- Boost - lists.preview.boost.org

[move] New version with move-aware containers library
by Ion Gaztañaga 03 May '09

03 May '09

Hi to all, Sorry for the delay, but finally I've found some time to uploaded a new version of Boost.Move: Sandbox: https://svn.boost.org/trac/boost/browser/sandbox/move ZIP file with docs: http://www.drivehq.com/web/igaztanaga/move.zip Online docs: http://www.drivehq.com/web/igaztanaga/libs/move/ Changes: -> Changed folder from boost/move_semantics to boost/move. -> I've extracted all the containers from Boost.Interprocess, put them in boost/container and they only depend on official Boost SVN code (I haven't tested it, but it should be also nice with Boost 1.38). There is no documentation for these containers. -> Fixed a bug with EDG-based compilers. -> Note: I haven't applied yet the patch for non-sfinae compilers. With this changes, I think move emulation is quite portable (at least MSVC, EDG and GCC). Now that containers are independent from Boost.Interprocess, I would like to ask about the next step. If the move library is added, where should we put those containers? I would like to just drop common (vector, list, ...) Boost.Interprocess containers and just maintain these move-aware ones. If move emulation is considered mature enough, we should also plan when could we port all custom move-emulations present in several boost libraries to the "official" move library. Best, Ion

8 20

[ANN] Boost 1.39.0
by Beman Dawes 02 May '09

02 May '09

Boost 1.39.0 is available for download from SourceForge. See http://sourceforge.net/project/showfiles.php?group_id=7586&package_id=8041&… New Libraries: Signals2. Updated Libraries: Asio, Flyweight, Foreach, Hash, Interprocess, Intrusive, Program.Options, Proto, PtrContainer, Range, Unordered, Xpressive. Updated Tools: Boostbook, Quickbook. Plus the usual bug fixes to numerous libraries. Please report any problems to the Boost users or developers mailing list. Thanks, --Beman Dawes, Daniel James, and all the other Boosters who help get releases out

2 1

[gil] Google analytics script tags
by Daniel James 02 May '09

02 May '09

Hi, If you've missed it, there was a recent thread about using google analytics in offline copies of boost and the consensus was firmly against them. The gil documentation includes the google analytics javascript but doesn't seem to use it. So I assumed that their inclusion was a mistake and went ahead and removed them on trunk. I hope that wasn't presumptuous. But I did this on the generated documentation as I couldn't work out how it is generated. It'd useful to have some brief instructions and any necessary files. Ideally we should be able to generate them with boost build but a manual doxygen call is fine. thanks, Daniel

3 4

[chrono] monotonic_clock ofr cygwin
by vicente.botet 02 May '09

02 May '09

Hi, The Boost.Chrono do not compiles on cygwin. The MONOTONIC_CLOCK is not available. I have just defined the monotonic_clock as system_clock and all works right. Attached the patch. Do you agree to commit it? Best, Vicente

2 1

Intentions of Boost.TR1
by David Abrahams 02 May '09

02 May '09

I just saw this post on comp.lang.c++.moderated. I wonder if the Boost.TR1 documentation shouldn't clarify its intended use or degree of conformity. I don't think TR1 conformity is something we have been paying much attention to, is it? -- Dave Abrahams BoostPro Computing http://www.boostpro.com

5 4

German Boost book updated
by Boris Schaeling 02 May '09

02 May '09

For those who know German: I've added five more chapters to my Boost book at http://www.highscore.de/cpp/boost/. There are now altogether 16 chapters covering 29 Boost libraries. The book does not try to replace the documentation but is written in a tutorial style to give developers a kick start. It explains what the numerous libraries are good for and how they are used without going too much into detail. Any feedback is, as always, appreciated! Boris

2 2

Boostcon Boost++0x Session
by Michael Caisse 02 May '09

02 May '09

Dave - I am sorting out which sessions to attend and wish there was more of me. For the Boost++0x sessions reading a tutorial is a prerequisite. Is the tutorial on the boostcon site and I just failed to find it? Thank you- michael -- ---------------------------------- Michael Caisse Object Modeling Designs www.objectmodelingdesigns.com

