threads : 260 % faster Keep in mind the Q6600 is not a real equivalent to a quad-cpu machine and that the default memory allocator is not multithreading friendly. It can be heavily customized. I already offer the possibility to choose between quick sorting and merge sorting, and for quick sorting I offer two pivot algorithms : fast or secure. Then you can of course specify a predicate and a fallback sorting algorithm for when you run out of threads. If you don't care to customize, no problem! Just write: parallel_sort<2>(v.begin(), v.end()); // sort with 2 threads There is still some work to be done, especially in the documentation area. Nevertheless, I think it's time to open up the code and start getting feedback. J As an added bonus I also offer to add my implementation of the smoothsort algorithm (with some hints from Steven). This algorithm is faster on sorted input but slower on random input. Cheers -- EA

6 20

SCons script for building Boost
by Sergey Popov 03 Feb '09

03 Feb '09

I've searched the ML archive and found out that scons wasn't considered for adoption because nobody offered to reimplement boost build instructions in terms of SCons. I'd like to work on this. Is there a chance I'll be given commit access so I could make a boost-SCons branch? ;) I've attached my early WIP build script to this posting.

4 4

Re: [boost] Review process question
by Christophe Henry 03 Feb '09

03 Feb '09

"John Maddock" schrieb im Newsbeitrag news:<343DA106BE604F22B7EF1BFF0A8515B8@acerlaptop>... > >> It is enough to specify which revision that is to be reviewed. > > > > Is it possible to change it after the submission request? It would allow > > me > > to make some bugfixes (if necessary) while being in the queue. > > I think bugfixes at that stage are fine: as long as there is a stable > version available somewhere, and as long as all the reviewers are refering > to the same version during the review - in other words try and keep things > stable during the review itself. > > HTH, John. > > _______________________________________________ > Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost > I think I got it now. Thanks! Christophe

1 0