- Boost-users - lists.preview.boost.org

[Coroutine 1.54] Compile error with bool return from coroutine-function
by Rob Desbois 27 Apr '14

27 Apr '14

(Caveat: I've been unable to find anything about this; hopefully haven't missed anything obvious.) (Apology: I'm aware Coroutine has a new interface as of 1.55; I haven't determined whether it is also affected.) Given the following code: #include <cassert> #include <boost/bind.hpp> #include <boost/coroutine/coroutine.hpp> using coro_t = boost::coroutines::coroutine< bool() >; void get_a_bool( coro_t::caller_type& yield ) { while (true) yield(true); } int main() { coro_t producer( boost::bind(&get_a_bool, _1) ); assert( producer ); } I get this error on compilation (g++ 4.8.2, Boost 1.54): rob{code 0} 839# clear && g++ boost-coroutine-bool-return.cpp -g -W -Wall --std=c++11 -L. -lboost_context -lboost_coroutine -o boost-coroutine-bool-return In file included from /usr/include/boost/coroutine/v1/detail/coroutine_base_resume.hpp:26:0, from /usr/include/boost/coroutine/v1/detail/coroutine_base.hpp:21, from /usr/include/boost/coroutine/v1/coroutine.hpp:30, from /usr/include/boost/coroutine/coroutine.hpp:13, from boost-coroutine-bool-return.cpp:11: /usr/include/boost/coroutine/detail/holder.hpp: In instantiation of ‘struct boost::coroutines::detail::holder<bool>’: /usr/include/boost/coroutine/v1/detail/coroutine_base_resume.hpp:103:28: required from ‘void boost::coroutines::detail::coroutine_base_resume<Signature, D, void, 1>::resume(boost::coroutines::detail::coroutine_base_resume<Signature, D, void, 1>::arg_type) [with Signature = void(bool); D = boost::coroutines::detail::coroutine_base<void(bool)>; boost::coroutines::detail::coroutine_base_resume<Signature, D, void, 1>::arg_type = bool]’ /usr/include/boost/coroutine/v1/detail/coroutine_op.hpp:268:9: required from ‘D& boost::coroutines::detail::coroutine_op<Signature, D, void, 1>::operator()(boost::coroutines::detail::coroutine_op<Signature, D, void, 1>::arg_type) [with Signature = void(bool); D = boost::coroutines::coroutine<void(bool), 1>; boost::coroutines::detail::coroutine_op<Signature, D, void, 1>::arg_type = bool]’ boost-coroutine-bool-return.cpp:29:13: required from here /usr/include/boost/coroutine/detail/holder.hpp:39:14: error: ‘boost::coroutines::detail::holder<Data>::holder(boost::coroutines::detail::coroutine_context*, bool) [with Data = bool]’ cannot be overloaded explicit holder( coroutine_context * ctx_, bool force_unwind_) : ^ /usr/include/boost/coroutine/detail/holder.hpp:35:14: error: with ‘boost::coroutines::detail::holder<Data>::holder(boost::coroutines::detail::coroutine_context*, Data) [with Data = bool]’ explicit holder( coroutine_context * ctx_, Data data_) : At a glance it looks like there's no way a coroutine-function could possibly return bool due to the overload collision. Is this right? Did I miss something completely obvious in the docs? It's easy enough to work around with a transparent bool wrapper, I just want to be sure I'm not being a muppet :-D TIA -- rob -- Rob Desbois http://theotherbranch.wordpress.com/

2 1

Boost PropertyTree: read_json() followed by write_xml produces invalid XML
by Marcel Loose 27 Apr '14

