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

[Boost-announce] [boost] [review] Metaparse Review Result
by Christophe Henry 11 Jul '15

11 Jul '15

Hi all, First of all, the final result. Metaparse is formally unconditionally accepted into Boost. We had 11 reviews, all "Yes" votes. 2 "No" reviews based on formality were later reversed to a "Yes". The review was lively, with a general agreement that the library is very useful, the design of high quality and the documentation great. Here are some noticeable quotes which helped me take my decision: - "thank you Abel for submitting such an amazing library. The quality of the documentation and code is wonderful and it has been a pleasure to review Metaparse." (Michael Caisse) - (Design) "Of high quality. Based on the MPL paradigm, but I don't view this as a problem." (Peter Dimov) - "Part of the role of boost is to push compilers to the limit (and sometimes beyond). This library certainly has the capabilities to do so." (Roland Bock) Worth noticing is the review of Gordon Woodhull. I myself wrote a new front-end for MSM using Metaparse, and other use cases were very welcome and Gordon managed to produce one for his compile-time graph library by the end of the review, which I found impressive (for both the reviewer and the library). Thank you for investing time in this review. Another important factor for me for acceptance is how well the author is going to maintain his library. We have too many abandoned libraries. Abel went out of his way to oblige remarks and demands made by reviewers. Long before the reviews he also had obliged my many critics and demands. He also showed himself open for experiments. All this while keeping a very calm and professional tone. This is for me a very good sign of a library which will be well-maintained. Finally, there were some repeated important comments which I see as adressed to the rest of Boost that the library is targeted towards library writers. >From Peter Dimov: > - Are you knowledgeable about the problem domain? "I'm not sure if anybody is. :-)" It all depends on us, library and potential library writers to become knowledgeable about this domain and make this library move from potentially useful to very useful. We need to use it to write more cool libraries. I notice a recurring theme where Boost would be dead or on the way to doom etc etc. Some very popular Boost libraries are now part of the Standard, and so what? We need to make new ones. Metaparse is one of them. Let's build more, using it! Yes Reviews(11): Andrzej Krzemienski Bjorn Reese Gordon Woodhull Lee Clagett Louis Dionne Michael Caisse Paul A. Bristow Peter Dimov Roland Bock Edward Diener Niall Douglas I want to thank all the reviewers who invested their time in this review and those who provided some usefuls comments. I also want to thank Abel for presenting this great library. Christophe Henry Metaparse Review Manager

1 0

Slow decoding in thread
by Gonzalo Garramuno 10 Jul '15

10 Jul '15

I am writing a movie player using boost. I use a condition_variable to control a mutex in one thread that waits for packets of video to be ready for decoding, while another thread feeds those packets. My problem is that the decode thread may end up doing a lot of work when playing backwards and it is not ready to wait on the condition variable when the message comes from the other thread and so it hangs (or so I think). This is my wait loop. I apologize for the use of macros. #define SCOPED_LOCK(mutex) Mutex::scoped_lock lk_##mutex(mutex) #define CONDITION_WAIT( cond, mutex ) cond.wait( lk_##mutex ); void aviImage::wait_image() { mrv::PacketQueue::Mutex& vpm = _video_packets.mutex(); SCOPED_LOCK( vpm ); for(;;) { if ( stopped() || ! _video_packets.empty() ) break; CONDITION_WAIT( _video_packets.cond(), vpm ); } return; } When it returns it tries to decode one or more frames of video taken from _video_packets. The condition variable is notified whenever the program calls push_back() to add a packet. That is: class PacketQueue { .... inline void push_back( AVPacket& pkt ) { Mutex::scoped_lock lk( _mutex ); // assert( pkt.dts != MRV_NOPTS_VALUE ); if ( av_dup_packet(&pkt) < 0 ) { std::cerr << "Could not duplicate packet - not added" << std::endl; return; } _packets.push_back( pkt ); _bytes += pkt.size; _cond.notify_one(); } }; Can somebody suggest an alternative to this scheme? -- Gonzalo Garramuño

