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

Re: [boost] linking on OSX
by David M. Cotter 22 Oct '09

22 Oct '09

>> please be explicit, are you saying i should type exactly this: >> >> ./bjam macosx-version-min=10.4 architecture=combined >> linkflags="-Wl,-force_cpusubtype_ALL" --with-thread -a -n install > > I don't know how to use bjam, nor do I have any wish to learn, so I > can't help you with its command line arguments. > Did doing that give you -Wl,-force_cpusubtype_ALL for the g++ > command that creates libboost_thread.dylib? apparently the extra -Wl is not necessary, since when i leave it off, it still gets sent, probably due to the macosx-min-version flag. WITH: > ./bjam macosx-version-min=10.4 architecture=combined linkflags="-Wl -force_cpusubtype_ALL" --with-thread -a -n install "g++" -dynamiclib -Wl,-single_module -install_name "libboost_thread.dylib" -o "bin.v2/libs/thread/build/darwin-4.2.1/ release/architecture-combined/macosx-version-min-10.4/threading-multi/ libboost_thread.dylib" "bin.v2/libs/thread/build/darwin-4.2.1/release/ architecture-combined/macosx-version-min-10.4/threading-multi/pthread/ thread.o" "bin.v2/libs/thread/build/darwin-4.2.1/release/architecture- combined/macosx-version-min-10.4/threading-multi/pthread/exceptions.o" "bin.v2/libs/thread/build/darwin-4.2.1/release/architecture-combined/ macosx-version-min-10.4/threading-multi/pthread/once.o" - headerpad_max_install_names -mmacosx-version-min=10.4 -Wl,-dead_strip - no_dead_strip_inits_and_terms -arch i386 -arch ppc -Wl - force_cpusubtype_ALL note: see it's in there TWICE? is the problem that i'm missing a comma? WITHOUT: > ./bjam macosx-version-min=10.4 architecture=combined linkflags="- force_cpusubtype_ALL" --with-thread -a -n install "g++" -dynamiclib -Wl,-single_module -install_name "libboost_thread.dylib" -o "bin.v2/libs/thread/build/darwin-4.2.1/ release/architecture-combined/macosx-version-min-10.4/threading-multi/ libboost_thread.dylib" "bin.v2/libs/thread/build/darwin-4.2.1/release/ architecture-combined/macosx-version-min-10.4/threading-multi/pthread/ thread.o" "bin.v2/libs/thread/build/darwin-4.2.1/release/architecture- combined/macosx-version-min-10.4/threading-multi/pthread/exceptions.o" "bin.v2/libs/thread/build/darwin-4.2.1/release/architecture-combined/ macosx-version-min-10.4/threading-multi/pthread/once.o" - headerpad_max_install_names -mmacosx-version-min=10.4 -Wl,-dead_strip - no_dead_strip_inits_and_terms -arch i386 -arch ppc -force_cpusubtype_ALL

2 3

[conversion] Release v0.3 : Changing the order of assign_to parameters + `tie' function
by vicente.botet 22 Oct '09

22 Oct '09

Hi, I have just released the version 0.3 of Boost.Conversion , which includes since my last publicly announced version 0.1: Version 0.3.0, Oct 22, 2009 Changing the order of to and from parameters on assign_to + Added `tie' function Incopatibility: * Changing the order of to and from parameters on assign_to. * Now boost/conversion/convert_to.hpp and boost/conversion/assign_to.hpp files are separated. New Features: * Added <boost/conversion.hpp> global file. * Added mca() function. * Added convert_to_via function. Test: * Added test for the new features Version 0.2.0, Mai 16, 2009 Adding array + fusion::tuples + Adaptation to Boost 1.39 New Features: * conversion between boost::array of explicitly convertible types. * conversion between Boost.Fusion sequences of explicitly convertible types. You can found all the detail for dowloading, documentation, ... in http://svn.boost.org/trac/boost/wiki/LibrariesUnderConstruction#Boost.Conve… I'd appreciate it if people could take a look, test it on their compiler and give me some feedback. Best regards, _____________________ Vicente Juan Botet Escribá