2 3

how to build boost version 1_34_1 with gcc in non-standard location?
by Oliver Kullmann 01 May '09

01 May '09

Hello, I'm wrestling here for a long time with the following problem, and I would be very glad if somebody could help me here (seems to be straight-forward, but I can't find anything anywhere on the subject): I have to build some C++ code which doesn't compile with the system gcc version 4.3.1. So I have installed gcc version 4.1.2 locally. Now that C++ code uses Boost version 1_34_1, and doesn't compile with higher versions. Alright, so I install 1_34_1; very unfortunately, linking is necessary, and here is the problem: Obviously I have to build Boost 1_34_1 with the locally build gcc 4.1.2, and now how can I tell the Boost-build where to find gcc??? What I found is to use PACKAGEPATH/bjam/bjam --toolset=gcc-4.1.2 --toolset-root=PATH_TO_GCC_DIR --prefix=INSTALLATION_DIR --build-dir=BUILD_DIR "-sGCC_ROOT_DIRECTORY=PATH_TO_GCC_DIR" install --without-python However, as I had to found out after some time, this doesn't have any effect --- Boost just builds with the system compiler, and ignores all these specifications of PATH_TO_GCC_DIR. Now what is the right way to do it?? I would be really glad for help. Oliver

2 2

[Graph] transitive reduction
by boese-wolf＠math.uni-frankfurt.de 30 Apr '09

30 Apr '09

Hi! May I propose the following implementation for a transitive reduction to be added to the boost graph library? I'm a novice and the quality of the code I produced surely does not meet boosts standards, but maybe with some help, I'm able to improve it. So, please don't shoot me. #ifndef MY_OWN_TRANSITIVE_REDUCTION_HPP_INCLUDED #define MY_OWN_TRANSITIVE_REDUCTION_HPP_INCLUDED #include <vector> #include <algorithm> //std::find #include <boost/concept/requires.hpp> #include <boost/concept_check.hpp> #include <boost/graph/topological_sort.hpp> #include <boost/graph/graph_traits.hpp> #include <boost/property_map.hpp> //I actually didn't care which of those heades include some of the //others and whicht could be left out, for this reason, I simply added a //header when I used a function, described to be in this header by the //boost graph documentation //also I didn't got all of the concepts thin. Am I suppose to check //for all concepts, which are needed for functions I call? (As if I //wouldn't do that, the users would see the functions called by //complaining about missings concepts, which would be clearly an error //message revealing internal implementation and should therefore be avoided?) //the pseudocode which I followed implementing this algorithmn was taken //from the german book Algorithmische Graphentheorie by Volker Turau //it is proposed to be of O(n + nm_red ) where n is the number //of vertices and m_red is the number of edges in the transitive //reduction, but I think my implementation spoiled this up at some point //indicated below using namespace boost; template <typename Graph, typename GraphTR, typename G_to_TR_VertexMap, typename VertexIndexMap> BOOST_CONCEPT_REQUIRES( ((VertexListGraphConcept< Graph >)) ((IncidenceGraphConcept< Graph >)) ((MutableGraphConcept< GraphTR >)) ((ReadablePropertyMapConcept< VertexIndexMap, typename graph_traits<Graph>::vertex_descriptor >)) ((Integer< typename property_traits< VertexIndexMap >::value_type >)) ((LvaluePropertyMapConcept< G_to_TR_VertexMap, typename graph_traits<Graph>::vertex_descriptor >)), (void)) transitive_reduction( const Graph& g, GraphTR& tr, G_to_TR_VertexMap g_to_tr_map, const VertexIndexMap & g_index_map ) { typedef typename graph_traits<Graph>::vertex_descriptor Vertex; typedef typename graph_traits<Graph>::vertex_iterator VertexIterator; typedef typename std::vector<Vertex>::size_type size_type; std::vector<Vertex> topo_order; topological_sort( g, std::back_inserter( topo_order ) ); std::vector<size_type> topo_number_storage( num_vertices(g) ); iterator_property_map<size_type*, VertexIndexMap, size_type, size_type&> topo_number( &topo_number_storage[0], g_index_map ); { typename std::vector<Vertex>::reverse_iterator it = topo_order.rbegin(); size_type n = 0; for(; it != topo_order.rend(); ++it,++n ) { topo_number[ *it ] = n; } } std::vector< std::vector< bool > > edge_in_closure(num_vertices(g), std::vector<bool>( num_vertices(g), false)); { typename std::vector<Vertex>::reverse_iterator it = topo_order.rbegin(); for( ; it != topo_order.rend(); ++it ) { g_to_tr_map[*it] = add_vertex(tr); } } for( typename std::vector<Vertex>::iterator it = topo_order.begin(); it != topo_order.end(); ++it ) { size_type i = topo_number[ *it ]; edge_in_closure[i][i] = true; std::vector<Vertex> neighbors; //I have to collect the successors of *it and traverse them in //ascending topological order. I didn't know a better way, how to //do that. So what I'm doint is, collection the successors of *it here { typename Graph::out_edge_iterator oi,oi_end; for( tie(oi, oi_end) = out_edges( *it, g ); oi != oi_end; ++oi ) { neighbors.push_back( target( *oi, g ) ); } } { //and run through all vertices in topological order for( typename std::vector<Vertex>::reverse_iterator rit = topo_order.rbegin(); rit != topo_order.rend(); ++rit ) { //looking if they are successors of *it if( std::find( neighbors.begin(), neighbors.end(), *rit) != neighbors.end() ) { size_type j = topo_number[ *rit ]; if( not edge_in_closure[i][j] ) { for(size_type k = j; k < num_vertices(g); ++k) if( not edge_in_closure[i][k] ) //here we need edge_in_closure to be in topological order, edge_in_closure[i][k] = edge_in_closure[j][k]; //therefore we only access edge_in_closure only through //topo_number property_map add_edge(g_to_tr_map[*it], g_to_tr_map[*rit], tr); } //if ( not edge_in_ } //if (find ( } //for( typename vector<Vertex>::reverse_iterator } // { } //for( typename vector<Vertex>::iterator } //void transitive_reduction #endif //#ifndef MY_OWN_TRANSITIVE_REDUCTION_HPP_INCLUDED Best regards, Eric

