December 2005 - Boost-users - lists.preview.boost.org

Compiling Archive Output Code
by Merrill Cornish 16 Dec '05

16 Dec '05

I am using Boost 1.33.0 with MinGW 3.4.2. I plan to use the Boost serialization library in a project I am working on. Following the tutorial (but allowing for XML archives rather than text) I have successfully included code like the following in my classes: #include <fstream> #include "boost/archive/xml_oarchive.hpp" #include "boost/archive/xml_iarchive.hpp" class ClassWhatever { private: friend class boost::serialization::access; template<class Archive> void serialize(Archive& ar, const unsigned int version) { ar & mResults; ar & mInputs; ar & mDisplayObject; }//serialize() . . . }; This code compiles without error in all classes. Now, I'm writing unit tests for the classes and I want to trigger the save and later reload of these classes. Following the same serialization tutorial, I tried this: . . . { ofstream outStream(archiveFile); boost::archive::xml_oarchive outArchive(outStream); outArchive << testObj_out; } // archive and stream closed when destructors are called . . . and similar code for the later load from the archive. When I compile this, I get massive errors, a few of which are shown below. (I've manually wrapped the long lines in an attempt to make it more readable--but I may have just made it worse.) I'm sort of guessing that there is a header or two missing. Any suggestions? Also, there are "polymorphic_..." versions of the archive files available, but I don't remember seeing them mentioned in the documentation. I have polymorphic classes. Should I be using the polymorphic versions of the headers? Merrill ======================================== E:/Dev-Cpp/Workflow/boost/include/boost/archive/basic_xml_oarchive.hpp: In member function `void boost::archive::basic_xml_oarchive<Archive>:: save_override(T&, int) [with T = boost::shared_ptr<WkFlo::Successor>, Archive = boost::archive::xml_oarchive]': E:/Dev-Cpp/Workflow/boost/include/boost/archive/detail/interface_oarchive.hpp:85: instantiated from `Archive& boost::archive::detail::interface_oarchive<Archive>:: operator<<(T&) [with T = WkFlo::Successor_ptr, Archive = boost::archive::xml_oarchive]' E:/Dev-Cpp/Workflow/test/src/Successor_Test.cc:111: instantiated from here E:/Dev-Cpp/Workflow/boost/include/boost/archive/basic_xml_oarchive.hpp:86: error: incomplete type `boost::STATIC_ASSERTION_FAILURE< false>' used in nested name specifier E:/Dev-Cpp/Workflow/boost/include/boost/archive/basic_xml_oarchive.hpp:86: error: size of array has non-integral type `<type error>' E:/Dev-Cpp/Workflow/boost/include/boost/archive/basic_xml_iarchive.hpp: In member function `void boost::archive::basic_xml_iarchive<Archive>:: load_override(T&, int) [with T = boost::shared_ptr<WkFlo::Successor>, Archive = boost::archive::xml_iarchive]': E:/Dev-Cpp/Workflow/boost/include/boost/archive/xml_iarchive.hpp:74: instantiated from `void boost::archive::xml_iarchive_impl<Archive>:: load_override(T&, int) [with T = boost::shared_ptr<WkFlo::Successor>, Archive = boost::archive::xml_iarchive]' E:/Dev-Cpp/Workflow/boost/include/boost/archive/detail/interface_iarchive.hpp:84: instantiated from `Archive& boost::archive::detail::interface_iarchive<Archive>:: operator>>(T&) [with T = WkFlo::Successor_ptr, Archive = boost::archive::xml_iarchive]' E:/Dev-Cpp/Workflow/test/src/Successor_Test.cc:121: instantiated from here . . . and so on

2 1

[BGL] Remove all vertices from graph with zero degree
by Dmitry Bufistov 15 Dec '05

15 Dec '05

Hi all again! I would really appreciate any ideas how to implement (subject) efficiently with bearing in mind "Iterator" stability. Graph representation is the following: typedef adjacency_list< listS, ///<Store out-edges of each vertex in a std::list vecS, ///<Store vertex set in std::vector bidirectionalS, ///<Graph is directed but with access to in & out edges VertexProperties, EdgeProperties, GraphProperties > Graph; filtered_graph seems isn't what I need. I would like to right as follow Graph g; remove_empty_vertices(g); //Now g doesn't contain zero degree vertices Many thanks in advance, regards, --dima

3 2

Boost binary serialization & AMD64
by Leon Mergen 15 Dec '05

15 Dec '05

Hello, I'm currently having problems using serializations that were stored binary on an x86 and loading them on an amd64. Anyone knows whether there are any compatibility issues between those architectures and storing things binary ? Thanks in advance. Regards, Leon Mergen

1 0

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