- Boost-users - lists.boost.org

using shared pointer in a linked data structure
by Zumichu 11 Jun '03

11 Jun '03

Basically I'm having trouble understanding exactly how to use shared_ptr . If I have a class: GraphNode { public: AddChild(shared_ptr<GraphNode> n) { m_Children.push_back(n); shared_ptr<GraphNode> thisptr(this); n->m_Parents.push_back(thisptr); } DeleteChild(shared_ptr<GraphNode> n); private: vector<shared_ptr<GraphNode> > m_Children; vector<shared_ptr<GraphNode> > m_Parents; }; main() { shared_ptr<GraphNode> n1(new GraphNode()); shared_ptr<GraphNode> n2(new GraphNode()); n1->AddChild(n2); } doing it this way gives me a run-time error that seems to have to do with updating the child node's parent vector with the parent node's this pointer. Am I not using the shared_ptr right?

2 1

Re: [crc] Simple example please
by Vladimir Prus 11 Jun '03

11 Jun '03

Daryle, Daryle Walker wrote: > One compiler finished after the above 'fix'. The other (a variant of > GCC 3.1) complained about missing symbols at link-time; it could not > create the global objects hidden in Boost.CRC. (I haven't worked on > the code in a long time; there could be quirks on compilers written > after Boost.CRC.) > > When I ran the program under the compiler that worked, the "readsome" > function always returned zero! There were no suitable substitutes; the > "read" function doesn't tell you how many characters it read (bad if > the last block isn't full) and the multi-character "get" functions use > delimiters. So this program is stuck at single-character reads at the > moment. Maybe I should switch to the stream-buffer method you used. > > Can you try this out to make sure the problems aren't just mine? With gcc 3.3 on Linux, 'readsome' variant, 'get' variant and 'sbumpc' variant all produce the same output. However, I'm not sure that 'readsome' variant should work. Here's what "man std::iostream" tells about 'readsome'. if A > 0, extracts min(A,n) [ A = rdbuf()->in_avail() ] ... The goal is to empty the current buffer, and to not request any more from the external input sequence controlled by the streambuf. Concerning 'in_avail()': If a read position is available, returns the number of characters available for reading before the buffer must be refilled. Otherwise returns the derived showmanyc(). Lastly, about 'showmanyc()': Base class version does nothing, returns zero. So, this all leads to conclusion that 'readsome' is required to read all characted currently available in stream buffer, but may ignore anything else. I believe that the right way is to use 'read' function, together with 'gcount()' with tells how many characters were read by the last unformatted input operation. The main loop will become: do { ifs.read(buffer, BLOCK_SIZE); result.process_bytes(buffer, ifs.gcount()); } while(ifs); Of course, this variant produces the same results for me. Can you try with your compiler? - Volodya P.S. In fact, I still don't understand what 'in_avail()' is good for...

1 0

Re: [crc] Simple example please
by Vladimir Prus 11 Jun '03

11 Jun '03

Daryle, Daryle Walker wrote: > One compiler finished after the above 'fix'. The other (a variant of > GCC 3.1) complained about missing symbols at link-time; it could not > create the global objects hidden in Boost.CRC. (I haven't worked on > the code in a long time; there could be quirks on compilers written > after Boost.CRC.) > > When I ran the program under the compiler that worked, the "readsome" > function always returned zero! There were no suitable substitutes; the > "read" function doesn't tell you how many characters it read (bad if > the last block isn't full) and the multi-character "get" functions use > delimiters. So this program is stuck at single-character reads at the > moment. Maybe I should switch to the stream-buffer method you used. > > Can you try this out to make sure the problems aren't just mine? With gcc 3.3 on Linux, 'readsome' variant, 'get' variant and 'sbumpc' variant all produce the same output. However, I'm not sure that 'readsome' variant should work. Here's what "man std::iostream" tells about 'readsome'. if A > 0, extracts min(A,n) [ A = rdbuf()->in_avail() ] ... The goal is to empty the current buffer, and to not request any more from the external input sequence controlled by the streambuf. Concerning 'in_avail()': If a read position is available, returns the number of characters available for reading before the buffer must be refilled. Otherwise returns the derived showmanyc(). Lastly, about 'showmanyc()': Base class version does nothing, returns zero. So, this all leads to conclusion that 'readsome' is required to read all characted currently available in stream buffer, but may ignore anything else. I believe that the right way is to use 'read' function, together with 'gcount()' with tells how many characters were read by the last unformatted input operation. The main loop will become: do { ifs.read(buffer, BLOCK_SIZE); result.process_bytes(buffer, ifs.gcount()); } while(ifs); Of course, this variant produces the same results for me. Can you try with your compiler? - Volodya P.S. In fact, I still don't understand what 'in_avail()' is good for...