2 3

[ANN] GCC 4.4.0 Released
by Eric Niebler 29 Apr '09

29 Apr '09

This might interest some here: http://gcc.gnu.org/ml/gcc/2009-04/msg00590.html It'd be nice to get some test coverage. Any takers? -- Eric Niebler BoostPro Computing http://www.boostpro.com

5 6

[multi-index] Changing hash_index_node.hpp to double linked list
by brad higgins 29 Apr '09

29 Apr '09

I am using boost multi-index, with a hashed-non-unique lookup. In my data, I am expecting many hash collisions. When I erase the multi- index, I find very poor performance. On investigation, I found that the hash collision data structure is a single linked list, and that removing an element involves traversing the entire list. Removing all of the elements involves traversing the list many times. Since my data has many collisions, I am traversing a very long list very many times. I know that the hash table is not designed for use with many collisions, but it is still faster in my testing on insert and serialization than the ordered_non_unique index specifier. I may end up using ordered_non_unique to avoid the worst-case erase() performance, and it may be good enough for the insert() performance. However, I would like to investigate making the hash collisions double- linked. I tried editing boost/multi_index/detail/hash_index_node.hpp, to make it a double-linked list, but I am seeing one of the included tests fail (test_safe_mode). I can't spot any bugs in my hash_index_node.hpp changes. What, apart from hash_index_node.hpp, needs to be updated, to support a double- linked list? boost/multi_index/detail/bucket_array.hpp? Others? Thanks, Brad