27 Apr '14

Hi all, I was quite surprised and disappointed to find out that I was unable to convert a simple JSON file into XML, using the Boost PropertyTree. As input I used the example JSON file from the documentation (see example.json below). I wrote a small program (see example.cc below) that reads in this JSON file, writes it to disk as XML file, and finally tries to read back this XML file. This last operation fails with an exception (see output below); the XML file that is produced (see example.xml below) is invalid, because it contains empty tags (<>). Am I doing something wrong? Best regards, Marcel Loose. $ cat example.json { "menu": { "foo": true, "bar": "true", "value": 102.3E+06, "popup": [ {"value": "New", "onclick": "CreateNewDoc()"}, {"value": "Open", "onclick": "OpenDoc()"} ] } } $ cat example.cc #include <boost/property_tree/ptree.hpp> #include <boost/property_tree/json_parser.hpp> #include <boost/property_tree/xml_parser.hpp> #include <fstream> using namespace std; using namespace boost::property_tree; int main() { ptree pt; { read_json("example.json", pt); ofstream ofs("example.xml"); write_xml(ofs, pt, xml_writer_make_settings(' ', 2)); } { ifstream ifs("example.xml"); read_xml(ifs, pt); } return 0; } $ ./example terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<boost::property_tree::xml_parser::xml_parser_error> >' what(): <unspecified file>(7): expected element name Aborted $ cat example.xml <?xml version="1.0" encoding="utf-8"?> <menu> <foo>true</foo> <bar>true</bar> <value>102.3E+06</value> <popup> <> <value>New</value> <onclick>CreateNewDoc()</onclick> </> <> <value>Open</value> <onclick>OpenDoc()</onclick> </> </popup> </menu>

3 3

Library advice
by Mathew Benson 27 Apr '14

27 Apr '14

Is there a boost library that I can use to dynamically deserialize structures from untyped buffers? I need to write code that receives a buffer of data from another computer, with some data serialized with bitfield structs. Yes, I do understand the difficulty of doing this and that it would have been easier if the data was serialized properly before transmission. I have no control over the transmitted data, but I can get the data format defined accurately. I fully understand the numerous edge cases and pitfalls with doing this and don't take the task lightly. Therefore, I'm looking for libs and open source code to make it easier. The incoming data format can and frequently does change so I can't hard code the format, but the format is defined in an XML file. I wrote an internal database containing field type, offset, and bit lengths, and populated via the XML file. My XML file is auto generated to match the as built interface definition, bit for bit. I wrote code that deserializes most of the data, but I need to rewrite it as a general purpose function to work for all cases, including bitfield structs. For now, I can assume same CPU (little endian) and the same compiler (gcc). However, both source and destination CPUs and OS's might be 32 bit or 64 bit. In future versions, I would like to add support for endian conversion and differing compiler packing. Does boost have something that would help? Sent from my Packard Bell 386

3 3

[Range] Extending the library for a base class and all its derived - enable_if
by Albert Yiamakis 26 Apr '14

26 Apr '14

Hello, I am looking to make a class, whose definition I cannot touch, model a Range concept. To that end I use method 2, http://www.boost.org/doc/libs/1_55_0/libs/range/doc/html/range/reference/ex… However, I am unable to specialize the relevant metafunctions (range_const_iterator, range_mutable_iterator, range_iterator) since they do not have an additional template parameter that would permit me to employ enable_if. I have found this impossible without touching either the class' definition or the metafunctions. Perhaps there is some other trick I am not aware of, in which case I would greatly appreciate hearing about it. If such a trick does not exist, then should these metafunctions feature an additional template parameter? Given that this is one of the methods to extend the library, it would make sense that one wishes to do so on conditions that enable_if can check. Albert

4 9

[Interprocess] Named objects with 'Managed External Buffer' (newbie)
by Jorge Bellon 25 Apr '14