1 0

RE: [Boost-Users] to_simple_string and Windows?
by Tom Matelich 11 Jun '03

11 Jun '03

UNICODE is enabled by the _UNICODE define. Assuming you're still using MSVC 6 (remember to rebuild all often since your dependencies may not getting generated correctly), just go to your Project/Settings/C/C++ tab and in Preprocessor definitions, remove the UNICODE symbol from the list. Otherwise, remove it from your Makefile. > -----Original Message----- > From: adutoit [mailto:adutoit@stonethree.com] > Sent: Tuesday, June 10, 2003 7:01 AM > To: Boost-Users(a)yahoogroups.com > Subject: Re: [Boost-Users] to_simple_string and Windows? > > > It is not a compile-time error, but rather a run-time error. > Ie. if I compile any program which utilises > boost::to_simple_string with any > optimisation whatsoever; it simply aborts at the point where I do my > function call to "to_simple_string". That is irrespective of > my using a > type date or ptime as the function argument... > > I must stress again that "to_simple_string" works fine when I > compile it in > Linux. > > O, another question I have. Where do I disable Unicode > support (wchar_t > usage in fact!) alltogether for Boost? > > Previously I had great fun with lexical_cast.hpp which > utilised wchar_t > versions of some of the streaming operators because I had a > silly UNICODE > define as part of my project options. The solution was to > use /Zc:wchar_t > project option in M$VC6. > > But UNICODE defined caused problems in network browsing code > I used which > `normally' returned a LPSTR instead of the LPWSTR if UNICODE > is defined. > Long story. I did a temp hack in lexical_cast.hpp by > # define DISABLE_WIDE_CHAR_SUPPORT > #undef BOOST_NO_INTRINSIC_WCHAR_T > and getting rid of the redefinition of BOOST_NO_INTRINSIC_WCHAR_T. > > Now is there a way to do this cleanly, compile-time so that I > do not need to > keep on hacking the source code? > > Greetings, > Andre. > > > > ----- Original Message ----- > From: "Jeff Garland" <jeff(a)crystalclearsoftware.com> > To: <Boost-Users(a)yahoogroups.com> > Sent: Tuesday, June 10, 2003 2:51 PM > Subject: RE: [Boost-Users] to_simple_string and Windows? > > > > > Darn; here I go replying to my own post again... > > > Anyway; so far I have narrowed the problem down to one of > optimisation. > I > > > did a debug compilation and had no problem. A moment of > reflection led > me > > > to the conclusion that optimisation might have something > to do with it. > For > > > the record; I am using Boost in Windows, with Visual > Studio as IDE only, > and > > > the Intel 7.0 compiler. Oh; and STLport instead of > Windows' crappy STL > > > implementation. > > > > > > I.e. turning on "Maximise speed" optimisation causes > "to_simple_string" > to > > > fail on my build system. > > > > > > I'll try to setup the simplest proggie I can to > illustrate this problem. > > > > > > Any thoughts? > > > > I'm not aware of anyone else having a problem. Perhaps you can post > > the details of the compilation failure? > > > > Jeff > > > > > > Info: <http://www.boost.org> > > Wiki: > <http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl> > > Unsubscribe: <mailto:boost-users-unsubscribe@yahoogroups.com> > > > > > > Your use of Yahoo! Groups is subject to > http://docs.yahoo.com/info/terms/ > > > > > ------------------------ Yahoo! Groups Sponsor > ---------------------~--> > Get A Free Psychic Reading! Your Online Answer To Life's > Important Questions. > http://us.click.yahoo.com/Lj3uPC/Me7FAA/ySSFAA/EbFolB/TM > -------------------------------------------------------------- > -------~-> > > Info: <http://www.boost.org> > Wiki: <http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl> > Unsubscribe: <mailto:boost-users-unsubscribe@yahoogroups.com> > > > Your use of Yahoo! Groups is subject to > http://docs.yahoo.com/info/terms/ > > ----------------------------------------------------------------------- DISCLAIMER: Information contained in this message and/or attachment(s) may contain confidential information of Zetec, Inc. If you have received this transmission in error, please notify the sender by return email. -----------------------------------------------------------------------