3 3

[encrypted strings]
by Sid Sacek 29 Apr '09

29 Apr '09

Does boost have any compile-time classes for string encryption? Is it even possible? When a hacker dumps an executable, they can see all of the strings the program might use, and some of those strings may contain sensitive information. Does boost have any classes that can encode the strings at compile-time? Ideally, the third string in the code below would never compile the "secret" string into the final binary. Regards, -Sid void foo( void ) { char const * s1 = "abc"; string s2 = "xyz"; encrypted_string s3( "secret" ); }

11 28

Re: [boost] [multi-index] Changing hash_index_node.hpp to double linked list
by JOAQUIN M. LOPEZ MUÑOZ 28 Apr '09

28 Apr '09

________________________________________ De: boost-bounces(a)lists.boost.org [boost-bounces(a)lists.boost.org] En nombre de brad higgins [bhiggins(a)arbor.net] Enviado el: martes, 28 de abril de 2009 21:56 Para: boost(a)lists.boost.org Asunto: Re: [boost] [multi-index] Changing hash_index_node.hpp to double linked list > Hi Joaquín, > Thanks for the quick response. I have attached the files I edited. > > Thank you very much for the help, Please allow a day or two till I find time to delve into this. I'll get back to you. Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

1 0

[interprocess] patch for AIX
by Etienne PIERRE 28 Apr '09

28 Apr '09

Hello, Currently interprocess doesn't work on AIX for two reasons : - detail/os_thread_functions.hpp uses ::pthread_equal but this fails because pthread_equal is a macro under AIX and it fails because of the :: prefix - in detail/atomic.hpp there is no implementation of PowerPC under AIX so I added one that uses the combination of lwarx and stwcx for each function to implement. I saw that under AIX, the detail/atomic_count.hpp and detail/sp_counted_base.hpp falls back to a pthread implementation, I could submit a patch for a real atomic implementation under AIX but I wonder if there are other places in boost where this should be done ? Thanks, Etienne PIERRE

2 2

[Boost.any] Nightly trunk testing failures
by K. Noel Belcourt 28 Apr '09

28 Apr '09

Hi, Most nightly trunk testing results are failing with this error message. Have their been recent changes to Boost.any that could have caused this? # Making "/var/scratch/boost/results" directory... # Starting tests ("/var/scratch/boost/tools_bjam/bin.linuxx86/bjam" "- sBOOST_BUILD_PATH=/var/scratch/boost:/var/scratch/boost/tools_bb" "- sBOOST_ROOT=/var/scratch/boost/boost" "--boost=/var/scratch/boost/ boost" "--boost-build=/var/scratch/boost/tools_bb" "--debug- configuration" -l300 intel-10.0 -d2 preserve-test-targets=off --dump- tests -j8 link=static "--build-dir=/var/scratch/boost/results" >>"/ var/scratch/boost/results/bjam.log" 2>&1)... # Running Boost.Build tests Traceback (most recent call last): File "run.py", line 60, in <module> runner(root) File "/var/scratch/boost/tools_regression_src/regression.py", line 193, in __init__ self.main() File "/var/scratch/boost/tools_regression_src/regression.py", line 551, in main getattr(self,action_m)() File "/var/scratch/boost/tools_regression_src/regression.py", line 516, in command_regression self.command_test() File "/var/scratch/boost/tools_regression_src/regression.py", line 317, in command_test self.command_test_boost_build() File "/var/scratch/boost/tools_regression_src/regression.py", line 354, in command_test_boost_build "boost/bin.v2/libs/any/test/any_test.test")); OSError: [Errno 2] No such file or directory: '/var/scratch/boost/ results/boost/bin.v2/libs/any/test/any_test.test' We get the exact same error message on all Sandia testers, and from the looks of it, most other testers are having the same problem. Any help in getting testing going again would be appreciated. Thanks. -- Noel

