September 2003 - Boost-users

Icompatibility between MS cl preprocessor and boost macro constructs
by m.doppler＠icoserve.com 22 Sep '03

22 Sep '03

Hi, the header files boost/iterator_adaptors.hpp and boost/regex/v3/instances.hpp contain macro constructs that are icompatible with the Microsoft compiler cl when using them in a source file, preprocessing the source to a another source file and afterwards compiling the preprocessed source file. For iterator_adaptors.hpp, cl version 13.00 is affected, for instances.hpp cl 13.00 and 13.10. The problem can be reproduced by compiling the following simple source file (test.cpp) in the base boost directory (boost-1.30.2): #include "boost/iterator_adaptors.hpp" #include "boost/regex.hpp" int main(void) { return 0; } The commandline to preprocess the file is: cl /TP /GX /EHsc /Zc:wchar_t /Zc:forScope /c /I. /P test.cpp The commandline to compile the previously preprocessed file is: cl /TP /GX /EHsc /Zc:wchar_t /Zc:forScope /c /I. test.i Note that when directly compiling the same test program using the following commandline the problem disappears: cl /TP /GX /EHsc /Zc:wchar_t /Zc:forScope /c /I. test.cpp One might expect that the last command line should produce the same result as the first two, but unfortunately this is not the case. I have attached a patch file that solves the icompatibility with the preprocessor by incorporating the following changes: - In instances.hpp the patch adds single spaces between the BOOST_DEFAULT_ALLOCATOR and the end of some template declarations. - In iterator_adaptors.hpp the conditional BOOST_WORKAROUND macros are refactored from inside the template declarations of indirect_iterator_generator and indirect_iterator_pair_generator to the front of them by duplicating the template struct definitions minus the BOOST_WORKAROUND code in the else branch of the conditional macro. In case you wonder why we need to preprocess the source file into a seperate file before compiling it: This is required to use distcc (http://distcc.samba.org) with Microsofts cl. Is there a chance this patch gets accepted for the next boost version? Michael --- michael doppler ----------------------------------- developer icoserve it gmbh innrain 98, 6020 innsbruck tel: +43 - 512 - 567622 - 250 fax: +43 - 512 - 567622 - 111 web: www.icoserve.com email: m.doppler(a)icoserve.com ---------------------------------------------------------------

2 1

boost::TokenizerFunction
by Zenaan Harkness 21 Sep '03

21 Sep '03

TokenizerFunction "Expression Semantics" table (documentation - eg. on debian: /usr/share/doc/libboost-doc/HTML/libs/tokenizer/index.htm), states that "tok is constructed", as part of "Precondition". However I note two things: 1: on looking at the source code for the provided example functions, each constructs a Token as it's first (or second) line of code in the call to operator(). 2: the "Semantics" says "The return value [bool] indicates whether a new token was found in the sequence". If there is no token available (eg. we are already at end of stream(/string) and will return false) why waste effort constructing storage for a token? Why does the doco say that the token must be constructed as a precondition? Is the code wrong, or the doco - ie. should my custom function construct a new (empty) token, or follow doco and have user construct one? It would seem potentially more efficient for the user to construct one, as they may be reused. I am trying to use (my own custom) function directly, in a for loop. Why would I use tokenizer class to wrap it? There is no example on using a function directly, thus these string of questions which I would like some answers on. The documentation for TokenizerFunction doesn't seem to have a proper doxygen-like api description - at least not on my local debian installation. So I have to revert to the source. As I'm a newbie, it is requiring quite a bit of going back and forth to try and figure it all out - eg., should I use a tokenizer (container), token_iterator, or just for-loop on a specific (custom) tokenizer function?? Finally, as I've been trying to figure out how to write my own TokenizerFunction. Delving into the code for the example TFs has eventually led me there, but a simpler example in the documentation would be very useful for other newbies, I'm sure. - I know, I should post the code. Well, I'm not done yet, and I'm implementing a function I need specifically, as opposed to just for learning or doco, but perhaps later in the week I can send something along if no one beats me to it. Once I understand it all a bit better. Simple one-hit functions can be a good way to learn how to code what you need (ie. not templatized). A progression from a one-hit to a templatized version would be ideal for learning. TIA Zen -- Mr Zenaan Harkness Phone +61 (0)412 166 990 Please respect the confidentiality of this email.

