- Boost - lists.boost.org

[1.52.0] Open for merging
by Marshall Clow 28 Aug '12

28 Aug '12

The release branch (https://svn.boost.org/svn/boost/branches/release) is open for merging all stable changes from trunk, including bug fixes and major upgrades to existing libraries. Breaking changes should be coordinated via the mailing list with the developers of the other libraries affected. New libraries may be added, but require permission of a release manager. In other words, merging *does not* require permission from a release manager, except for new libraries. Boost Calendar with the release schedule: http://www.boost.org/development/index.html Release schedule explanation: http://svn.boost.org/trac/boost/wiki/ReleaseSchedule Release practices for developers: http://svn.boost.org/trac/boost/wiki/ImprovingPractices -- The Release Managers, Beman Dawes Daniel James Eric Niebler Marshall Clow Rene Rivera Vladimir Prus

3 2

[contract] a possible future extension
by Andrzej Krzemienski 28 Aug '12

28 Aug '12

Hi Lorenzo, This is definitely outside the scope of the review, but I was wondering whether it would be feasible for the macros in Boost.Contract to handle base class aliases, as described in this proposal: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2881.html The positive "side effect" of your framework is that it has the potential to add various missing language features. I am wondering if these aliases proposed in N2881 are worth the effort. Regards, &rzej

2 1

a (useful?) crtp trick
by pavel 28 Aug '12

28 Aug '12

hi all i post this in the hope someone finds this technique useful to make long story short, here is a snippet: template<typename t> //tag deducing util struct base_tag { typedef decltype(get_tag((t*)42)) type; }; template<typename type, typename tag = typename base_tag<type>::type> struct Base //crtp base class { friend tag get_tag(type*); //helper declaration, never defined }; template<typename type> //no mention of tag void f(const Base<type>&) //tag is deduced automatically {} struct Foo : Base<Foo, void> //derived type {}; int main() { Foo foo; f(foo); } the idea is that if we have an associated type (or types) along with the "real" type in the crtp pattern we can make that associated type deduced automatically and even pretend there is no such tag leaving it e.g. for advanced usage more on this here: http://programmizm.sourceforge.net/blog/2012/because-we-can -- Pavel P.S. if you notice a grammar mistake or weird phrasing in my message please point it out

3 7

release manager for 1.52?
by Oliver Kowalke 28 Aug '12

28 Aug '12

Who is the release manager for 1.52?

2 1

[atomic] lockfree failure on msvc 64-bit
by Christoph Macheiner 27 Aug '12

27 Aug '12

As indicated by http://www.boost.org/development/tests/trunk/developer/output/cm_x64-boost-…. In boost/atomic/detail/interlocked.hpp, BOOST_ATOMIC_LLONG_LOCK_FREE must be defined to 2 if _WIN64 is defined (cas64strong.hpp is included afterwards). Because the library has not yet been released, I didn't bother to create a ticket and because the change is so trivial I didn't create a patch as well (if either or both is required or wanted, please let me know). Question: If I recall correctly, long long can be made lockfree also in 32-bit. If I am right, is there a reason why it hasn't been done? Thanks. --Christoph.

