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

Re: [Boost-users] boost builds dll's on mingw, can't use
by Merrill Cornish 15 Dec '05

15 Dec '05

Imran, >>> .....mingw32\bin\ld.exe: cannot find -lboost_thread OK, I see two problems here. First, are you calling ld.exe directly to link your executable, or are you calling either gcc or g++? I initially made the mistake of calling ld.exe directly, but found the linker couldn't even find the C++ functions in std namespace. I was told to use gcc or g++ because they know to automatically include the standard runtime libraries. Second, -lboost_thread would work only if you have a -L<directory> where <directory> is where the Boost libraries are AND the Boost library file was named "libboost_thread.dll" or "libboost_thread.lib". That is, the -l option only takes care of the "lib" prefix and the ".dll" or ".lib" file extension, depending on whether you are linking to shared or static libraries. When Boost builds its libraries, it add a suffix to the library name that records the major build options (which compiler, static? debug? multi-thread? etc.). So, I'm betting that in your Boost library directory you do NOT have a file named libboost_thread.dll or libboost_thread.lib. You need to include that filename suffix (and I'm talking about the filename and not the file extension) in the argument to the -l option. What is the full filename of the boost_thread library in the Boost libs directory? Merrill

1 0

Re: [Boost-users] boost builds dll's on mingw, can't use
by Merrill Cornish 15 Dec '05

15 Dec '05

Imran, >>>Is the problem not the extensions (.dll/.lib vs .so/.a)? >>>Why didn't boost name them properly when I specified mingw? .dll is for shared libraries on Windows. .lib is for static libraries on Windows. .so and .a are for shared/static on Unix/Linux. It shouldn't matter whether you are using GNU make or not. MinGW uses g++ to do the linking, and it's g++ that has the file ordering requirement on the command line. On the other hand, if you aren't using g++ (or gcc, which calls g++), then maybe your linker doesn't have the ordering problem. Is the error you are getting about unresolved references? Merrill

2 1

Re: [Boost-users] boost builds dll's on mingw, can't use
by Merrill Cornish 15 Dec '05

15 Dec '05

Imran, I've using Boost 1.33.0 with the latest MinGW on Windows XP, and it works fine. Could you be a bit more specific as to why MinGW "can't use" the Boost libraries? Do you mean that references are left unresolved? Here is an excerpt from the part of my my Makefile that deals with libraries: BOOSTLIBS = -L${boost_libs} \ -llibboost_date_time${boost_suffix} \ -llibboost_filesystem${boost_suffix} \ -llibboost_regex${boost_suffix} \ -llibboost_serialization${boost_suffix} \ -llibboost_thread${boost_suffix} \ -llibboost_program_options${boost_suffix} where ${boost_libs} expands into the pathname of the directory the Boost libraries are in and ${boost_suffix} expands into "-mgw-mt-sd-1_33", which says the libraries are compiled for MinGW, multi-threaded, static, debug, and Boost version 1.33. I put ${BOOSTLIBS} as the LAST thing on the call to g++ to link my .exe file. Notice that the positioning is important. The linker reads the files listed on the command line from left to right. When it comes to a library file, it resolves whatever outstanding references to that library it has AT THAT POINT in the scan. If it sees another reference to that library later in the left to right scan after seeing the library, that later reference goes unresolved. The linker needs to see ALL references to library functions BEFORE it sees the libraries themselves. So, put the libraries last. The Boost library build process creates the libraries both with the Boost version suffix ("1_33" in my case) and without. You can use either one as they are identical files. The idea is that if you use the library file versions WITHOUT the version number suffix, then you won't have to change your Makefile when a new version comes out. On the other hand, if you want to explicitly tie your program to one and only one version of the Boost libraries, you can use the filename WITH the version numbers. Merrill