1 0

Q: exceptions and -fPIC
by Andrey Nechypurenko 19 Sep '03

19 Sep '03

Hi, I am on the following platform: Linux 2.6.0-test5 #11 Wed Sep 10 20:28:31 CEST 2003 i686 unknown Reading specs from /usr/lib/gcc-lib/i386-slackware-linux/3.2.2/specs Configured with: ../gcc-3.2.2/configure --prefix=/usr --enable-shared --enable-threads=posix --enable-__cxa_atexit --disable-checking --with-gnu-ld --verbose --target=i386-slackware-linux --host=i386-slackware-linux Thread model: posix gcc version 3.2.2 Boost version 1.30.2 compiled with: bjam "-sTOOLS=gcc" I have a question regarding the following boost::filesystem use case: try { for(fs::directory_iterator dir_itr(full_path); dir_itr != end_iter; ++dir_itr ) { // use dir_itr } } catch(const std::exception &ex) { std::cout << ex.what() << std::endl; } In case of full_path pointed to the directory without read permissions the corresponding exception is thrown and catched in the catch block. My problem is that if I compile my application with -fPIC option, I am unable to catch the exception anymore and the program is simply aborted. I need to use -fPIC because of other library which has own build system and if I follow suggested guidlines for Makefile, I get -fPIC automatically. I would verry appreciate if somebody could tell me are there any ways to make the code mentioned above work when it is compiled with -fPIC? Should I recompile boost with -fPIC? If yes, what is the right place to set this parameter? BTW how it is supposed to write a shared library which includes boost code? As I understand, than my code must be compiled with -fPIC... Thanks, Andrey. P.S. Here is example I am using and the makefile: // ttt.cpp #include <iostream> #include <boost/filesystem/operations.hpp> #include <boost/filesystem/path.hpp> namespace fs = boost::filesystem; int main(int argc, char **argv) { try { fs::path full_path = fs::system_complete(fs::path(argv[1], fs::native)); fs::directory_iterator end_iter; for(fs::directory_iterator dir_itr(full_path); dir_itr != end_iter; ++dir_itr ) { std::cout << dir_itr->leaf() << std::endl; } } catch(const std::exception &ex) { std::cout << ex.what() << std::endl; } return 0; } Makefile: CPPFLAGS := -I/usr/local/src/boost LDFLAGS := -L/usr/local/src/boost/libs/filesystem/build/bin/libboost_filesystem.a/gcc/debug/runtime-link-dynamic/ LIBS := $(LIBS) -lboost_filesystem ttt: ttt.cpp c++ $(CPPFLAGS) -g -ftemplate-depth-128 -fPIC -Wall $(LDFLAGS) -o $@ ttt.cpp $(LIBS) __________________________________ Do you Yahoo!? Yahoo! SiteBuilder - Free, easy-to-use web site design software http://sitebuilder.yahoo.com

1 0

Spirit library - Problem with the copy constructor of the class rule<> ?
by fritz.birnbacher＠t-online.de 18 Sep '03

18 Sep '03

Hello, I am new in using boost and I played a little bit with the spirit library. I tried following code using MSVC 7.1 and boost 1.30.2 #include"stdafx.h" #include<spirit.hpp> using namespace boost::spirit; rule<> GetRule() { rule<> r = anychar_p; return r; } void Parse1() { char s[] = "ABCD"; parse(s, GetRule()); } void Parse2() { rule<> r = anychar_p; char s[] = "ABCD"; parse(s,r); } int _tmain() { Parse1(); return 0; } Parse1() throws an exception (access violation) Parse2() works correct. Is it a compiler problem or a problem in spirit or my problem ? Thanks

2 1

may it change ?
by maxch 17 Sep '03

17 Sep '03

Hi All, I am a new user of Boost. but when I look the source of Boost, I found that almost all the implementations is put in head files together, not as usual to put into CPP file. I know, it must be intended, however, could it possible to move the implementations into cpp file in future? Sorry if this question has been posted. Thanks!