25 Apr '14

Good afternoon, First, I apologize if I'm asking something obvious or outside the acceptable topics. I'm starting to use Boost so I'm quite newbie. I'm interested in using the managed_external_buffer in order to perform buffer copies into a created buffer that was created previously. Therefore, I will be storing mostly raw data into that buffer. Basically, I'm trying to do the following: Suppose a program uses the following int array: int numbers[10] = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 }; Eventually, I have to copy a chunk of that array into a pre-allocated buffer (i.e. the managed_external_buffer). Let the chunk be defined with: typedef struct { void* addr; size_t size; } chunk_t; However, these chunks can overlap, so I will sometimes merge two overlapping chunks into a bigger one. Example: // Chunk 1 chunk1_addr = (void*)&numbers[3]; chunk1_size = 3 * sizeof(int); // Chunk 2 chunk2_addr = (void*) &numbers[4]; chunk2_size = 4 * sizeof(int); The desired result is a single chunk containing {3, 4, 5, 6, 7}. The proposed problem is the merge, but how to find those chunks in the managed_external_buffer, so my question is the following: - Would it be possible to use named objects with another data type other than CharType and use managed_external_buffer::find to look for them? The thing is I could generate a hash or a numerical ID to identify the chunk. Generating a string using that number could be done but it would not be efficient. Otherwise, I guess I would have to keep a map (or other collection) with the <chunk ID, chunk pointer> pairs, define the managed_external_buffer with boost::interprocess::null_index (since I won't be using named objects) and use it through allocate and deallocate. Is this correct? Thanks in advance, Jorge WARNING / LEGAL TEXT: This message is intended only for the use of the individual or entity to which it is addressed and may contain information which is privileged, confidential, proprietary, or exempt from disclosure under applicable law. If you are not the intended recipient or the person responsible for delivering the message to the intended recipient, you are strictly prohibited from disclosing, distributing, copying, or in any way using this message. If you have received this communication in error, please notify the sender and destroy and delete any copies you may have received. http://www.bsc.es/disclaimer

2 1

[msm] compilation warning with boost-1.55.0
by samriti katoch 25 Apr '14

25 Apr '14

Hi, With boost-1.55.0 I see this warning from msm code: In file included from /build/boost/fusion/include/make_map.hpp:10, from /build/boost/msm/front/euml/common.hpp:40, from /build/boost/msm/front/euml/operator.hpp:15, from /build/abc.cc:15: cc1plus: warnings being treated as errors /build/boost/msm/back/fold_to_list.hpp:35: warning: 'boost::msm::back::states_' defined but not used What is a good way to remove this warning? Thanks! Samriti

2 2

Free Workshop: Test-Driven Development in C++ (Salt Lake City, UT)
by legalize+jeeves＠mail.xmission.com 25 Apr '14

25 Apr '14

[Please do not mail me a copy of your followup] This will be a practice run of my presentation for C++ Now! 2014. Workshop announcement: <http://bit.ly/RQgOtQ> Excerpt from the announcement: Test-Driven Development in C++, Wednesday, April 30, 2014 Have you heard about that new thing the kids are doing called test-driven development and wondered what it's all about? Is it about testing? Is it about design? Is it good or is it bad? Come join me for an evening workshop where I will show you what it's all about and you'll have a chance to try it out for yourself on a problem large enough to be interesting, but small enough that you can make progress on it during the workshop. The workshop will use the Boost.Test and Turtle libraries. Thanks to Fusion-io for the meeting facilities, WiFi access, food (pizza from The Pie) and refreshments. -- "The Direct3D Graphics Pipeline" free book <http://tinyurl.com/d3d-pipeline> The Computer Graphics Museum <http://computergraphicsmuseum.org> The Terminals Wiki <http://terminals.classiccmp.org> Legalize Adulthood! (my blog) <http://legalizeadulthood.wordpress.com>