1 1

Query: metis_input_exception
by Mayank Bakshi 10 Jul '15

10 Jul '15

1 0

crc updating how?
by U.Mutlu 09 Jul '15

09 Jul '15

Hi, I need to calc the crc chunkwise (ie. calc crc of a block, then continue with the next block etc.). Is this (not) possible with the boost crc? In other words: how can the type "crc_32_type" be initialized with a user supplied uint32_t value? boost::crc_32_type crc; crc = 0x1234; // <-- this brings compile error crc.process_bytes(buf, len); error: no match for ‘operator=’ (operand types are ‘boost::crc_32_type {aka boost::crc_optimal<32ul, 79764919u, 4294967295u, 4294967295u, true, true>}’ and ‘const uint32_t {aka const unsigned int}’) Thx

2 3

[crc] bug in boost::crc
by U.Mutlu 09 Jul '15

09 Jul '15

/* boost-crc-bug-demo.cpp Written by U.Mutlu on 2015-07-09-Th Compile: g++ -O2 -Wall -std=gnu++11 boost-crc-bug-demo.cpp Run: ./a.out Output: crc("The quick brown fox jumps over the lazy dog")=414fa339 crc("The quick brown fox ")=88b075e2 crc("The quick brown fox jumps over the lazy dog")=4ac56d0f ERROR See also: http://www.boost.org/doc/libs/1_58_0/libs/crc/crc.html#usage Conclusion: It doesn't do what it should and also advertises; ie. there's a bug */ #include <boost/crc.hpp> #include <boost/cstdint.hpp> #include <iostream> int main() { uint32_t expected; { const std::string sTxt = "The quick brown fox jumps over the lazy dog"; boost::crc_optimal<32, 0x04C11DB7, 0xFFFFFFFF, 0xFFFFFFFF, true, true> crc; crc.process_bytes(sTxt.data(), sTxt.size()); std::cout << "crc(\"" << sTxt << "\")=" << std::hex << crc.checksum() << std::endl; // save the result for the following test expected = crc.checksum(); } // test: doing the same as above, but now in 2 steps, ie. chunk- or blockwise: std::string sTxt; uint32_t result = 0xFFFFFFFF; for (size_t i = 0; i <= 1; ++i) { sTxt += !i ? "The quick brown fox " : "jumps over the lazy dog"; // using the special ctor for initial remainder stored in "result" boost::crc_optimal<32, 0x04C11DB7, 0xFFFFFFFF, 0xFFFFFFFF, true, true> crc(result); crc.process_bytes(sTxt.data(), sTxt.size()); std::cout << "crc(\"" << sTxt << "\")=" << std::hex << crc.checksum() << std::endl; result = crc.checksum(); } // eval: const std::string sRes = result == expected ? "SUCCESS" : "ERROR"; std::cout << sRes << std::endl; return 0; }

3 4

[Interprocess] posix robust mutex and EOWNERDEAD
by Malko 07 Jul '15

07 Jul '15

Hi, I lurked a bit and found this: http://boost.2283326.n4.nabble.com/boost-interprocess-gt-1-45-robust-mutexe… I didn't feel like necroposting, and it's /dev anyway, so allow me to ask here about updates on the matter. Is there anything, even if only "experimental" or "beta", that you have that would allow me to handle the EWOENERDEAD result? I'd rather use your code before putting my hands on your library, since it would be easier to let my manager (and his manager) deal with a "boost beta" rather than my personal modification to boost. If there is none, I would really need to modify your files for my project to work, and I would appreciate if you could confirm a couple of things for me. In http://stackoverflow.com/a/1179766/4598277 it is mentioned to modify two files: 1. posix/thread_helpers.hpp to set the attribute (first time ever I was happy to get an exception :D) 2. posix/mutex.hpp to handle the EOWNERDEAD return value Can you please confirm if this would be enough? It would seem so to me, but I have only studied parts of your code. Thank you for the assistance. -- View this message in context: http://boost.2283326.n4.nabble.com/Interprocess-posix-robust-mutex-and-EOWN… Sent from the Boost - Users mailing list archive at Nabble.com.

