July 2005 - Boost-users - lists.boost.org

Is there Boost Network library that provides...
by BRIDGES Dick 20 Jul '05

20 Jul '05

I have the following requirements. One TCP socket - R&W ("event driven": R & W request arrivals not predictable) - threadsafe. If it's of any help, readS update the local state machine in response to events external to a "server". Events external to the local device are reported back to the server via writeS. Communication in both directions is in the form of discrete "messages". OS is, let's say, Linux. An IOStream wrapper would be nice. On the other hand, I must have fairly low level control over the socket (specify interface(s) used, set various SO_*, etc.) It just doesn't seem that I should be required to roll my own to provide what appears to me to be a fairly basic set of requirements. On the other hand, I haven't seen anything on Boost that provides the features I'm looking for. It looks like the closest would be the new iostreams library *after* Jonathan Turkanis adds socket support to Boost version 1.34. Is that pretty much it? Anyone got any other ideas that might work now? Suggestions sincerely appreciated, Dick Bridges "For every complex problem, there is a solution that is simple, neat, and wrong." H.L. Mencken

1 0

boost::array MSVC 7.1 /GL bug
by Johan Råde 20 Jul '05

20 Jul '05

Hi, I just discovered a bug when using boost::array with MSVC 7.1. I have project with about 100 cpp files and 20.000 lines of code. It uses about a dozen different boost libraries. When I compile with the compiler option /GL (whole program optimization) then I get an internal compiler error. It is certainly a compiler bug. The option /GL is a somewhat experimental feature. The boost::array source code contains a templatized assignment operator with type conversion. If I add a regular assignment operator, then the code compiles. In fact, I get a clean /W4 build :-) I have tried to reproduce the bug in a smaller example, but with no success. Smaller projects compile just fine. I propose that the following lines be added to array.hpp: // assignment without type conversion (sometimes needed with MSVC 7.1 with option /GL) array<T,N>& operator= (const array<T,N>& rhs) { std::copy(rhs.begin(),rhs.end(), begin()); return *this; } --Johan Råde Department of Mathematics Lund University Lund, Sweden

1 0

beginner question about boost Unit test framework
by J Liu 20 Jul '05

20 Jul '05

I am using Boost 1.31 with VC 6.0. I am testing the simplest example provided in the boost document. It seems "boost::unit_test" is not a valid namespace. After I change it to "boost::unit_test_framwork", it can compile but can not link, the error I got is "LINK : fatal error LNK1561: entry point must be defined". The exmaple I used: #include <boost/test/unit_test.hpp> using namespace boost::unit_test; test_suite* init_unit_test_suite( int argc, char* argv[] ) { test_suite* test= BOOST_TEST_SUITE( "const_string test" ); return test; } Thanks for any help

4 9

[date-time] Local time to UTC conversion?
by Charles BROSSOLLET 20 Jul '05

20 Jul '05

Hello, I'm looking for a way to convert a local time to UTC time. I found c_local_adjustor::utc_to_local, but is there the reciprocal? thanks, -- Charles Brossollet

2 1

Re: [Boost-users] Container/Algorithm question
by JOAQUIN LOPEZ MU?Z 20 Jul '05

20 Jul '05

----- Mensaje original ----- De: Elisha Berns <e.berns(a)computer.org> Fecha: Sábado, Julio 16, 2005 3:15 am Asunto: Re: [Boost-users] Container/Algorithm question > Joaquin, > > Sorry to be the bearer of bad news, but for some reason the multi- > indexcontainer starts to fail for some larger number of pairs (a few > hundred). The container doesn't/won't insert them, and to be > honest, I > don't have a clue at this point why. The code is so simple that I fail to see how it can be failing, so let's insert some traces: std::pair<iterator,bool> insert(const value_type& x) { iterator it0=impl.find(boost::make_tuple(x.first)); if(it0!=impl.end()){ typename impl_type::nth_index_iterator<1>::type it1= impl.template get<1>().find(x.second); if(it1!=impl.template get<1>().end()){ std::cout<<x.first<<","<<x.second<<std::endl; std::cout<<it0->first<<","<<it0->second<<std::endl; std::cout<<it1->first<<","<<it1->second<<std::endl; return std::make_pair(it0,false); } } return impl.insert(x); } Can you insert the std::cout output sentences above to your code and report what it's showing? If everything is working OK, when an insertion fails, x.first==it0->first and x.second=it1->second. Isn't it so? Can you provide the data the container is fed with when the first (erroneous) insertion failure happens? > If you're up for it you > might want > to stress test the code you sent me with thousands of unique pairs and > redundant pairs. > I'll examine this closer next Monday if we haven't located the problem before. Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