1 0

circular buffer in a multithreaded program
by Sebastian Gesemann 24 Apr '14

24 Apr '14

Hello! I had some troubles with using boost::circular_buffer in a multithreaded application and I'm not really sure whether I actually ran into a bug or wether I expected just too much that wasn't promised in the thread-safety section. Anyhow, I wanted to share my story and a theory of what possibly went wrong. Maybe this posting even contains some ideas of how circular_buffer could be made a little more thread-safe without ading any mutex- or atomics-based synchronization. What I tried to do: I tried to grant one thread write access to a sub-range in a circular_buffer while another thread had read-access to a non-overlapping sub-range. At no time any capacity change was needed. Occasionally, in a reader thread I would "consume" some of the values that were already "committed" by a writer and make more room for the writer to store new data by - locking a mutex - calling cb.erase_first(n); - calling cb.resize(cb.capacity()); - notify writer that there is new free space to write stuff into - unlocking the mutex As for iterator invalidation, erase_first just invalidates iterators to the elements being removed. And resize in this case just increases the size again to the capacity which should not invalidate any iterators If I understand correctly. But the program did not work. The first thing that came to my attention was that the checked iterators of circular_buffer don't support this kind of multithreading as they register and unregister themselves into a linked list without any synchronization. But even after disabling the checked iterators via BOOST_CB_DISABLE_DEBUG, the program still did not work. I could not figure out why (with enough certainty) and eventually replaced circular_buffer with a custom vector-based data structure. My guess is that the iterators of a circular_buffer somehow depend on some part of the state of the buffer that is modifed by erase_first. Then, I would have a data race. But such a dependency does not really have to be there. In my own implementation, an iterator is "self-sufficient" in the sense that it does not need to access any of the buffer's state (indices, counters and whatnot) to access the right elements. And this way, erasing some unneeded elements from the buffer while some other thread still works on other values in the buffer via iterators does not cause a data race. In my opinion, it's reasonable to expect something like this to work using a circular_buffer. Maybe someone knows more about how circular_buffer iterators are implemented and/or wants to improve the circular_buffer implementation w.r.t. thread-safety to support this kind of use. Since I switched to a custom data structure that is able to deal with this situation, I did not investiage this issue with circular_buffer any further. I already lost about a day because of this. I hope, this is of interest to somebody who wants to avoid exessive copying in the standard "producer-consumer / bounded-buffer pattern". Cheers! Sebastian