1 0

memory corruption in async_read_until using asio::ssl
by Jim Page 27 Aug '12

27 Aug '12

Afternoon All I am seeing an occasional memory corruption bug, using valgrind, in our web server application that is using boost::asio::ssl, and it seems to be related to calls to async_read_until, where several chunks of data are required before the terminating regex is found. What appears to be happening is that engine::perform() causes a write (via ssl3_read_bytes()) on a deleted memory region, the buffer into which decrypted data is written. This region was delete'd via invocation of the detail::io_op::handler, and in turn via read_until_delim_op::operator(), due to prepare() being called on the streambuf, and a std::vector<char> is delete'd in order to create a larger one. It seems like one of the chain of handlers is using a pointer which is becomes invalid when the streambuff is expanded after failing to find the regex terminator after one or more reads, and prepare() is called before the next async_read_some() on the underlying stream. I have included the valgrind output at the end - sorry for the large size. I am pretty certain that I have ruled out the obvious (multiple simultaneous attempts to read using the same streambuf), and I am only using one thread for the io_service. This seems a pretty common potential failure mode, so perhaps I am barking up the wrong tree - I'd appreciate anyone's opinion before I start a potentially tricky debug effort requiring modification to the boost libraries on our production servers. I'm using boost 1.49. There's a lot of optimization on the application which is why some calls are optimized out of the valgrind dump. The machines are running Centos 5.8, x86_64. Compiler is gcc 4.4.6. Thanks for looking Jim Page ==29870== Thread 22: ==29870== Invalid write of size 1 ==29870== at 0x4A088F8: memcpy (mc_replace_strmem.c:587) ==29870== by 0x3400821EDE: ssl3_read_bytes (in /lib64/libssl.so.0.9.8e) ==29870== by 0x340081E750: ??? (in /lib64/libssl.so.0.9.8e) ==29870== by 0x66A75B: boost::asio::ssl::detail::engine::perform(int (boost::asio::ssl::detail::engine::*)(void*, unsigned long), void*, unsigned long, boost::system::error_code&, unsigned long*) (engine.ipp:219) ==29870== by 0x7E3326: boost::asio::ssl::detail::io_op<boost::asio::basic_stream_socket<boost::asio ::ip::tcp, boost::asio::stream_socket_service<boost::asio::ip::tcp> >, boost::asio::ssl::detail::read_op<boost::asio::mutable_buffers_1>, boost::asio::detail::rea d_until_expr_op<boost::asio::ssl::stream<boost::asio::basic_stream_socket<bo ost::asio::ip::tcp, boost::asio::stream_socket_service<boost::asio::ip::tcp> > >, std::allocator<char>, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > >, boost::_bi::bind_t<void, boost::_mfi::mf2<void, HttpConnection, boost::system::error_code const&, char const*>, boost::_bi::list3<boost::_bi::value<boost::shared_ptr<HttpConnection> >, boost::arg<1> (*)(), boost::_bi::value<char const*> > > > >::operator()(boost::system::error_code, unsigned long, int) (engine.ipp:151) ==29870== by 0x7E45C2: boost::asio::detail::wait_handler<boost::asio::ssl::detail::io_op<boost::asi o::basic_stream_socket<boost::asio::ip::tcp, boost::asio::stream_socket_service<boost::asio::ip::tcp> >, boost::asio::ssl::detail::read_op<boost::asio::mutable_buffers_1>, boost::asio::detail::read_until_expr_op<boost::asio::ssl::stream<boost::asio ::basic_stream_socket<boost::asio::ip::tcp, boost::asio::stream_socket_service<boost::asio::ip::tcp> > >, std::allocator<char>, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > >, boost::_bi::bind_t<void, boost::_mfi::mf2<void, HttpConnection, boost::system::error_code const&, char const*>, boost::_bi::list3<boost::_bi::value<boost::shared_ptr<HttpConnection> >, boost::arg<1> (*)(), boost::_bi::value<char const*> > > > > >::do_complete(boost::asio::detail::task_io_service*, boost::asio::detail::task_io_service_operation*, boost::system::error_code const&, unsigned long) (bind_handler.hpp:46) ==29870== by 0x619783: boost::asio::detail::task_io_service::run(boost::system::error_code&) (task_io_service_operation.hpp:37) ==29870== by 0x79F4FE: BayeuxServer::run() (io_service.ipp:59) ==29870== by 0x364800DB8E: ??? (in /usr/lib64/libboost_thread.so.1.49.0) ==29870== by 0x3AE300677C: start_thread (in /lib64/libpthread-2.5.so) ==29870== by 0x3AE20D325C: clone (in /lib64/libc-2.5.so) ==29870== Address 0x1b508610 is 0 bytes inside a block of size 1,024 free'd ==29870== at 0x4A05A33: operator delete(void*) (vg_replace_malloc.c:346) ==29870== by 0x658E54: std::vector<char, std::allocator<char> >::_M_fill_insert(__gnu_cxx::__normal_iterator<char*, std::vector<char, std::allocator<char> > >, unsigned long, char const&) (new_allocator.h:95) ==29870== by 0x68B805: boost::asio::basic_streambuf<std::allocator<char> >::reserve(unsigned long) (stl_vector.h:851) ==29870== by 0x7E2EF2: boost::asio::detail::read_until_expr_op<boost::asio::ssl::stream<boost::asio ::basic_stream_socket<boost::asio::ip::tcp, boost::asio::stream_socket_service<boost::asio::ip::tcp> > >, std::allocator<char>, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > >, boost::_bi::bind_t<void, boost::_mfi::mf2<void, HttpConnection, boost::system::error_code const&, char const*>, boost::_bi::list3<boost::_bi::value<boost::shared_ptr<HttpConnection> >, boost::arg<1> (*)(), boost::_bi::value<char const*> > > >::operator()(boost::system::error_code const&, unsigned long, int) (basic_streambuf.hpp:207) ==29870== by 0x7E32DE: boost::asio::ssl::detail::io_op<boost::asio::basic_stream_socket<boost::asio ::ip::tcp, boost::asio::stream_socket_service<boost::asio::ip::tcp> >, boost::asio::ssl::detail::read_op<boost::asio::mutable_buffers_1>, boost::asio::detail::read_until_expr_op<boost::asio::ssl::stream<boost::asio ::basic_stream_socket<boost::asio::ip::tcp, boost::asio::stream_socket_service<boost::asio::ip::tcp> > >, std::allocator<char>, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > >, boost::_bi::bind_t<void, boost::_mfi::mf2<void, HttpConnection, boost::system::error_code const&, char const*>, boost::_bi::list3<boost::_bi::value<boost::shared_ptr<HttpConnection> >, boost::arg<1> (*)(), boost::_bi::value<char const*> > > > >::operator()(boost::system::error_code, unsigned long, int) (read_op.hpp:59) ==29870== by 0x7E493C: boost::asio::detail::reactive_socket_recv_op<boost::asio::mutable_buffers_1, boost::asio::ssl::detail::io_op<boost::asio::basic_stream_socket<boost::asio ::ip::tcp, boost::asio::stream_socket_service<boost::asio::ip::tcp> >, boost::asio::ssl::detail::read_op<boost::asio::mutable_buffers_1>, boost::asio::detail::read_until_expr_op<boost::asio::ssl::stream<boost::asio ::basic_stream_socket<boost::asio::ip::tcp, boost::asio::stream_socket_service<boost::asio::ip::tcp> > >, std::allocator<char>, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > >, boost::_bi::bind_t<void, boost::_mfi::mf2<void, HttpConnection, boost::system::error_code const&, char const*>, boost::_bi::list3<boost::_bi::value<boost::shared_ptr<HttpConnection> >, boost::arg<1> (*)(), boost::_bi::value<char const*> > > > > >::do_complete(boost::asio::detail::task_io_service*, boost::asio::detail::task_io_service_operation*, boost::system::error_code const&, unsigned long) (bind_handler.hpp:118) ==29870== by 0x613778: boost::asio::detail::epoll_reactor::descriptor_state::do_complete(boost::asi o::detail::task_io_service*, boost::asio::detail::task_io_service_operation*, boost::system::error_code const&, unsigned long) (task_io_service_operation.hpp:37) ==29870== by 0x619783: boost::asio::detail::task_io_service::run(boost::system::error_code&) (task_io_service_operation.hpp:37) ==29870== by 0x79F4FE: BayeuxServer::run() (io_service.ipp:59) ==29870== by 0x364800DB8E: ??? (in /usr/lib64/libboost_thread.so.1.49.0) ==29870== by 0x3AE300677C: start_thread (in /lib64/libpthread-2.5.so) ==29870== by 0x3AE20D325C: clone (in /lib64/libc-2.5.so)