2 3

iterator over boost::tuple
by Uthpal Urubail 06 Jul '15

06 Jul '15

Experts, What is the best way to iterator over boost::tuple and return value? Example: For(int i=0;i<boost::tuples::length<T>::value;i++) { Boost_tuple_example.get<i>(); } Regards, UJ

5 4

Boost python and python sets
by Will Hall 06 Jul '15

06 Jul '15

Hi, I'm struggling to find a way to support python sets with boost. The Python API is well defined, so am unable to switch to Python list type. The API needs to support Python sets for both input and output values. I have output values (from C++ to Python) working as follows with to_python namespace bp = boost::python; template<class T> struct to_python; template<class T> struct to_python<std::set<T> > { to_python() { bp::to_python_converter< std::set<T>, to_python< std::set<T> > >(); } static PyObject* convert(const std::set<T>& set) { bp::list values; for(typename std::set<T>::const_iterator iter=set.begin(); iter!=set.end(); ++iter) { values.append(bp::object(*iter)); } PyObject* result = PySet_New(values.ptr()); return result; } static const PyTypeObject* get_pytype() { return &PySet_Type; } }; BOOST_PYTHON_MODULE(helpers) { to_python< std::set<int> >(); bp:class_<MyClass>() // Define constructor with const std::set& arg // Define method that returns const std::set& arg } But I have no idea how to write an equivalent "from_python" to convert Python set to std::set. Any help would be very appreciated!

1 0

{Serialization] Deserialization of const references
by Mccall, Kurt E. (JSC-EG411) 06 Jul '15

06 Jul '15

As near as I can tell from the documentation http://www.boost.org/doc/libs/1_57_0/libs/serialization/doc/serialization.h… , the following is the way to handle const reference members, which are initialized in the constructor. Serialization seems to be working, from an inspection of the resulting XML file, but after deserialization, the values in the object given by the const reference are wrong and are all zero. Here is my question: Is load_construct_data() below correct? It looks like it is initializing the constant reference from a temporary, but I'm doing the same thing as the example in the documentation. class MyClass1 { friend boost::serialization::access; … }; class MyClass2 { friend boost::serialization::access; const MyClass1 & mc1_; MyClass2 ( const MyClass1 &mc1) : mc1_(mc1) { } ; }; namespace boost { namespace serialization { template<class Archive> inline void save_construct_data( Archive &ar, const MyClass2 *t, const unsigned int file_version) { // mc1_ is a reference and requires special handling -- save it // as a pointer const MyClass1 *mc1 = &t->mc1_ ; ar << boost::serialization::make_nvp("mc1_", mc1); } template<class Archive> inline void load_construct_data( Archive &ar, MyClass2* t, const unsigned int file_version) { MyClass1 *mc1; ar >> boost::serialization::make_nvp("mc1_", mc1); // NOTE: IS THIS INITIALIZING THE REFERENCE FROM A TEMPORARY? ::new(t) MyClass2(*mc1); } Thanks, Kurt

1 0

Problems to use Boost Graph
by Oodini 04 Jul '15

04 Jul '15

Hello, I am trying to understand the boost::dijkstra_shortest_paths() function. I don't understand the second and third arguments. The documentation says they're supposed to be properties map : OUT: predecessor_map(PredecessorMap p_map) The predecessor map records the edges in the shortest path tree, the tree computed by the traversal of the graph. Upon completion of the algorithm, the edges (p[u],u) for all u in V are in the tree. The shortest path from vertex s to each vertex v in the graph consists of the vertices v, p[v], p[p[v]], and so on until s is reached, in reverse order. The tree is not guaranteed to be a minimum spanning tree. If p[u] = u then u is either the source vertex or a vertex that is not reachable from the source. The PredecessorMap type must be a Read/Write Property Map whose key and value types are the same as the vertex descriptor type of the graph. but that doesn't work. I have to use an iterator on the properties map, by using : boost::make_iterator_property_map(pred_map.begin(),vertex_index) Does the documentation need reviewers ?

3 4