2 1

newbie boost thread build under Linux help
by Mark Sizer 11 Jun '03

11 Jun '03

I had such success with Boost under Windows, that I'm now trying to get it up and running under Linux. This is not a happy experience. I got it to work, but this can't be the way it's supposed to be done: I'm using boost_1_30_0. I installed the bjam RPM (boost_jam_3.1.4-1.i386.rpm). bjam "-sTOOLS=gcc" from a bash command line. I had all sorts of very strange error messages, so I set about upgrading my Linux environment. I boostrapped the gcc 3.3 compiler (I had 2.6 something). Same error messages. I upgraded libstdc++ to libstdc++-v3.0.97-gcc-3.1.x.tar.gz (I'm seeing the gcc version mismatch as I type this, but it is the latest libstdc++). The "merging into an existing compiler" instructions seemed really complicated so I installed it and re-bootstrapped the compiler. Got a more sensible error message: Unknown compiler version please run the configure tests and report the results. (from gnu.hpp) I hacked gnu.hpp to allow the 3.3 compiler. The error message changed to being unable to find the std:: headers (first one was <utility> from select_stdlib.hpp). I seem to have at least four versions: /usr/include/g++-2/utility /usr/include/g++-3/utility /usr/include/g++-v3/utility (judging by the dates, this seems the correct one) /usr/local/include/c++/utility Obviously, none of these are in the compiler's include path. bjam "-sTOOLS=gcc" "-sGCC_INCLUDE_DIRECTORY=/usr/include/g++-v3" Wasn't helpful: "bits/c++config.h" can't be found. That's in /usr/include/g++-v3/i686-pc-linux-gnu. However, there doesn't seem to be a way to specify more than one include directory (neither multiple "-s" nor delimited paths in a single "-s" worked). So I copied everything from i686-pc-linux-gnu/bits to ../../bits. It built. This can't be normal. How is this process supposed to work? Thanks, - Mark

2 1

Mutex scoped_lock compile errors
by Darren Vincent Hart 10 Jun '03

10 Jun '03

Esteemed Boost Gurus, I am trying to compile a method with a scoped_lock call. If I use a mutex object declared in the class implementation file (logger.cpp) it builds, if I use a mutex of the same type declared as a private member of the logger class (in logger.h), I get the following compiler error: logger.cpp: In member function `void pel::logger::info(const std::string&, const std::string&, int, const std::string&) const': logger.cpp:36: error: no matching function for call to ` boost::detail::thread::scoped_lock<boost::mutex>::scoped_lock(const boost::mutex&)' /usr/include/boost/thread/detail/lock.hpp:62: error: candidates are: boost::detail::thread::scoped_lock<boost::mutex>::scoped_lock(const boost::detail::thread::scoped_lock<boost::mutex>&) /usr/include/boost/thread/detail/lock.hpp:67: error: boost::detail::thread::scoped_lock<Mutex>::scoped_lock(Mutex&, bool = true) [with Mutex = boost::mutex] make: *** [logger.o] Error 1 The relevant method from the implementation file is as follows: void logger::info(const std::string& msg, const std::string& file, int line, const std::string& function) const { // io_mutex is declared above in the pel namespace boost::mutex::scoped_lock scoped_lock(io_mutex); // no problem // mutex is a private member of the logger class boost::mutex::scoped_lock scoped_lock2(mutex); // line 36, fails std::cout << "...blah..." << std::endl; } Can someone explain why these are treated differently by the compiler ? Thanks, Darren Hart dvhart(a)byu.edu