1 0

Boost.Process 0.5 released
by Boris Schaeling 27 Aug '12

27 Aug '12

Boost.Process 0.5 has been released! Documentation and library can be found at: <http://www.highscore.de/boost/process0.5/> Why version 0.5? We have no official process management library in Boost yet. But we had countless drafts since 2006. In order not to get completely lost between all those drafts I gave them version numbers. For example, 0.1 was the very first draft in 2006. 0.31 is the draft documented at <http://www.highscore.de/cpp/process/>. And 0.4 was the draft which was rejected in a formal review in spring 2011. Version 0.5 is the latest attempt to create a process management library. What's new? Version 0.5 has a very different design than all other drafts before. The design was proposed by Jeff Flinn who introduced it briefly at BoostCon 2011 when we gave a presentation on Boost.Process. The goals of this new version are: * A lightweight implementation (no overly complicated code anymore). * A fully extensible library (no need for arbitrary "extension points"). * Support for system-level cross-platform concepts only (no Java-like constructs). Other goals like making the library easy to use were of course also targeted. But these were goals before. The goals above are important as those things have been criticized in earlier drafts and were as far as I see the main reasons why version 0.4 had been rejected. What now? Please have a look at the documentation and think about whether this library could fit into Boost. If you have more time, please download and play around with the new version. Concentrate on the bigger picture: Does the design make sense? Is it easy to understand? Can you do what you want to do? Does it look and feel like modern C++? This is not a formal review. But it would help to get some feedback to see a tendency. After six years, two Google Summer of Code projects, countless drafts and still a lot of interest from many people in a process management library in Boost, we might be able to get somewhere this time. (If you are on Windows: There is a bug in Boost.Iostreams which has been fixed in the release candidate 1 of Boost 1.51.0. See the documentation if you run into a problem when closing pipes.) Anything missing? We had a closed beta test in the previous two weeks with about ten people who had contacted me because of their interest in Boost.Process. I got already some good feedback and have to change a few details. These are only details though and shouldn't stop anyone from using this new version now. If everyone is happy with the overall design and architecture and it's only details we have to discuss, it's already a huge step forward. Last but not least: Boost.Process 0.5 wouldn't exist without the help of Intra2net AG (<http://www.intra2net.com/>). Intra2net AG sponsored a project to support the development of Boost.Process. It was this sponsorship which made it possible to create this new version. Thanks to them! Boris