3 4

mpl::print and the sun compiler
by Dmitry Goncharov 28 Apr '09

28 Apr '09

The sun c++ compiler doesn't issue a warning about comparison between signed and unsigned integer expressions. The attached patch makes mpl::print produce "implicit int is not supported" warning. With the patch applied the compiler output looks like this: $ CC +w -c print.cpp "/home/dgoncharov/boost/trunk/boost/mpl/print.hpp", line 55: Warning: Implicit int is not supported in C++. "print.cpp", line 3: Where: While specializing "boost::mpl::print<int>". "print.cpp", line 3: Where: Specialized in non-template code. 1 Warning(s) detected. Br, Dmitry

1 0

Re: [boost] [encrypted strings]
by Kasra 28 Apr '09

28 Apr '09

Sid Sacek Wrote: >This sounds very cool. Let me know when you're done. I would definitely >use it. For your needs the library is fully implemented, you don't need to look into RC4 implementation since the library already has many algorithms. On your encrypted string problem, it is best to user during installation techniques. Create an encrypted archive where the user has to provide a product key for the installation to commence. Read decrypted contents from the archive, encrypt it using any encryption algorithm you like however, use a PRNG to generate the key. Seed the PRNG from the available hardware and host OS and etc. Then each time you want the plaintext you regenerate your PRNG then decrypt your ciphertext. -- Kasra

2 1

Re: [boost] [asio] send blocks with MSG_DONTWAIT (linux)
by Stefan Pledl 28 Apr '09

28 Apr '09

> You can switch your socket in the non-blocking mode with > basic_stream_socket::non_blocking_io(). > Though asio should handle MSG_DONTWAIT as well. > > BR, Dmitry Hi Dimitry, that's right, but i need a blocking read and a non-blocking wirte. With basic_stream_socket::non_blocking_io() both, read and write are non-blocking. When i use C-functions read is blocking and write is non-blocking (when passing MSG_DONTWAIT). Best regards Stefan

1 0

Help with VMS port of CMake
by Bill Hoffman 27 Apr '09

27 Apr '09

Hi, I am trying to port CMake to VMS. For the first pass I am going to use the GNV tools and gmake. I am very close, but I am having trouble with files that have more than one "." in the name. CMake uses files like that in several places, as do many CMake based projects. Assuming I use POSIX paths on VMS, does anyone on this list know what magic the GNV cat program uses to handle the following: This works: cat /idisk1/devel/Windows-cl.cmake.in I wrote a simple C++ program that uses fstream to open argv[1], and get the following results: These work: ./test. Windows-cl.cmake.in ./test. Windows-cl^.cmake.in ./test. idisk1:[devel]Windows-cl^.cmake.in This does not: ./test. /idisk1/devel/Windows-cl.cmake.in Any help would be appreciated. Sorry if this is not very boost related, but in previous CMake discussions on the list the topic of VMS came up, so I thought there might be someone that could help me out. Thanks. -Bill

1 0

[reference_wrapper] feature request
by Thorsten Ottosen 27 Apr '09

27 Apr '09

Hi, It came up again in the RangeEx review that we really need reference_wrapper<T> to implement a forwarding operator(). Personally I don't care much for more than one or two arguments, but I guess others will. Thanks -Thorsten

2 1

Re: [boost] [encrypted strings]
by Kasra 27 Apr '09

27 Apr '09

Hi, I am working on a Crypto library for that could do compile time encryption/decryption, however, requires 'constexpr' support from the compiler: template<class T> constexpr void encrypt(T& cipher_text, const T& plain_text) const throw(bad_state); template<class T> constexpr void decrypt(T& cipher_text, const T& plain_text) const throw(bad_state); -- Kasra

2 1

Re: [boost] [encrypted strings]
by raindog＠macrohmasheen.com 27 Apr '09

27 Apr '09

Boost does not have one. If I were you however I would look into a commercial protector such as themida (www.oreans.com) or armadillo or asprotect. ------Original Message------ From: Sid Sacek Sender: boost-bounces(a)lists.boost.org To: boost(a)lists.boost.org ReplyTo: boost(a)lists.boost.org Sent: Apr 27, 2009 8:21 AM Subject: [boost] [encrypted strings] Does boost have any compile-time classes for string encryption? Is it even possible? When a hacker dumps an executable, they can see all of the strings the program might use, and some of those strings may contain sensitive information. Does boost have any classes that can encode the strings at compile-time? Ideally, the third string in the code below would never compile the "secret" string into the final binary. Regards, -Sid void foo( void ) { char const * s1 = "abc"; string s2 = "xyz"; encrypted_string s3( "secret" ); } _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost Sent from my Verizon Wireless BlackBerry

