- Boost - lists.boost.org

[lockfree/win32] looking for help
by Tim Blechmann 08 Jul '12

08 Jul '12

hi all, i've been trying to sort out some compile problems of boost.lockfree, but now i find myself in a position that for some reasons, the tests on on the windows platform fail ... not only with msvc, but also with gcc/mingw. the main problem that i have is that two test cases compile, but fail at runtime ... unfortunately the error messages are not necessarily expressive: mingw: -- Run [2012-07-04 16:08:25 UTC]: fail EXIT STATUS: -1073741819 -- msvc: -- Running 8 test cases... unknown location(0): fatal error in "spsc_queue_test_buffering": stack overflow ..\libs\lockfree\test\spsc_queue_test.cpp(435): last checkpoint: "spsc_queue_test_buffering" entry. *** 1 failure detected in test suite "Master Test Suite" EXIT STATUS: 201 -- i am pretty clueless, what is happening there ... would it be possible that someone who has a win32 machine can send me a backtrace of this crash? thanks, tim

2 3

[Lockfree] examples with clang
by Rich E 08 Jul '12

08 Jul '12

Hi all, I've been attempting to test Lockfree (while at the same time learning to work with boost from trunk, so bare with me I'm a bit new) using Clang 4.0, the version that will ship with Xcode 4.5. It appears that the preprocessor check is only for gcc. If I change: #if __cplusplus < 201103L to #if __cplusplus < 201103L && ( defined( __clang__ ) && ! __has_include(<atomic>) ) and compile the examples with (from libs/lockfree/examples): ../../../bjam toolset=clang cxxflags="-std=c++0x -stdlib=libc++" linkflags="-stdlib=libc++" Then the examples work correctly (queue output: "produced 40000000 objects. consumed 40000000 objects."). I'm sure there is a more elegant way to check for libc++ existence when compiling with clang, possibly with boost/config.hpp? Also, I think a message in the dummy main would be nice to inform the user that lockfree won't work with their current setup / compile method. Cheers, Rich

2 1

[context] a couple of build fixes for Solaris (10)
by Luke Elliott 07 Jul '12

07 Jul '12

First is to get the definition of mprotect() that takes void * rather than char * by defining _XPG4_2, which causes a compilation failure with Solaris Studio 12.3. Second is a workaround for a bug in their csignal header which uses the same include guard that signal.h uses, which csignal includes (leaving SIGSTKSZ undefined). Attached if of any help. Regards Luke Elliott.

3 3

memory leak in fileiter.cpp
by GAURAV GUPTA 07 Jul '12

07 Jul '12

There is a possible memory leak in fileiter.cpp in Regex component. I have raised a ticket https://svn.boost.org/trac/boost/ticket/7084 along with the patch. Please verify and commit it. -- Regards, Gaurav

1 0

[gil::io_new] Help with gcc compiler
by Christian Henning 07 Jul '12

07 Jul '12

Hi there, I'm in the process of making my gil extension ready for inclusion. The code is now stable and I'm trying to get gcc/clang to work. Since I'm no expert with gcc I would like to ask for some help with various compiler errors. Could someone get the latest from here: http://gil-contributions.googlecode.com/svn/trunk/gil_2 and try to compile the following code: ------------------- #include <boost/gil/extension/io_new/bmp_all.hpp> using namespace std; using namespace boost::gil; int main() { { const char* bmp_filename = ""; const char* bmp_out = ""; rgb8_image_t img; read_image( bmp_filename, img, bmp_tag() ); write_view( bmp_out + "all_formats_test.bmp", view( img ), bmp_tag() ); } } ----------------- It should generate some errors which I don't quite understand. I have posted the errors here: http://pastebin.com/Wps3YH1N I'm using gcc 4.4.1 . The complete test suite compiles with Visual Studio 2010 without any warnings. Thanks a lot! Christian

5 10

noncopyable && C++11
by Ben Pope 06 Jul '12

06 Jul '12

Hi all, boost::noncopyable interferes with the implicit generation of move construction and move assignment, so I was wondering if anybody had any thoughts on something like this: class noncopyable { protected: noncopyable() = default; noncopyable(noncopyable&&) = default; noncopyable& operator=(noncopyable&&) = default; noncopyable(noncopyable const&) = delete; noncopyable& operator=(noncopyable const&) = delete; }; Which would replace the existing technique when C++11 support is on. I guess it could be argued that non copyable should imply non moveable. However, you won't hear me defending that position; plenty of thought went into move semantics with the defaults chosen to have sa[f|n]e behaviour. The derived class will still be subject to that criteria, I'm just trying to prevent noncopyable from artificially inhibiting move semantics. Would anybody want nonmovable? nonmoveable? Any thoughts / comments / omissions gratefully received. Ben

2 2

[iostreams] cin wrapper
by Daniel Larimer 06 Jul '12

06 Jul '12

I am attempting to 'wrap' std::cin such that all 'blocking' occurs in a dedicated 'cin' thread. This is required because in a cooperatively scheduled thread, no fiber can call cin and block. So I went about attempting a boost::iostreams::source: class cin_source : public boost::iostreams::source { streamsize read( char* s, streamsize n ) { if( n == 0 ) return n; auto r = std::cin.readsome(s,n); if( r <= 0 ) { // post read request to another thread. // cooperatively wait for it to finish... return n; } return r; } }; The problem I have is that this code causes read to be called with 'n == 0'; boost::iostreams::stream<cin_source> my_cin; char buf[100]; my_cin.read( buf, 100 ); It doesn't matter how I attempt to read from the stream, n always equals 0. What am I doing wrong? Dan