9 46

[rational] Over/underflow checked update ready for comments
by Dan Searles 26 Aug '12

26 Aug '12

>> Quick overview of the update: the original usage and >> code of the template class remains mostly as before, ie >> boost::rational<int> RationalVar; should work as >> in the original version. Adding 'true' as a second >> template parameter, however, allows an exception to >> be raised if an operation with a class variable results >> in a value that can't accurately be represented >I haven't looked at anything else, but I dislike using true/false as the template argument. I'd prefer meaningful names. _____ >Rob Stewart robert.stewart(a)sig.com >Software Engineer using std::disclaimer; >Dev Tools & Components >Susquehanna International Group, LLP http://www.sig.com Replaced the bool template parameter with an enum: enum RationalChecktype {RationalNoChecking, RationalCheckforOverflow};

1 0

[contract] "Duplicated globals" problem
by Andrzej Krzemienski 26 Aug '12

26 Aug '12

Hi Lorenzo, I think that installing broken contract handlers will not work correctly when a program is linking with DLLs on Windows. I do not know how dynamic linking works on Linux. Here is a sketch of a sample program that fails to work as expected. The program consists of two projects: one builds a DLL, the other builds an executable that links with the DLL: DLL: ==================================== # include <contract.hpp> # include "testdll.h" CONTRACT_FUNCTION ( int (give1) () postcondition(auto result = return, result == 1) ) { return 2; // obviously wrong } DLL_MODE void test_contracts() { int i = give1(); } ==================================== EXE: ==================================== # include <contract.hpp> # include "../testdll/testdll.h" void PostconditionHandler( contract::from const& context ) { std::cerr << "ERROR!!!" << std::endl; } int main() { contract::set_postcondition_broken(&PostconditionHandler); test_contracts(); } ==================================== I expect that the program would use the broken postcondition handler I installed, but it still uses the default one. The cause is that Contract is using some global object to store the installed handler, but when a program is linked with a DLL both the program and the DLL store their own copy of the global. The global is duplicated. THe one in the EXE registers my callback, but the global in the DLL still keeps the default handler. The only way to solve this issue is to -- at least in the cases when users' programs link with DLLs -- ship Contract with a small DLL that stores the unique copy of the globals. This is how Boost.Log handles this problem. Quoting the documentation ( http://boost-log.sourceforge.net/libs/log/doc/html/log/installation.html) *"If your application consists of more than one module (e.g. an exe and one or several dll's) that use Boost.Log, the library must be built as a shared object. If you have a single executable, you may build the library as a static library." * Regards, &rzej* *

2 1

complex acos and atanh
by Stephen Montgomery-Smith 26 Aug '12

26 Aug '12

Hey people, I was trying out your complex arc-trig functions, to compare them with some that I have written: http://people.freebsd.org/~stephen/catrig.c. I am finding that your acos function is getting some of them wrong. For example, your program evaluates acos(1.00000002785941 + I*5.72464869028403e-200) as 0 - I*0.000236048349018331 whereas it should be 2.42520172707401e-196 - I*0.000236048349018331. Also your complex atanh function does quite badly in some edge cases. I presume it is because Hull, Fairgrieve and Tang did not write a paper on that function. If you email Fairgrieve, he may give you a copy of the paper that was never published. He sent me a copy, but doesn't want it widely distributed. I devised my own algorithm at http://people.freebsd.org/~stephen/catrig.c, which actually turns out to be very similar to the unpublished algorithm of Hull, Fairgrieve and Tang. Anyway, I would be happy to work with whomever is currently working on these algorithms. I think someone must be, because I see that someone fixed a bug in asin since 1.48.0. Best regards, Stephen

2 2