2 1

vc 8 secure stl/runtime and boost
by Delfin Rojas 15 Dec '05

15 Dec '05

Hi there, Can anybody tell me if Boost.Build defines _SECURE_SCL as 0 when building is using the vc-8 toolset? If not, has anybody noticed any slowdown in boost libraries for using the new "secure" runtime? What about _HAS_ITERATOR_DEBUGGING in debug builds? Is it defined as 0 by Boost.Build when using the VC-8 toolset? I just spent a few minutes figuring out a compiler error with std::lower_bound that happened to be caused by the "Checked Iterators" feature, so I was wondering if boost disables this. Thanks Delfin Rojas

4 5

boost builds dll's on mingw, can't use
by Imran Akbar 15 Dec '05

15 Dec '05

Hi, I've been having problems linking to boost using mingw. I installed boost linking to mingw, but it installed the necessary boost-mgw files with either .dll or .lib extensions, which mingw can't use! I tried renaming the .lib's to .a's and the .dll's to .so's, which didn't work. then I tried adding a 'lib' prefix to each .a or .so, which still didn't work... there isn't actually a boost_thread library file, it's got some version number at the end - would that be a problem? Did I install it correctly? I'm linking to it using -lboost_thread in mingw (gcc). Any help much appreciated. Thanks, Imran

1 0

Seg fault with boost::any / derived system
by Stephen Gross 15 Dec '05

15 Dec '05

In the following snippet, I've got a simple inherited system (base and derived). I create a boost::any and assign it a derived instance. Then I try to any_cast it to the base. It compiles, but at runtime it segfaults. Is this expected behavior? Is there a way to make this work? Thanks! --Steve ===== CODE FOLLOWS ===== class base { public: virtual void go() const { ; } }; class foo1 : public base { virtual void go() const { std::cout << "foo1" << std::endl; } }; void test1() { foo1 f; boost::any a = f; boost::any_cast<base> (&a)->go(); } =========================

2 1

initializing dynamic_bitsets
by Dave Steffen 15 Dec '05

15 Dec '05

Folks, I used dynamic_bitset for the first time last night - nifty! I am curious, though, about how to construct a bitset of all 1's. Right now I'm doing this: dynamic_bitset foo(n); foo.set(); Is there a more efficient way to generate a bitset will all bits set? Thanks very much! ------------------------------------------------------------------------ Dave Steffen, Ph.D. "We warn the reader in advance that the proof Software Engineer IV presented here depends on a clever but highly Numerica Corporation unmotivated trick." ph (970) 419-8343 x27 fax (970) 223-6797 -- Howard Anton, "Elementary Linear Algebra" dgsteffen(a)numerica.us

2 2

another problem with crc.hpp
by Clarke, Trevor 15 Dec '05

15 Dec '05

I'm compiling some code that uses crc.hpp. On windows (VS2003) it works fine. On Solaris 8 (Sun Studio 8), I get the following error. (offending code included below) >> Assertion: unhandled expression in abi2_mangler::entity_expression (../lnk/v2mangler.cc, line 1584) while processing /.../boost/crc.hpp at line 57. If I delete lines 47-58, I get the same error on line 45. If I delete lines 41-59, I get the same error on line 39 which is a comment. A google search turns up a thread in sun's forums dealing with a bug in the Forte 7 compiler. However, that only showed up when compiling with -g so I removed the -g from my compile line and still get the error. Has anyone else seen this? 36 // workaround is to make up a dummy function argument that encodes the template 37 // arguments. Calls to such template functions need all their template 38 // arguments explicitly specified. At least one compiler that needs this 39 // workaround also needs the default value for the dummy argument to be 40 // specified in the definition. 41 #if defined(__GNUC__) || !defined(BOOST_NO_EXPLICIT_FUNCTION_TEMPLATE_ARGUMENTS) 42 #define BOOST_CRC_DUMMY_PARM_TYPE 43 #define BOOST_CRC_DUMMY_INIT 44 #define BOOST_ACRC_DUMMY_PARM_TYPE 45 #define BOOST_ACRC_DUMMY_INIT 46 #else 47 namespace boost { namespace detail { 48 template < std::size_t Bits, BOOST_CRC_PARM_TYPE TruncPoly, 49 BOOST_CRC_PARM_TYPE InitRem, BOOST_CRC_PARM_TYPE FinalXor, 50 bool ReflectIn, bool ReflectRem > 51 struct dummy_crc_argument { }; 52 } } 53 #define BOOST_CRC_DUMMY_PARM_TYPE , detail::dummy_crc_argument<Bits, \ 54 TruncPoly, InitRem, FinalXor, ReflectIn, ReflectRem> *p_ 55 #define BOOST_CRC_DUMMY_INIT BOOST_CRC_DUMMY_PARM_TYPE = 0 56 #define BOOST_CRC_DUMMY_PARM_TYPE 57 #define BOOST_ACRC_DUMMY_PARM_TYPE , detail::dummy_crc_argument<Bits, \ 58 TruncPoly, 0, 0, false, false> *p_ 59 #endif ------------------------------ Trevor R.H. Clarke tclarke at ball com <mailto:tclarke@ball.com> Ball Aerospace & Technologies Corp