2 2

Boost.Python - implementing slicing on extension container class
by Steve Horne 16 Sep '03

16 Sep '03

I'm currently writing a set of container classes using Boost.Python to wrap an existing C++ library. Progress has been very good - not even two days and I've wrapped all the existing C++ methods (at least 100). But of course a library designed for C++ is not a library designed for Python. Job 2 is to make it more Pythonic. The current problem is implementing slicing in the __getitem__, __setitem__ and __delitem__ methods. These are already working with single keys. To handle slicing, I need to identify in the key object is a slice object. This is the sticking point right now. I can access the type of the object using p.attr ("__class__") - but how do I get access to the slice type object from __builtins__ in order to compare? I could probably use a workaround such as extracting and checking a string representation from p.attr ("__class__").str (), but it seems to me that the ability to import symbols from another module (whether __builtins__ or some other Python or extension module). This could reduce to the ability to convert a PyObject* to an object instance - even more useful for using any C API functions that don't have a Boost.Python wrapper - but there seems to be no obvious way to do this. In addition, my containers are currently declared as follows... class_<classname, boost::noncopyable>("classname") ... ; The boost::noncopyable is there because the containers don't have copy constructors due to certain issues using the library in C++. Therefore, I'd like to better understand the implications of this. For instance, in Python the copy constructor is often used to make copy the entire container (as opposed to just rebinding names). Presumably Boost.Python supports this using the C++ classes copy constructor. This raises the questions... 1. How can I create an object to return to Python, containing an instance of the C++ container, without removing boost::noncopyable? (I need to do this to implement slicing too) This, I think, is roughly what is done in the Inheritance section of the tutorial so I probably need something like the 'factory' function. So, to reduce it to a simple copy-function example, the following should be about right... Base* container::make_copy () { container *temp = new container (); temp->do_copy (*this); return temp; } ... .def("make_copy", container::make_copy, return_value_policy<manage_new_object>()) I think this should be right, but confidence is low. 2. How can I support the Python copy constructor without having a C++ copy constructor. I'm using boost-1.30.2 with Python 2.2.3 and Visual Studio 7. Any advice would be appreciated. Thanks.

2 4

Web page errors
by Reed Hedges 16 Sep '03

16 Sep '03

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Looks like I just found the *real* list :) The web page (http://boost.org/more/mailing_lists.htm#users) still links to the Yahoo page, and seems to have some other problems (some repeated text below the horizontal line). Would it be possible to fix that soon so as not to confuse others like me too much? thanks reed - -- Reed Hedges reed(a)zerohour.net http://zerohour.net/~reed -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.1 (Darwin) iD8DBQE/ZHFjFK83gN8ItOQRApIQAJ9Mav7GFLBokSwHBb5s+3jzGF/UuwCeMkq7 FN80pQo0c7osCwWm0mNFEsk= =sLUi -----END PGP SIGNATURE-----

2 1

building Jam and Boost on Solaris
by banibrata dutta 16 Sep '03

16 Sep '03

hi, is it possible to build Jam and Boost on Solaris. going thru the Boost.Jam documentation, for building Jam from scratch, i figured that the latest Jam tarball, doesn't seem to have a Makefile. in that case how do i build Jam, and how do i build Boost? any help would be appreciated. thanks, bdutta. _________________________________________________________________ Interact with peers. Learn from experts. http://server1.msn.co.in/sp03/teched/index.asp Be tech-empowered!