5 10

Memory leak in boost::filesystems::parent_path()?
by B.W.H. van Beest 24 Apr '14

24 Apr '14

Hi, The following minimal program (only 7 lines) causes valgrind to report 2 errors, in line6: Some somebody point me to why it is wrong? a.cpp: 1 #include <boost/filesystem.hpp> 2 3 int main( int, char** ) 4 { 5 boost::filesystem::path path0( "/a/b/c" ); 6 boost::filesystem::path path1 = path0.parent_path(); 7 } Compiling and running with valgrind: g++ -g -lboost_system -lboost_filesystem a.cpp -o a.exe valgrind ./a.exe This gives the following output. Note the errors. ==13303== Memcheck, a memory error detector ==13303== Copyright (C) 2002-2012, and GNU GPL'd, by Julian Seward et al. ==13303== Using Valgrind-3.8.1 and LibVEX; rerun with -h for copyright info ==13303== Command: ./a.exe ==13303== ==13303== Invalid read of size 8 ==13303== at 0x5B0A5C8: wcscmp (in /lib64/libc-2.18.so) ==13303== by 0x52CA5C4: std::moneypunct<wchar_t, false>::~moneypunct() (in /usr/lib64/libstdc++.so.6.0.18) ==13303== by 0x52CA638: std::moneypunct<wchar_t, false>::~moneypunct() (in /usr/lib64/libstdc++.so.6.0.18) ==13303== by 0x52C1304: std::locale::_Impl::~_Impl() (in /usr/lib64/libstdc++.so.6.0.18) ==13303== by 0x52C143C: std::locale::~locale() (in /usr/lib64/libstdc++.so.6.0.18) ==13303== by 0x5AA775E: __cxa_finalize (in /lib64/libc-2.18.so) ==13303== by 0x503DD72: ??? (in /usr/lib64/libboost_filesystem.so.1.53.0) ==13303== by 0x400EE69: _dl_fini (in /lib64/ld-2.18.so) ==13303== by 0x5AA73D8: __run_exit_handlers (in /lib64/libc-2.18.so) ==13303== by 0x5AA7424: exit (in /lib64/libc-2.18.so) ==13303== by 0x5A90BEB: (below main) (in /lib64/libc-2.18.so) ==13303== Address 0x604c038 is 0 bytes after a block of size 8 alloc'd ==13303== at 0x4C284B7: operator new[](unsigned long) (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==13303== by 0x52CBC81: std::moneypunct<wchar_t, false>::_M_initialize_moneypunct(__locale_struct*, char const*) (in /usr/lib64/libstdc++.so.6 .0.18) ==13303== by 0x52C3807: std::locale::_Impl::_Impl(char const*, unsigned long) (in /usr/lib64/libstdc++.so.6.0.18) ==13303== by 0x52C42D1: std::locale::locale(char const*) (in /usr/lib64/libstdc++.so.6.0.18) ==13303== by 0x50451B1: boost::filesystem::path::codecvt() (in /usr/lib64/libboost_filesystem.so.1.53.0) ==13303== by 0x5046C7E: boost::filesystem::path::parent_path() const (in /usr/lib64/libboost_filesystem.so.1.53.0) ==13303== by 0x401843: main (a.cpp:6) ==13303== ==13303== Invalid read of size 8 ==13303== at 0x5B0A5C8: wcscmp (in /lib64/libc-2.18.so) ==13303== by 0x52CA4D4: std::moneypunct<wchar_t, true>::~moneypunct() (in /usr/lib64/libstdc++.so.6.0.18) ==13303== by 0x52CA548: std::moneypunct<wchar_t, true>::~moneypunct() (in /usr/lib64/libstdc++.so.6.0.18) ==13303== by 0x52C1304: std::locale::_Impl::~_Impl() (in /usr/lib64/libstdc++.so.6.0.18) ==13303== by 0x52C143C: std::locale::~locale() (in /usr/lib64/libstdc++.so.6.0.18) ==13303== by 0x5AA775E: __cxa_finalize (in /lib64/libc-2.18.so) ==13303== by 0x503DD72: ??? (in /usr/lib64/libboost_filesystem.so.1.53.0) ==13303== by 0x400EE69: _dl_fini (in /lib64/ld-2.18.so) ==13303== by 0x5AA73D8: __run_exit_handlers (in /lib64/libc-2.18.so) ==13303== by 0x5AA7424: exit (in /lib64/libc-2.18.so) ==13303== by 0x5A90BEB: (below main) (in /lib64/libc-2.18.so) ==13303== Address 0x604c268 is 0 bytes after a block of size 8 alloc'd ==13303== at 0x4C284B7: operator new[](unsigned long) (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so) ==13303== by 0x52CB6B1: std::moneypunct<wchar_t, true>::_M_initialize_moneypunct(__locale_struct*, char const*) (in /usr/lib64/libstdc++.so.6. 0.18) ==13303== by 0x52C3854: std::locale::_Impl::_Impl(char const*, unsigned long) (in /usr/lib64/libstdc++.so.6.0.18) ==13303== by 0x52C42D1: std::locale::locale(char const*) (in /usr/lib64/libstdc++.so.6.0.18) ==13303== by 0x50451B1: boost::filesystem::path::codecvt() (in /usr/lib64/libboost_filesystem.so.1.53.0) ==13303== by 0x5046C7E: boost::filesystem::path::parent_path() const (in /usr/lib64/libboost_filesystem.so.1.53.0) ==13303== by 0x401843: main (a.cpp:6) ==13303== ==13303== ==13303== HEAP SUMMARY: ==13303== in use at exit: 0 bytes in 0 blocks ==13303== total heap usage: 483 allocs, 483 frees, 34,115 bytes allocated ==13303== ==13303== All heap blocks were freed -- no leaks are possible ==13303== ==13303== For counts of detected and suppressed errors, rerun with: -v ==13303== ERROR SUMMARY: 2 errors from 2 contexts (suppressed: 2 from 2)

3 4

cmake + CUDA 6 + clang + boost errors on OS X Mavericks
by John Owens 24 Apr '14

24 Apr '14

Greetings, I’m using boost (1.55) (installed via macports) on: - OS X Mavericks - using clang (5.1, the Xcode default version, LLVM 3.4) - with CUDA 6 - and cmake (2.8.12.2). It has a large number of compile errors, shown here: http://pastebin.com/Qg2V0UwM that begin with: /opt/local/include/boost/type_traits/is_abstract.hpp(72): error: identifier "__is_abstract" is undefined I’d be delighted to have any advice on how to do cmake + CUDA 6 + clang + boost on OS X! JDO

1 0