1 0

[iostreams] Decompressing n lines at a time and parsing
by Mark Van De Vyver 15 Dec '05

15 Dec '05

Hi, I'd appreciate any advice on how to acheive the following using iostreams: 1) Read in part of a compressed file. 2) Decompress the chunk just read. 3) Pass this to boost::tokenizer (or boost::spirit?) 4) Parse the decompressed text, one line at a time. 1) Read in the next chunk of compressed data. The docs don't seem to have this as an example, and I'm a novice at programming so am a little wary of trying something 'off-the-cuff'. It seems 4) can be a little sticky - when the chunk of compressed data does not end at the 'end of a line' in the decompressed data.... I'd rather spend some time becoming familar with iostreams - it has more 'head-room'. Nonethelss, this may not be what iostreams was designed for, so will it be more effective to use the the zlib library, using gzgets()? Finally, looking at some posts it seems that if using boost::iostreams the correct buffer is an array_source and array_sink? Would appreciate any suggestions. Regards Mark

2 2

Help ! invalid wide-character constant error in regex test_locale.cpp (WinXp CHT)
by Sam Gw 15 Dec '05

15 Dec '05

Hi there, I got some problem when try to compile unit testing source code. When I try to compile test_locale.cpp , I got compile error message like below: \libs\regex\test\regress\test_locale.cpp(133) : error C2002: invalid wide-character constant I guess the problem relates with my Tradition Chinese System & Unicode. But I don't know how to solve. And when I try to define BOOST_NO_WREGEX, it success. (I could compile and execute example -> credit_card_example.cpp success. ) operation system : WinXP professional Tradition Chinese SP2. Compiler : Vistual C++ sp6 + stlport 5.0 + platform SDK Feb 2003 Line below is source code: TEST_REGEX_SEARCH_L("[[:lower:]]+", perl, "\xde\xdf\xe0\xe1\xe2\xe3\xe4\xe5\xe6\xe7\xe8\xe9\xea\xeb\xec\xed\xee\xef\xf0\xf1\xf2\xf3\xf4\xf5\xf6\xf8\xf9\xfa\xfb\xfc\xfd\xfe\xf7", match_default, make_array(1, 32, -2, -2)); My compile command: E:\boost_1_33_1\libs\regex>cl /MD /Zm400 /GF /Gy -GX -GR /Oityb1 E:\boost_1_33_1\libs\regex\test\regress\test_locale.cpp -link /LIBPATH:..\..\stage\lib /LIBPATH:E:\boost_1_33_1\bin\boost\libs\regex\build\libboost_regex.lib\msvc-stlport\release\threading-multi\ /LIBPATH:%STLPORT_PATH%\lib user32.lib

2 1