1 0

Re: [boost] linking on OSX
by David M. Cotter 21 Oct '09

21 Oct '09

please be explicit, are you saying i should type exactly this: ./bjam macosx-version-min=10.4 architecture=combined linkflags="-Wl,- force_cpusubtype_ALL" --with-thread -a -n install cuz that didn't do anything On Oct 21, 2009, at 11:03 AM, Peter O'Gorman wrote: > On 10/21/2009 12:26 PM, David M. Cotter wrote: >> you can see that the build system is in fact emitting the correct >> command for linking, however it is apparently being ignored on PPC > >> >> malkier:lib davec$ otool -f -v ./libboost_thread.dylib >> Fat headers >> fat_magic FAT_MAGIC > >> architecture *ppc7400* >> cputype CPU_TYPE_POWERPC >> cpusubtype *CPU_SUBTYPE_POWERPC_7400* >> capabilities 0x0 > >> >> so i'm dead in the water on this project. any insight or help that >> solves this would win you free mac Karaoke software. :) and my >> gratitude. >> > > Change it to -Wl,-force_cpusubtype_ALL, the flag is not passed along > by gcc to the linker. > > If you build all your object files with the -mmacosx-version-min > flag, as far as I know, you shouldn't be getting ppc7400 in the > cpusubtype. > > Peter > -- > Peter O'Gorman > http://pogma.com

2 2

[BGL] Multithread problem inserting edges and exploring the same graph
by Cosimo Calabrese 21 Oct '09

21 Oct '09

Hi to all, I've a problem on BGL. My environment is WinXP, Visual Studio 2005, Boost 1.37. I'm trying to use BGL in the following way. I've a single graph in my application, that is an adjacency_list< vecS, vecS, ... >, and I've 2 kind of threads that work on the graph: - some "Explorer" threads, that executes a Dijkstra's algorithm repeatedly; - some "Integrator" thread, that call the add_edge() function repeatedly. The program often crashes, because it cannot dereference some graph elements, some auto_ptr... The crash isn't sistematic, so I think there's a concurrency problem. I think that the problem is in the use of the vectors in the implementation of the graph. In general, if I've an iterator 'it' that point on an vector's element A, and I insert another element B before A, then the iterator is now invalid. I suppose that it happens something of similar in the graph. If I read the out-edge list of a vertex at the same time that I add another out-edge to the same vertex, the reading may be inconsistent. So I've tried with an adjacency_list< listS, vecS, ... >, because if I add an element to a list, the previous reference are still valid, but I've experienced the same problem again. If what I've said is correct, then I think that with an adjacency_matrix I will not have problems, but I've approximately 5 millions of vertices, and the graph is sparse. I think that is a waste of memory. And what about the exploring time? Another question: if I launch *only* a set of Explorer thread, it seems that there aren't problems with EdgeList = listS and EdgeList = vecS too. Can someone confirm to me that the exploring algorithms, like Dijkstra, don't modify the graph structure? Are they read-only algorithms? Can I safely launch multiple exploration threads simultaneously? Can someone help me? Many thanks in advance, Cosimo Calabrese.

5 24

Re: [boost] Preliminary submission: generic linear algebra algorithms library Boost LA
by Daniel Pfeifer 21 Oct '09

21 Oct '09

Hi Emil, I had a look at the documentation and examples of your library and I think it is awesome so far. Your library could be a C++ complement to the builtin vector and matrix types and functions of GLSLang. (see http://www.opengl.org/registry/doc/GLSLangSpec.1.50.09.pdf pages: 23, 53ff, 60, 81ff). Since you already have mastered swizzling, implementing the rest should be peanuts. My wishlist: * component wise multiplication and division of two vectors implemented as operators (as in GLSL). * all the vector and matrix related functions described in chapter 8 of the GLSL spec. * operator[] As for the name: If LA is too misleading, I would propose "kd-math" or "math-kd". There already is another project called math3d, and since your library is not limited to three dimensions, kd (as in kd-tree) suits quite well in my opinion. cheers, Daniel Emil Dotchevski wrote: > I remember some past discussions on this mailing list about the need > for similar library in Boost, so I hope it is OK to skip the interest > query and request a preliminary review for Boost LA. > > Boost LA provides a set of generic linear algebra algorithms, > primarily for working with vectors and matrices of static size. The > library also defines vector and matrix data types, however it allows > users to introduce their own types by specializing the vector_traits > and matrix_traits templates. > > Documentation and source code available at > http://www.revergestudios.com/boost-la, released under the Boost > Software License. > > Thanks, > Emil Dotchevski > _______________________________________________ > Unsubscribe & other changes: > http://lists.boost.org/mailman/listinfo.cgi/boost

2 1

Re: [boost] linking on OSX
by David M. Cotter 21 Oct '09

21 Oct '09

okay i've stared at the doc for bjam but can find NOTHING which indicates how to tell it to build for 10.4 anyone have an idea on that? >>> This load command (see mach-o/loader.h) is LC_DYLD_INFO_ONLY. >>> >>> I can only guess, but did you compile boost with -mmacosx-version- >>> min=10.6, but built your app with an earlier deployment target, >>> and when you attempt to run it, you're doing so on Leopard? >> >> i did not specify a min version when i built. so what should i >> do? this is what i suspect: >> >> re-build boost using -mmacosx-version-min=10.4 >> >> and all will be well? > > Should work, though I don't know if you'll also have to point boost > at the 10.4 SDK with -isysroot. I suspect not.