3 2

Bug in boost::format with 0 valued integer
by Simon Clift 15 Sep '03

15 Sep '03

Hi folks, There appears to be a bug in boost::format, an unitialized value problem spotted when I was using valgrind on a program of mine. The following test code illustrates the (admittedly minor?) problem: #include <boost/format.hpp> #include <iostream> using namespace std; int main() { int dummy = 5; cout << boost::format( "%05d" ) % dummy << endl; dummy = 0; cout << boost::format( "%05d" ) % dummy << endl; }; I compiled this with -g and when run with valgrind it seems that the put_last function in feed_args.hpp sends an uninitialized value into the standard library <<(int) operator when the integer to be printed is 0 and leading zeros are specified. Valgrind reports (clipped for size) on the second ::format call (but not the first) of the test code: ==28879== Conditional jump or move depends on uninitialised value(s) ==28879== at 0x40289470: (within /usr/lib/libstdc++.so.5.0.3) ... ==28879== by 0x4028F36B: std::ostream::operator<<(int) (in /usr/lib/libstdc++.so.5.0.3) ==28879== by 0x804CF8D: void boost::io::detail::(anonymous namespace)::put_last<char, std::char_traits<char>, int>(std::basic_ostream<char, std::char_traits<char> >&, int&) (feed_args.hpp:114) ==28879== by 0x804C9A2: _ZN5boost2io6detail31_GLOBAL__N_test00Fill.cppfx7Zjb3putIcSt11char_traitsIcERiEEvT1_RKNS1_11format_itemIT_T0_EERSbIS9_SA_SaIS9_EERSt19basic_ostringstreamIS9_SA_SE_E (feed_args.hpp:176) ==28879== by 0x804BF44: void boost::io::detail::distribute<char, std::char_traits<char>, int&>(boost::basic_format<char, std::char_traits<char> >&, int&) (feed_args.hpp:237) ==28879== by 0x804ABB6: _ZN5boost2io6detail4feedIcSt11char_traitsIcERiEERNS_12basic_formatIT_T0_EESA_T1_ (feed_args.hpp:246) ==28879== by 0x804A14B: boost::basic_format<char, std::char_traits<char> >& boost::basic_format<char, std::char_traits<char> >::operator%<int>(int&) (format_class.hpp:69) ... Hope this helps. Otherwise I'm enjoying using boost! Best regards --- Simon -- Simon Clift E-mail: ssclift at math dot uwaterloo dot ca Ph.D. Student Scientific Computation Group University of Waterloo Waterloo, Ontario, Canada N2L 3G1

2 2

re: copy_if
by Witz 15 Sep '03

15 Sep '03

> > template<class II, class OI, class PRED> > > std::pair<II,OI> copy_if(II ibegin, II iend, OI obegin, OI oend, > > PRED p) > > { > > for(; ibegin != iend; ++ibegin) > > { > > if(p(*ibegin)) > > { > > *obegin = *ibegin; > > if(++obegin == oend) > > break; > > } > > } > > return std::make_pair(ibegin, obegin); > > } > > > > This should check whether obegin == end before copying, in case obegin > == oend initially. I don't know whether it should stop as soon as it > finds obegin == oend or wait until it knows there's another item to be > copied. The former behaviour saves a little time in the case that no > more items are wanted once the output buffer is full. The latter > behaviour avoids the occasional need for another pass that copies > nothing, if copy_if is called repeatedly to fill an output buffer which > is then flushed. > Ben. I've tended to rely on the common sense of the user (after all i cannot detect other possible errors such as 'oend < obegin') but i suppose the following addition at the start would seem sensible: ASSERT(obegin != oend); As for your second point - the amount of time saved by 'the former behaviour' depends on how expensive the predicate is and how many times it will be called if no more satisfactory items are available. As the entries in the output buffer are overwritten as and when required there is no need in any general sense to clear it between calls and this surely obviates any advantage your proposed approach may have. Witz

1 0

Re: Boost.Python - implementing slicing on extension container class
by Steve Horne 15 Sep '03

15 Sep '03

I have solved the issues I was having from my previous email. The cause of my problem was mainly that the information I needed was in the 'Embedding' section of the Boost.Python tutorial - I am not doing embedding so it did not occur to me to look there.

1 0

How do I browse the current 'boost-users' list?
by Andrew Bache 15 Sep '03

15 Sep '03

Is it possible to browse items on this list? I posted this via email having got to this page ... http://lists.boost.org/mailman/listinfo.cgi/boost-users What is not immediatly obvious (to me) is how I browse the current status of the list. Is there a URL that displays headers of the most recent posts? Can someone please point me at it? (There is a URL on the above page for an archive). Email reply please - for obvious reasons. Thanks Andy

