June 2015 - Boost-users - lists.preview.boost.org

fusion error when compiling graphviz code using VC2010
by Carl Wallace 30 Jun '15

30 Jun '15

We have been using Boost's graphviz support for several years and have upgraded through a few versions of Boost with no issues in some long stable code. However, when we moved to 1.58 our Windows build broke. We are still using VC2010, with plans to move to 2015 in the not-too-distant future. We have not yet tried 1.58 in VC2015. The error message is very large but emanates from here: boost/include/boost-1_58\boost/fusion/container/list/cons.hpp(86): error C2664: 'boost::xpressive::detail::alternates_list<Head,Tail>::alternates_list(cons t Head &,const boost::fusion::nil &)' : cannot convert parameter 1 from 'const boost::fusion::cons_iterator<Cons>' to 'const boost::xpressive::detail::static_xpression<Matcher,Next> &’ I am not familiar with the fusion or xpressive libraries, but after picking through the error messages and review of diff of changes since 1.54 (the version we are moving from), this looks like a template specialization problem. The problem is rooted in a regex that uses xpressive in the escape_dot_string function in the graphviz library. In 1.58 the function at the line identified in the error snip above is defined as follows: template <typename Sequence> BOOST_CONSTEXPR BOOST_FUSION_GPU_ENABLED cons( Sequence const& seq, typename boost::enable_if< mpl::and_< traits::is_sequence<Sequence>, mpl::not_<is_convertible<Sequence, Car> > > // use copy to car instead >::type* /*dummy*/ = 0 ) : car(*fusion::begin(seq)), cdr(fusion::next(fusion::begin(seq)), mpl::true_()) {} When we replace this with the following implementation from 1.54 the build works and it appears the desired functionality is working (though more testing is required). template <typename Sequence> BOOST_CONSTEXPR BOOST_FUSION_GPU_ENABLED cons( Sequence const& seq, typename boost::disable_if< mpl::or_< is_convertible<Sequence, cons>, // use copy ctor instead is_convertible<Sequence, Car> // use copy to car instead > >::type* /*dummy*/ = 0 ) : car(*fusion::begin(seq)), cdr(fusion::next(fusion::begin(seq)), mpl::true_()) {} Any thoughts? The new code in 1.58 builds and runs fine with no mods on OS X. We have not tried Linux yet.

2 1

[msm] is that correct behavior?
by Sam Fisher 29 Jun '15

29 Jun '15

Hi Christophe, For an fsm with orthogonal regions if any region rejects an event (guard returns false), the library blocks all anonymous transitions. for example, an fsm has two regions. Region 0 returns HANDLED_TRUE, region 1 returns HANDLED_GUARD_REJECT for the same event. region_processing_helper combines it into a single "HANDLED_TRUE | HANDLED_GUARD_REJECT" value. However function process_event_internal<> (state_machine.hpp) creates handle_eventless_transitions_helper that takes (handled == HANDLED_TRUE) as a condition for allowing transitions. should it be (handled & HANDLED_TRUE)? as a result, fsm is stuck in event that has an anonymous transition to another event. PS. I've seen the same condition (handled == HANDLED_TRUE) in do_post_handle_deferred function, but I do not use deferred events so I did not check that part.

2 1

[PBGL] Creating graph on multiple processes
by Alexander Frolov 29 Jun '15

29 Jun '15

Hi! I am trying to create undirected weighted graph on two mpi-processes. I use add_edge function to add edges (master (pid=0) process is responsible for this). For small graphs everything is ok. But when I create larger graphs (for instance, number of vertices is 2^20 and number of edges is 2^28) the program hangs supposedly in add_edge. For mpirun -np 1 there is no such problem. What can cause this issue? Thanks for any help! Best, Alex

1 1

Boost.DLL formal review
by Vladimir Prus 29 Jun '15

29 Jun '15

The formal review of Boost.DLL library by Antony Polukhin starts today. Boost.DLL is a C++98 library for comfortable work with DLL and DSO. Library provides a portable across platforms way to: - load libraries at runtime - import and export any native functions and variables - make alias names for C++ mangled functions and symbols - query libraries/objects and executables for sections and exported symbols - self loading and self querying - getting program and module location by exported symbol The documentation can be found at: http://apolukhin.github.io/Boost.DLL/index.html and the source can be obtained at: https://github.com/apolukhin/Boost.DLL Please post your reviews on the mailing list and if possible, answer the following questions: - Should the library be accepted? - How useful is it? - What's your evaluation of - Design - Implementation - Documentations - Tests - How much effort did you put into your evaluation? - Did you attempt to use the library? On what systems and compilers? This formal review has a couple of adjustments in place: - It will run for two calendar weeks, not for 10 days, as a number of people said 10 days is too little. That said, please try to submit a review during first week, so that necessary discussion can happen conveniently. - The documentation site has Google Analytics enabled, which will allow us to see which pages and navigation paths are most useful for reviewers. No personally identifiable information will be collected. I will post the data after the review is over. Thanks, Volodya

1 0

[Boost][DateTime] Non conforming iterators?
by dariomt＠gmail.com 29 Jun '15

29 Jun '15

Hi, The documentation for Boost.Date_Time iterators says that "Date iterators are a model of Bidirectional Iterator". However, I am not able to use them in some STL algorithms because of the lack of a required nested type. The following simplified example does not compile: #include <boost/date_time/gregorian/gregorian.hpp> #include <algorithm> void main() { using namespace boost::gregorian; date v[5]; std::copy_n( month_iterator( date(2015,06,25) ), 5, v ); } On VS2013 fails with error 'difference_type' : is not a member of 'boost::date_time::date_itr<boost::date_time::month_functor<boost::gregorian::greg_durations_config::date_type>,boost::gregorian::date>' Compilation on gcc and clang fails with a similar error complaining about the lack of 'difference_type'. Isn't a nested difference_type a requirement of BidirectionalIterator? Am I using these date iterators incorrectly? Regards