4 17

BOOST_FOREACH like macro for map
by mobi phil 21 Oct '09

21 Oct '09

Hello, new to this list... did not find anything similar on net. One of the annoyances with std::map is that if you work with const maps, then there is no "one line" code to get an element knowing the key. given the map const map<string, string> aStringMap; I cannot write const string &s = aStringMap["key"]; process(s); I should write: map<string, string>::const_iterator it=aStringMap.find("key") if( it != aStringMap.end() ) { process(it->second) } so would be nice to have a macro like BOOST_CONSTMAPFIND(aStringMap, "key", data) { processResult( data ); } the version if(BOOST_CONSTMAPFIND(l_map, l_key, l_data) ) { } Seems to be impossible to implement -- rgrds, mobi phil being mobile, but including technology http://mobiphil.com

4 10

Re: [boost] linking on OSX
by David M. Cotter 21 Oct '09

21 Oct '09

okay we now know what the problem is, the question for the BOOST folk is: how do i do this with bjam? On Oct 20, 2009, at 7:23 PM, Chris Suter wrote: > Hi David, > > On Wed, Oct 21, 2009 at 1:03 PM, David M. Cotter <me(a)davecotter.com> > wrote: > okay. everything is working fantastically, however, it won't run on > G3 machine running 10.4.11 > it works perfectly on a G4 PPC machine also running 10.4.11 > > is there some way to "tune" bjam to emit G3-compatible code? > cuz i'm getting this: > > Host Name: TAZ-2 > Date/Time: 2009-10-20 05:56:18.099 -0700 > OS Version: 10.4.11 (Build 8S165) > Report Version: 4 > > Command: FooBar Pro > Path: /Applications/FooBar Pro.app/Contents/MacOS/FooBar Pro > Parent: WindowServer [310] > > Version: FooBar Pro 1.0d50r33 (1.0) > > PID: 445 > Thread: Unknown > > Link (dyld) error: > > Library not loaded: @loader_path/../SharedSupport/ > libboost_thread.dylib > Referenced from: /Applications/FooBar Pro.app/Contents/MacOS/ > FooBar Pro > Reason: no suitable image found. Did find: > /Applications/FooBar Pro.app/Contents/MacOS/../SharedSupport/ > libboost_thread.dylib: no matching architecture in universal wrapper > /Applications/FooBar Pro.app/Contents/MacOS/../SharedSupport/ > libboost_thread.dylib: no matching architecture in universal wrapper > > This is most likely because you're only building for G4 machines and > greater. If you run otool -f -v <binary>, you'll probably see > cpusubtype set to CPU_SUBTYPE_POWERPC_7400 (which basically means G4 > or greater). > > To fix, as a minimum, you need to make sure that your deployment > target is set to 10.4 or less and make sure your architecture > setting is ppc rather than ppc7400. > > Kind regards, > > Chris

1 0

[interprocess] new problem with move-aware containers
by Thorsten Ottosen 20 Oct '09

20 Oct '09

Hi Ion, I'm using the boost::interprocess::multiset from trunk with vc9. After updating, some of my code stopped working. I use a set like multiset<const Node*,SomePredicate> and when calling set.insert( &obj ) I get compile errors. Specifically in interprocess\detail\move.hpp(79) I don't understand the details of these move-emulation tricks, but the compiler complains that T=U* for some U is not a legal base class. AFAICT, the compiler is right: template <class T> class rv : public T { rv(); ~rv(); rv(rv const&); void operator=(rv const&); }; -Thorsten

2 1

bzip with spreadsort
by Steven Ross 20 Oct '09

20 Oct '09

I've added fasterbzip2-1.0.5.zip to the Boost Vault as an example application of the Spreadsort hybrid radix sorting algorithm. As bzip is written in C, I couldn't use the template library directly and had to modify the source. Using the Spreadsort algorithm causes a measurable speed improvement on Windows for the first 100MB of wikipedia from: http://cs.fit.edu/~mmahoney/compression/textdata.html<http://cs.fit.edu/%7Emmahoney/compression/textdata.html> along with other data I tested. Please note that bzip doesn't spend all its time compressing, already has two stages of radix sort, and has specialized and manually inlined comparison functions, so it's heavily optimized to begin with and improvements are minor.

1 0