2 1

RE: [Boost-users] Re: copy_if
by Ben Hutchings 15 Sep '03

15 Sep '03

Witz <witz(a)nildram.co.uk> wrote: > You can find an implementation of copy_if in Boost Wiki here: > > http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl > ?STLAlgorithmExtensions > > An extension that i find useful is the following: > > template<class II, class OI, class PRED> > std::pair<II,OI> copy_if(II ibegin, II iend, OI obegin, OI oend, > PRED p) > { > for(; ibegin != iend; ++ibegin) > { > if(p(*ibegin)) > { > *obegin = *ibegin; > if(++obegin == oend) > break; > } > } > return std::make_pair(ibegin, obegin); > } > > Here we have a finite sized sink. Items are copied from the source if > they satisfy the condition and if there is room left in the sink. The > positions reached are returned on termination of the algorithm. This should check whether obegin == end before copying, in case obegin == oend initially. I don't know whether it should stop as soon as it finds obegin == oend or wait until it knows there's another item to be copied. The former behaviour saves a little time in the case that no more items are wanted once the output buffer is full. The latter behaviour avoids the occasional need for another pass that copies nothing, if copy_if is called repeatedly to fill an output buffer which is then flushed. Ben.

1 0

BGL: finding all minimum cut sets
by Trevon Fuller 14 Sep '03

14 Sep '03

Does anyone know how to find all of the minimum size cut sets of a graph (if more than 1 exists) using the BOOST graph library? The example in the BGL book just finds 1 minimum cut set (p. 106ff.).

1 0

Cygwin: Building a static libboost_thread
by Reed Hedges 14 Sep '03

14 Sep '03

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hello, I have tried every combination of BUILD parameters but I can't seem to get a static libboost_thread.a on gcc/Cygwin (in "no cygwin" mode: TOOLS=gcc-nocygwin). Is it possible? I am using 1.30.2, IIRC. (If on the other hand someone knows how to get DevC++ (with msys gcc) to link against the shared library that would be helpful too) thanks reed - -- Reed Hedges reed(a)zerohour.net http://zerohour.net/~reed -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.1 (Darwin) iD8DBQE/ZHFRFK83gN8ItOQRArZRAJ9M6ohiYEoPqIjXurvdJrx07YJ4TACfcHRg 5d3XguoTJNDWGgSId/zKcEY= =t/LR -----END PGP SIGNATURE-----

1 0

Test
by Victor A. Wagner, Jr. 13 Sep '03

13 Sep '03

I haven't received anything in over a week, this is just a test of the address, please ignore. Victor A. Wagner Jr. http://rudbek.com The five most dangerous words in the English language: "There oughta be a law"

1 1

Re: copy_if
by Witz 13 Sep '03

13 Sep '03

You can find an implementation of copy_if in Boost Wiki here: http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl?STLAlgorithm… An extension that i find useful is the following: template<class II, class OI, class PRED> std::pair<II,OI> copy_if(II ibegin, II iend, OI obegin, OI oend, PRED p) { for(; ibegin != iend; ++ibegin) { if(p(*ibegin)) { *obegin = *ibegin; if(++obegin == oend) break; } } return std::make_pair(ibegin, obegin); } Here we have a finite sized sink. Items are copied from the source if they satisfy the condition and if there is room left in the sink. The positions reached are returned on termination of the algorithm. Witz

2 1

Documentation links don't
by ygor＠comcast.net 13 Sep '03

13 Sep '03

The links for PDF/HTML documentation do not work. Thanks.

4 4

RE: [Boost-users] Re: tupl::get bag
by Ben Hutchings 12 Sep '03

12 Sep '03

nagual <nagual(a)promsoft.ru> wrote: <snip> > I correct line 12 & 13: > > tie(i1, i2, i3, c1, c2, c3).template get<num>() = t; > cout<<tie(i1, i2, i3, c1, c2, c3).template get<num>()<<endl; > > g++ (GCC) v 3.2.2 now compile ok. > > But cl (MSVC) v 13.00.9466 output not change. ;-c The problem there must be something else. VC++ 7.1 accepts the code with or without "template".