3 3

Serilization : Using serialization with socket++ for serializing objects over TCP connections : Problem
by nitin motgi 20 Jul '05

20 Jul '05

Hi All, I using boost::archive and socket++ libraries for serializing objects over TCP connections. Socket++ provides iosockinet class which is a derivative of iostream that supports i/o with two different buffers. Upon binding the iosockinet to a port I assign it to binary_iarchive and oarchive upon doing so .. the program aborts immediately. I am not sure what is causing this problem. Any help in this would be appreciated. ================= serialize_common.h ======================== class serialize_common { private: friend class boost::serialization::access; string name; int iVal; long lVal; unsigned int uiVal; float fVal; double dVal; template <class Archive> void serialize(Archive& ar, const unsigned int version) { ar & name; ar & iVal; ar & lVal; ar & uiVal; ar & fVal; ar & dVal; } public: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . }; ================= test.cpp ================================== /* 1. sockinetbuf is derived from sockbuf 2. iosockinet is derived from iosockstream which inturn is derived from iostream */ #include <sockinet.h> #include "serialize_common.h" int main() { serialize_common command; sockinetbuf si(sockbuf::sock_stream); si.bind(); cout << si.localhost() << ' ' << si.localport() << endl; si.listen(); while(1) { iosockinet s(si.accept()); boost::archive::binary_oarchive ao(s); <== Modifies s which comes non usable boost::archive::binary_iarchive ao(s); <== Aborts at this point char buf[1024]; while (s >> buf) { cout << buf << ' '; s << ::strlen(buf) << endl; } cout << endl; } return 0; } Thank you, Nitin

1 0

weak_ptr for intrusive_ptr's??
by Edson Tadeu 19 Jul '05

19 Jul '05

Is there an weak_ptr<> equivalent to the intrusive_ptr<>, to handle ownership cycles? Thanks.

2 1

cvs boost 1.33 date_time problem
by Serge Skorokhodov 19 Jul '05

19 Jul '05

Hi, I've built todays cvs boost library but date_time lib seems to have the same error as in 1.32 version. That is: wstring wts = _T("20021017T231217.12345"); wdate_facet* date_facet = new wdate_facet; wtime_facet* time_facet = new wtime_facet; ptime pt1( from_iso_string( convert_string_type<_TCHAR, char>(wts) ) ); long Y = pt1.date().year(); long M = pt1.date().month(); long d = pt1.date().day(); long h = pt1.time_of_day().hours(); long m = pt1.time_of_day().minutes(); long s = pt1.time_of_day().seconds(); long f = pt1.time_of_day().fractional_seconds(); wcout << L"Components of pt1:" << endl << L"Year: " << Y << endl << L"Month: " << M << endl << L"Day: " << d << endl << L"Hours: " << h << endl << L"Minutes: " << m << endl << L"Seconds: " << s << endl << L"Microseconds: " << f << endl; ptime pt2( gregorian::date(2002, 10, 17), posix_time::time_duration(23, 12, 17, 12345) ); wcout << pt2 << endl; Y = pt2.date().year(); M = pt2.date().month(); d = pt2.date().day(); h = pt2.time_of_day().hours(); m = pt2.time_of_day().minutes(); s = pt2.time_of_day().seconds(); f = pt2.time_of_day().fractional_seconds(); wcout << L"Components of pt2:" << endl << L"Year: " << Y << endl << L"Month: " << M << endl << L"Day: " << d << endl << L"Hours: " << h << endl << L"Minutes: " << m << endl << L"Seconds: " << s << endl << L"Microseconds: " << f << endl; Output is: Components of pt1: Year: 2002 Month: 10 Day: 17 Hours: 23 Minutes: 12 Seconds: 17 Microseconds: 0 2002-Oct-17 23:12:17.012345 Components of pt2: Year: 2002 Month: 10 Day: 17 Hours: 23 Minutes: 12 Seconds: 17 Microseconds: 12345 Another question is about wide char support. As far as I understand, wstring is supported only through stringstream and facets? TIA -- Serge

2 11

Re: [Boost-users] Error building date_time and ICE with multi_index
by Jasper, Alexander 19 Jul '05