1 0

[format] Any interest in decoupling it from std::basic_string?
by news.gmane.org 06 Jul '12

06 Jul '12

I.e. such that alternative string implementations could be used (GCC's vstring, boost's...). I've attempted two methods, both pretty simple. First is to simply template basic_format on the string type and propagate that change throughout the library (including similar for the altstream/stringbuf classes). Second was to add a template template parameter for the string, e.g.: template <class Ch, class Tr, class Alloc, template <typename, typename, typename> class StringT = std::basic_string> class basic_format; and propagate that. I would assume the first method is definitely out due to breaking the interface (especially the altstream classes which I would guess are supposed to look like the std ones). I've not looked at whether something such as templatizing the str() method on the string implementation is possible, as either of the above is sufficient for my own purposes and trivial which I suspect a templatized str() wouldn't be... Anyway, if there's any interest I can post a patch for either or both and if not continue using the first method (simpler diff from the original)! Best regards Luke Elliott.

5 11

[Range] Range adaptor approach for temporary range lifetime issue
by Michel Morin 06 Jul '12

06 Jul '12

When an rvalue range is adapted in range-based for loop, the lifetime of the rvalue range ends before the loop body begins. For example, the following code produces dangling references. for (auto x : std::string("Hello world!") | reversed) { /*...*/ } This is the temporary range lifetime issue. When I proposed BOOST_FOREACH extension for this issue, ( http://lists.boost.org/Archives/boost/2011/04/180591.php ) a solution with "moving an rvalue range into range adaptors" was suggested in the thread. I implemented this approach as a range adaptor (with the aid of rvalue references and decltype C++11 features): for (auto x : std::string("Hello world!") | moved | reversed) { /*...*/ } When an rvalue range is piped to `moved` adaptor, it returns `moved_range`. `moved_range` stores the following data: * Moved rvalue range: The rvalue range is moved and stored in `moved_range`. * Applied adaptors: When range adaptors are applied to`moved_range`, they are not invoked immediately; they are stored in fusion vector. The stored adaptors are invoked when `begin` or `end` member functions of `moved_range` are called. (To prevent repeated applications of the stored range adaptors, begin and end iterators are cached when they are first computed.) For example, in `std::string("Hello world!") | moved | reversed`, 1. By applying `moved` adaptor, `moved_range` is created. It stores `std::string` object moved from `std::string("Hello world!")`. It doesn't store any adaptors. 2. By applying `reversed` adaptor, another `moved_range` is created. It stores `std::string` object moved from the previous `moved_range`. It also stores `reversed` adaptor in fusion vector. Step 2 is repeated if more adaptors are applied. Then, in "for-init-statement" of range-based for loop, the stored adaptors are actually applied (and the result is cached). The free function version (`boost::adaptors::move`) supports `std::initializer_list`: for (auto x : move( {1, 2, 3, 4, 5} ) | reversed) { /*...*/ } Note that we cannot support it in the pipe syntax `{1, 2, 3, 4, 5} | moved` because the C++11 Standard does not allow to use list initializers in binary operators. Code attached. Comments? Regards, Michel

7 40

[exception] [type_traits] warning: unrecognized #pragma GCC system_header
by Krzysztof Czainski 06 Jul '12

06 Jul '12

Hello, Below is a list of files, that contain: #if defined(__GNUC__) #pragma GCC system_header #endif which causes a warning on my TI compiler. E:\boost\include\boost-1_50_0\boost\exception\all.hpp E:\boost\include\boost-1_50_0\boost\exception\current_exception_cast.hpp E:\boost\include\boost-1_50_0\boost\exception\detail\error_info_impl.hpp E:\boost\include\boost-1_50_0\boost\exception\detail\exception_ptr.hpp E:\boost\include\boost-1_50_0\boost\exception\detail\is_output_streamable.hpp E:\boost\include\boost-1_50_0\boost\exception\detail\object_hex_dump.hpp E:\boost\include\boost-1_50_0\boost\exception\detail\type_info.hpp E:\boost\include\boost-1_50_0\boost\exception\diagnostic_information.hpp E:\boost\include\boost-1_50_0\boost\exception\errinfo_errno.hpp E:\boost\include\boost-1_50_0\boost\exception\exception.hpp E:\boost\include\boost-1_50_0\boost\exception\get_error_info.hpp E:\boost\include\boost-1_50_0\boost\exception\info.hpp E:\boost\include\boost-1_50_0\boost\exception\info_tuple.hpp E:\boost\include\boost-1_50_0\boost\exception\to_string.hpp E:\boost\include\boost-1_50_0\boost\exception\to_string_stub.hpp E:\boost\include\boost-1_50_0\boost\throw_exception.hpp E:\boost\include\boost-1_50_0\boost\type_traits\is_virtual_base_of.hpp In other files, i.e. boost\integer.hpp, I've found this #pragma guarded against __GNUC__ >= 4: #if defined(__GNUC__) && (__GNUC__ >= 4) #pragma GCC system_header #endif which doesn't cause a warning on my compiler, because it defines __GNUC__ to 3. Could this be fixed in the files on the list above, or are there some __GNUC__ 3 compilers, that need this pragma to be enabled? Regards, Kris

3 5