1 0

copy_if
by Mr Zenaan Harkness 12 Sep '03

12 Sep '03

Does boost have an implementation of copy_if that I can't find yet? Just wondering, because it is an algorithm that, according to both stroustrup and o'rielly, should have been part of the standard but was essentially accidentally forgotten. Surely this (and possibly other necessary additions) would be part of an boost algorithm library, even if only one or two functions... I need copy_if, both these books I have say it's one of "the first generic algorithm that most programmers will probably write", so where is it? ??? (Yes, I can type it in, but boost is supposed to fill in the blanks, yes?) tia zen

1 0

boost::char_separator enhancement (or, how do I do it myself)
by Mr Zenaan Harkness 12 Sep '03

12 Sep '03

I'm teaching myself C++, and want to tokenize my stream, using boost::tokenizer. I want to be able to optionally tokenize with or without including the delimiter character - this is supported. However: the delimiter character, if included, becomes its own token. I want an option to have the delimiter character be part of the previous token (which is what makes sense for the processing I am doing on my files). ??? ta zen

1 0

Boost's pool implementation
by Thomas Harning 11 Sep '03

11 Sep '03

First off, I think the Boost pool is a great tool and for me has been one of the most useful of Boost's libraries. I have run into a problem however, due to the pool's usage of "malloc" and "free" as the names of functions. The problem is that I want to debug memory usage in my program and catch memory leaks, but using the leak detection in Visual Studio 7.1 for the CRT creates a macro for malloc and free, which conflicts with the pool's malloc and free and makes a general mess. Is there any better memory debug library/tools out there that are free and can get the line/file that the memory allocation occurred and the size perhaps and possibly a portion of the contained data as a character string? Perhaps I could write up a patch to rename the malloc and free from the boost pool [though I see that would be quite a pain to the existing users of boost's pool since code would have to be rewritten with whatever function is added]. Thanks. Thomas Harning Jr.

1 0

boost::pool base class new override
by Rodrigo Serafim 11 Sep '03

11 Sep '03

hi first question: where on earth was boost while I was doing my last year project? just thinking of the amount of trouble I could have saved if I knew about it... makes me bang my head against several walls ;) Well, I starter looking towards memory managment with other eyes after 90% of my bugs being directly originated form it. So after looking on the boost memory library a question popped: how can I use smart pointers and pool toguether. In particular intrusive pointers and/or pool, being the "and/or" essencial. If i have class C all made and ready and now I want it work with smart pointers i just have to make a base class RefCount, use multiple inheritance to contruct C_RC and i can have both things. Problem is, now i want to use it in pool too, but my base class RefCount calls delete. My solution; make yet another base class Poolable and override new and delete there. Well to my surprise it worked but only in very special cases. I have to use pool<> and cant use object_pool<C_RC_P> it crashes. Also it only works in MSVC6 in Release, in Debug it crashes. In a nutshell, my questions: - Is object_pool->malloc / free diferent in semantics from pool->malloc / free ? - I dont mind using pool<> but doing pool<>(sizeof(C_RC_P)) isnt that typesafe, can I make it? - Why does it crash in debug ? I send the code in attatchement. sorry abount the lengthy mail, too many question, too little time Thanks, Rodrigo Serafim

1 0

RE: [Boost-users] tupl::get bag
by Ben Hutchings 11 Sep '03

11 Sep '03

nagual <nagual(a)promsoft.ru> wrote: > I write example code for tuples::tuple::get use tuples::tie. > If compile icl (Intel C++) v 7.1 or bcc32 (Borland) v 5.6.4 all ok, > but if compile g++ (GCC) v 3.2.2 or cl (MSVC) v 13.00.9466 - > i have eror. > ;-c > I need to use it in MSVC. > Help me! <snip> Older compilers performed only minimal parsing of templates before they were instantiated, but the standard says they must do more parsing. This requires that you tell them which dependent names (names of things that depend on a template parameter) are being used as type or template names. You need to put the keyword "template" in front of "get<num>" so that the compiler can parse it correctly. Ben.

2 2

2D Shared Array
by Kevin J. Menard, Jr. 11 Sep '03

11 Sep '03

Hi, Is it possible to build a shared_array of shared_arrays? I would imagine there must be away, but I haven't been so successful at my attempts to find it :-/ I'm working on converting an existing project over to the boost smart pointer library. We've been using dynamically allocated 2D arrays to hold points in a matrix, and would like to take advantage of the automatic clean-up that the boost smart pointers can provide. That's why I would like to make a shared_array of shared_arrays of some data type. Any help would be appreciated. And if I'm approaching this all wrong, please let me know. I'm open to suggestions for other ideas, so long as the individual points can still be referenced in the form of: point[row][col]. Thanks, Kevin

1 0