19 Jul '05

Hi Bart! thanks *a lot". I tried it right away and it worked. I think this something that should be included in the new documentation dealing with stlport, that Jeff posted. W/o this the release binaries do not get build if one uses the procedure outlines there. BTW: great speedup compared to debug version. kind regards Alex > -----Ursprüngliche Nachricht----- > Von: boost-users-bounces(a)lists.boost.org > [mailto:boost-users-bounces@lists.boost.org]Im Auftrag von Bart > Gesendet: Dienstag, 19. Juli 2005 16:09 > An: boost-users(a)lists.boost.org > Betreff: Re: [Boost-users] Error building date_time and ICE with > multi_index > > > On Mon, 18 Jul 2005 22:21:28 +0200, Jasper, Alexander wrote: > > > Hi All! > > > > I'm currently starting to use more of boost, including libs > that need compiling. Namely date_time. > > I first of all have to thank you for creating boost. It > really boosts c++ development so much. And it is amazing what > can be done with c++ templates. > > > > I'm having problems linking the date_time lib. I'm using > the vc-7_1-stlport config. In the debug build everything is > fine. But no libs are build for release mode. At least there > are none w/o the g in their name in the stage/lib folder. The > pragma comment lib stuff complains when building my program > about not being able to find > libboost_date_time-vc71-mt-p-1_32.lib. And indeed, it is not there. > > > > I'm using this cmd-line to build date_time > > > > ..\..\build\bjam.exe -sTOOLS=vc-7_1-stlport > "-sBUILD=<define>BOOST_NO_STD_WSTRING" > "-sSTLPORT_4.6.2_PATH=D:\Alexander.Jasper\hlf\OpsMan-2\tao146_ > port\thirdparty\stlport" "-sstlport-iostreams=on" > --with-date_time stage > > > > Modify your BUILD variable setting to include "release": > "-sBUILD=release <define>BOOST_NO_STD_WSTRING" > > This worked for me. > > Bart Garst > > _______________________________________________ > Boost-users mailing list > Boost-users(a)lists.boost.org > http://lists.boost.org/mailman/listinfo.cgi/boost-users >

1 0

Error building date_time and ICE with multi_index
by Jasper, Alexander 19 Jul '05

19 Jul '05

Hi All! I'm currently starting to use more of boost, including libs that need compiling. Namely date_time. I first of all have to thank you for creating boost. It really boosts c++ development so much. And it is amazing what can be done with c++ templates. I'm having problems linking the date_time lib. I'm using the vc-7_1-stlport config. In the debug build everything is fine. But no libs are build for release mode. At least there are none w/o the g in their name in the stage/lib folder. The pragma comment lib stuff complains when building my program about not being able to find libboost_date_time-vc71-mt-p-1_32.lib. And indeed, it is not there. I'm using this cmd-line to build date_time ..\..\build\bjam.exe -sTOOLS=vc-7_1-stlport "-sBUILD=<define>BOOST_NO_STD_WSTRING" "-sSTLPORT_4.6.2_PATH=D:\Alexander.Jasper\hlf\OpsMan-2\tao146_port\thirdparty\stlport" "-sstlport-iostreams=on" --with-date_time stage after building stage/lib contains these files 18.07.2005 14:39 53.248 boost_date_time-vc71-mt-gdp-1_32.dll 18.07.2005 14:39 22.400 boost_date_time-vc71-mt-gdp-1_32.lib 18.07.2005 14:39 22.400 boost_date_time-vc71-mt-gdp.lib 18.07.2005 14:39 951.040 libboost_date_time-vc71-mt-gdp-1_32.lib 18.07.2005 14:39 951.040 libboost_date_time-vc71-mt-gdp.lib I have attached the bjam output to this email. But I can't find any error messages in there. Does anybody have an idea what is could be going wrong here? Thank would be great. And thanks a lot in advance. I have another issue with VC 7.1 with the multi_index lib. If I use it in conjunction with a smart_ptr it gives me an ICE. Does anybody know if that's a known bug of VC? Fixed in 8.0? To be specific this code fails: typedef boost::multi_index::multi_index_container< boost::shared_ptr<ps::Activity>, boost::multi_index::indexed_by< boost::multi_index::ordered_unique< boost::multi_index::tag<id>, BOOST_MULTI_INDEX_CONST_MEM_FUN(ps::Activity,unsigned int,getId)>, boost::multi_index::ordered_non_unique< boost::multi_index::tag<date>,BOOST_MULTI_INDEX_CONST_MEM_FUN (ps::Activity,boost::posix_time::ptime,getBegin)> > > TLegCache; It works ok with a regular pointer. This is supposed to work, isn't it? I have not yet tried with gcc. Another thing is that it consistently crashed with another class being index. Even with a regular pointer in that case. The typedef and the class are almost the same as above. The getter functions are const and return objects by value. I simply don't get it. This damn VC. My first impression was that things got A LOT better with 7.1. But apparently I'm wrong. Or is my code broken, after all? Thanks in advance for any hints or work arounds. kind regards Alex

4 3