2 1

[tokenizer] Deprecated class used by default
by Boris Schaeling 27 Apr '09

27 Apr '09

I just noticed that although the class char_delimiters_separator is deprecated it is used by default by Boost.Tokenizer. I don't know when char_delimiters_separator was deprecated in favor of char_separator but it goes back to 1.33.1 (see http://www.boost.org/doc/libs/1_33_1/libs/tokenizer/index.html) Any plan to change the default class soon (doesn't need to be in 1.39.0 of course as it's pretty late for such a change)? Boris

1 0

beta docs on the website
by Michael Marcin 27 Apr '09

27 Apr '09

Shouldn't http://beta.boost.org/doc/ have 1.39 docs at this point? -- Michael Marcin

2 1

[asio] send blocks with MSG_DONTWAIT (linux)
by Stefan Pledl 27 Apr '09

27 Apr '09

If i pass the flag MSG_DONTWAIT to basic_stream_socket::send(buffers, flags) the call blocks in the following situation. -> Connection is established (boost::asio::socket_base::send_buffer_size option(2048) is set) -> sending data form local endpoint to remote endpoint -> remote endpoint does not read the data due to software bug/crash etc. -> sendqueue (see netstat) on local endpoint runs full -> send on local endpoint blocks at this point I think the flag MSG_DONTWAIT should avoid this blocking call. If i do the same using C-function send(fd, buf*, len, flag) the call does not block. I have analyzed the code and i think the problem is in send function in asio/detail/reactive_socket_service.hpp socket_ops::send returns but socket_ops::poll_write blocks. I have added the following changes (see comments __SP__) and all works ok for me. Is the blocking desired or where is my mistake? modified code form asio/detail/reactive_socket_service.hpp ---------------------------------------------------------- // Send the given data to the peer. template <typename ConstBufferSequence> size_t send(implementation_type& impl, const ConstBufferSequence& buffers, socket_base::message_flags flags, boost::system::error_code& ec) { if (!is_open(impl)) { ec = boost::asio::error::bad_descriptor; return 0; } // Copy buffers into array. socket_ops::buf bufs[max_buffers]; typename ConstBufferSequence::const_iterator iter = buffers.begin(); typename ConstBufferSequence::const_iterator end = buffers.end(); size_t i = 0; size_t total_buffer_size = 0; for (; iter != end && i < max_buffers; ++iter, ++i) { boost::asio::const_buffer buffer(*iter); socket_ops::init_buf(bufs[i], boost::asio::buffer_cast<const void*>(buffer), boost::asio::buffer_size(buffer)); total_buffer_size += boost::asio::buffer_size(buffer); } // A request to receive 0 bytes on a stream socket is a no-op. if (impl.protocol_.type() == SOCK_STREAM && total_buffer_size == 0) { ec = boost::system::error_code(); return 0; } // Send the data. for (;;) { // Try to complete the operation without blocking. int bytes_sent = socket_ops::send(impl.socket_, bufs, i, flags, ec); // Check if operation succeeded. if (bytes_sent >= 0) return bytes_sent; // Operation failed. if ((impl.flags_ & implementation_type::user_set_non_blocking) || (ec != boost::asio::error::would_block && ec != boost::asio::error::try_again)) return 0; // __SP__ // do not poll if flag MSG_DONTWAIT is set if (flags & MSG_DONTWAIT) return 0; // __SP__ // Wait for socket to become ready. if (socket_ops::poll_write(impl.socket_, ec) < 0) return 0; } }

2 1