3 4

RE: [Boost-Users] Struggling with bind
by Jean-Marc Prud'homme 10 Jun '03

10 Jun '03

Hi Paul, _1 is a placeholder argument. When you write bind(&CDocumentAttachments::Execute, _1, _T("print")) you are creating a functor that can be called with a single argument of type CDocumentAttachments. This is what is required by the for_each algorithm. It is equivalent to bind<HINSTANCE>(mem_fn(&CDocumentAttachments::Execute), _1, _T("print")) Look at the mem_fn doc. You still need this kind of functor if you add more arguments to CDocumentAttachments::Execute. This should create the functor you need bind(&CDocumentAttachments::Execute, _1, _T("print"), false) Hope this helps, Jean-Marc -----Original Message----- From: Paul Thompson [mailto:friedegg@ntlworld.com] Sent: Tuesday, June 10, 2003 7:54 AM To: Boost-Users(a)yahoogroups.com Subject: [Boost-Users] Struggling with bind Hi all. I've been using boost::bind to work around deficiencies with std::mem_fun_ref and std::bind2nd under MSVC6. I have a class called CDocumentAttachments which had a member function thus: HINSTANCE Execute(LPCTSTR verb) const; I was using bind with this as follows: for_each(filenames.begin(), filenames.end(), bind(&CDocumentAttachments::Execute, _1, _T("print"))); "filenames" is a const vector of CDocumentAttachments objects. This worked fine. A little later on in development, I changed the function signature for CDocumentAttachments::Execute to: HINSTANCE Execute(LPCTSTR verb, bool visible) const; How will I pass the extra bool to bind? I can't find any examples of passing more than one parameter to a member function. I thought about (among many others) for_each(filenames.begin(), filenames.end(), bind(&CDocumentAttachments::Execute, _2, _T("print"), false)); but the compiler (VC6) complains with two of the following: k:\vc\include\boost\bind.hpp(760) : error C2679: binary '[' : no operator defined which takes a right-hand operand of type 'const class boost::arg<2>' (or there is no acceptable conversion) k:\vc\include\boost\bind\bind_template.hpp(33) : see reference to function template instantiation 'struct HINSTANCE__ *__cdecl boost::_bi::evaluator3<struct HINSTANCE__ *>::eval(const class boost::_bi::list3<class boost::arg<2>,class boost:: _bi::value<char *>,class boost::_bi::value<bool> > &,struct boost::_mfi::cmf2<struct HINSTANCE__ *,class CDocumentAttachments,char const *,bool>,class boost::_bi::list1<class CDocumentAttachments const &> &)' being compiled I know I'm being really thick here, and I'm sure there's something really obvious that I'm missing, but reading through the documentation hasn't helped me. Thank you very much. Paul -- Paul Thompson Programmer McLellan Software Design Ltd. Info: <http://www.boost.org> Wiki: <http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl> Unsubscribe: <mailto:boost-users-unsubscribe@yahoogroups.com> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

1 0

to_simple_string and Windows?
by adutoit 10 Jun '03

10 Jun '03

Hi. I have used the boost::date_time library in Linux with no problems, but the same can not be said of my experience with it in Windows... My current pain in Windows is with the "to_simple_string" method used to convert a boost::date_time value to guess what, a string. This works fine in Linux, but not in stupid Windows. The following line just borks: std::cout << to_simple_string(date(2003,4,30)) << std::endl; Any little unknown configuration tricks etc. I need to be aware of here? Has anyone used the "to_simple_string" method without problems in Windows? I'll appreciate any help in this regard. Greetings, A. [Non-text portions of this message have been removed]

2 3

New file uploaded to Boost-Users
by Boost-Users＠yahoogroups.com 10 Jun '03

10 Jun '03

Hello, This email message is a notification to let you know that a file has been uploaded to the Files area of the Boost-Users group. File : /uBLA Performance 2.pdf Uploaded by : sumohurd <matt(a)finray.net> Description : Simplistic uBLA Performance chart, second version You can access this file at the URL http://groups.yahoo.com/group/Boost-Users/files/uBLA%20Performance%202.pdf To learn more about file sharing for your group, please visit http://help.yahoo.com/help/us/groups/files Regards, sumohurd <matt(a)finray.net>

1 0

[crc] Simple example please
by Vladimir Prus 10 Jun '03

10 Jun '03

Hi Alex, > I'd like to calculate a value for a cyclic redundancy check (crc) for a > number of files. > > This isn't going to be transmitted anywhere, I simply want to compare the > values to see whether the files are exactly the same. I know I can simply > check byte for byte, but I've got a few thousand files and it's easier to > calculate the crc for each and see whether any are the same first. Here's the function that I use for the same purpose. Might not be optimal, but works: unsigned file_crc(const string& name) { #if defined(__GNUC__) && __GNUC__ < 3 ifstream ifs(name.c_str(), ios::binary); #else ifstream ifs(name.c_str(), ios_base::binary); #endif if (!ifs) return 0; else { using namespace boost; crc_32_type crc32; int c; while( (c = ifs.rdbuf()->sbumpc()) != -1) crc32.process_byte(char(c)); return crc32.checksum(); } } I believe some example of this kind should be included... Daryle, what do you think? HTH, Volodya

3 2