2 1

[multi_index] are indices in a random access index stable on addition, can we have an "index_of" method?
by Nick Stokes 27 Jun '15

27 Jun '15

I have a stream of pairs of elements of type T, and want to insert individual T's uniquely into a container and convert the pairs to a pair of indices into this container. Boost.MultiIndex seems to be cut out exactly for this job, but I not sure about how to obtain the indices or stability of them on addition: Take, namespace bmi = boost::multi_index; typedef bmi::multi_index_container< T, bmi::indexed_by < bmi::hashed_unique< bmi::identity<T> > , bmi::random_access<> >> MIC; MIC elements; std::vector<std::pair<int,int>> pairs; for( const std::pair<T,T>& pt : stream ) { auto t1 = elements.insert(pt.first).first; // not all insertions are successful of course auto t2 = elements.insert(pt.second).first; pairs.insert( index_of(t1), index_of(t2) ); } // where index_of() is auto index_of = [&]( MIC::const_iterator i ) { return elements.project<1>(i) - elements.get<1>().begin(); } So my questions are (1) Is the implementation of index_of() correct? (or better, is there something like this already for random access indices) (2) Is the way they are used in above loop stable, where elements are inserted after an index is obtained? Thanks Nick

2 1

Re: [Boost-users] [Test] How to Access Private Data from a Test
by Jürgen Hunold 26 Jun '15

26 Jun '15

Hi Steven, On Donnerstag, 25. Juni 2015 21:24:51 Steven Clark wrote: > This remark really puzzles me. > Do you feel that white-box testing should > never be done? Or should never be needed? Both, in fact. See Gavin's mail for the rationale. He explained it better than I would have :-) In fact, I've got a long way with black-box testing only in the past five years, so I have strong evidence that white-box testing is unnecessary. But this is on a fresh, test-driven developed code base. Designed to be black-box testable. > > On Jun 25, 2015, at 4:52 PM, Jürgen Hunold <jhunold(a)gmx.eu> wrote: > > In my opinion, a test should never access private data. Note the "In my opinion" :-) > > But there are times > > and legacy code where I had no other choice, too. And then I wrote the disclaimer that sometimes you can't do everything with black-box testing. I tried hard to get a test harness up for some legacy code in the years before and got stuck in multiple nested class hell, some d- Pointer magick and class hierarchies hard to instantiate. That is the reason why I was able to actually grep for the friend declarations needed in those cases. Not nice in my opinion, but "an ugly test is better than no test" (see http://www.artima.com/weblogs/viewpost.jsp?thread=203994 for more on a pragmatic view on unit tests). Which leads to the question if the OP was able to solve the issue or stopped reading after the first paragraph. Yours, Jürgen -- * Dipl.-Math. Jürgen Hunold ! * voice: ++49 4257 300 ! Fährstraße 1 * fax : ++49 4257 300 ! 31609 Balge/Sebbenhausen * jhunold(a)gmx.eu ! Germany

1 0

[Test] How to Access Private Data from a Test
by Merrill Cornish 26 Jun '15

26 Jun '15

In the Bad-Old-Days when I had to roll my own, I would define my tests as methods of a special test class. If those tests needed to access private data in the classes under test, I would declare special test class as a friend of the class under test--one extra statement and everything worked. How do I do that using Boost test (1.58.0) with the AUTO_TEST family of macros? In the labyrinth of macro expansions, I found a struct name with the argument name from BOOST_AUTO_TEST_CASE(), but a friend of that name doesn't work. The ideal case would be to declare the string name in BOOST_AUTO_TEST_SUITE() as a friend so one friend declaration does it all. Alternately, having to declare a separate friend for each BOOST_AUTO_TEST_CASE() would be tedious (especially for large classes), but it would work. At the moment, I'm using the ole private=public trick: #define private public #include "header_under_test" #undef private However, I've already discovered one system header, <sstream>, that won't work with that kludge. There may be others. Is there another workaround for Boost Tests? I can't be the only one who has to access private data in tests. Merrill Cornish

3 3

Interprocess - 3rd process cannot read/write the shared memory
by Lloyd 25 Jun '15

25 Jun '15

Hi, I am working on Windows 7, boost 1.57. I am using interprocess shared memory. I have two processes A and B. A creates the shared memory with read/write permission, then B opens it and writes to it. In this case it works fine. As the next step process A terminates and starts again and opens the same shared memory. But in this case, it seems that A and B are reading/writing to different memory! Is this a defined behavior? If I make B to reopen the memory, it works! Thanks, Lloyd

1 0

[datetime] little problem with periods
by Georg.Kellerer＠lfu.bayern.de 24 Jun '15

24 Jun '15

Hello, i am using Boost V1.5.8 under vs 2010 with a german windows 7 64 bit. When I tried to calculate a time period I always ran into the same problem. The period is calculated one day too short. I wrote a little program to illustrate my problem. It's important that you change the value of the variable test to the current date. Is this a bug or do I use the library in an incorrect way? Thanks a lot, Georg #include "boost/date_time.hpp" #include <iostream> using namespace std; using namespace boost::gregorian; int main() { date today = day_clock::local_day(); date test(2015,6,23); // Change value to current date date first(1965,10,26); date_period period(first, today); if (!period.contains(test)) cout << "shouldn't be our testdate in the period? " << endl; }

4 3