May 2007 - Boost-users - lists.preview.boost.org

Specifying library locations for g++
by Scott Meyers 31 May '07

31 May '07

I've built Boost 1.34 for VC8 and g++ 4.1.1 on WinXP. I'm testing my installation with a simple program that uses Boost.FileSystem and Boost.Regex. Compilation is no problem, and for VC8, all I have to do to get things to link is add the location of the Boost binaries to the LIB environment variable. I was hoping that all I had to do for g++ was do the same for the LIBRARY_PATH environment variable (per http://gcc.gnu.org/onlinedocs/gcc/Environment-Variables.html) but I get unresolved symbol errors during linking even after doing that. I played around with the -l option (per http://gcc.gnu.org/onlinedocs/gcc/Link-Options.html#Link-Options) but nothing I tried worked. I have the following relevant binaries in my Boost binary directory: libboost_regex-mgw41-1_34.a libboost_regex-mgw41-d-1_34.a libboost_regex-mgw41-mt-1_34.a libboost_regex-mgw41-mt-d-1_34.a libboost_regex-mgw41-mt-s-1_34.a libboost_regex-mgw41-mt-sd-1_34.a libboost_regex-mgw41-s-1_34.a libboost_regex-mgw41-sd-1_34.a boost_filesystem-mgw41-1_34.a boost_filesystem-mgw41-d-1_34.a boost_filesystem-mgw41-mt-1_34.a boost_filesystem-mgw41-mt-d-1_34.a libboost_filesystem-mgw41-1_34.a libboost_filesystem-mgw41-d-1_34.a libboost_filesystem-mgw41-mt-1_34.a libboost_filesystem-mgw41-mt-d-1_34.a libboost_filesystem-mgw41-mt-s-1_34.a libboost_filesystem-mgw41-mt-sd-1_34.a libboost_filesystem-mgw41-s-1_34.a libboost_filesystem-mgw41-sd-1_34.a Can somebody explain or point me to an explanation of what I need to do to get a program using Boost binaries to link? I'm hoping there's something as simple as with VC8 where I can just specify a directory to look in and have the linker magically figure everything out from there. I'd like to avoid explicitly having to add a new library to link against each time I start using a new compiled Boost library in a project. Incidentally, in the above list of files, I notice that for Boost.FileSystem, there are some archives that start with lib and some that do not, while for regex, all start with lib. Is this meaningful? Thanks, Scott

9 23

Getting a "direct" iterator from Pointer Container
by Lost 30 May '07

30 May '07

Hi I'm using boost pointer container. Is it possible to get a "direct" iterator from Pointer Container? That's the iterator get form a normal stl iterator that points to the pointer? Thanks.

2 1

[UUID] Review results
by Hartmut Kaiser 30 May '07

30 May '07

Hi all, now, as Andy has finished his implementation (here: http://boost-consulting.com/vault/, file: uuid_v8.zip) we can move forward with the results of the review. This review has been kind of special in the sense that - it revealed some serious problems with the library to review (a license incompatibility problem and a security related problem around the random number generator scheme used), - it also spawned lots of dicussions on how to solve these problems. In the end we reached full agreement on what has to be done and Andy agreed to implement the changes. Normally, the mentioned problems would have prevented the acceptance of the library into Boost. OTOH, I think the following facts warrant not to reject the library in the first place: - The issues have already been discussed and a consensus has been reached on how to resolve the problems - Almost all reviewers found the proposed library to be incredibly useful and wanted to see the library accepted given the problems will be resolved - Andy fixed all of the significant problems in a short timeframe The final decision about the acceptance of the library will be made after a mini review of the proposed library (to be scheduled), which should concentrate on the new code presented by Andy and whether the mentioned problems have been appropriately fixed. Regards Hartmut Review Manager

1 0

compiler warnings in boost::serialization
by Os 30 May '07

30 May '07

Hi, I get a huge number of compiler warnings from the serialization and archive libraries. (This is boost 1.34.0 and Visual Studio 2005.) While I can turn them off, I'd rather not have to. The problem ones are: C4100 unreferenced formal parameter C4244 conversion from '...' to '...', possible loss of data C4265 class has virtual functions, but destructor is not virtual and new to 1.34.0 C4189 local variable is initialized but not referenced Can anything be done about this? cheers, os. -- os(a)collective.co.uk http://www.collective.co.uk/ http://www.myspace.com/darkroomtheband

1 0

Re: [Boost-users] [Boost.Test] Embedding unit testing in DLL projects
by Gennadiy Rozental 30 May '07

30 May '07

> "Tanner Davis" <tanner.davis(a)innovision.com> wrote in message > news:6704.32919073409$1180477224@news.gmane.org... > Hello- Hi, Tanner > What is the best practice for embedding test cases into library (versus > executable) projects? I have successfully used the Boost.Test framework > to build and > create EXE targeted, testing specific (MSVS) projects. However, much of > what I need to unit-test is compiled into a Win32 DLLs, and I would much > rather > embed the test cases directly within that DLL project code > (inside define blocks of course), rather than having to define a separate > project to produce an EXE > to execute the tests for each DLL. When I say "I would rather", it > really means my manager demands it. It appears that the boost.test MACROs > nicely > obfuscate the normal main()-ish entry point for the testing. I have > searched for information on overriding or extending the kick-off of > whatever calls the > init_unit_test_suite() so I could write my own custom entry points for > testing the DLL as a post-build step, but have not found what I am looking > for. Boost 1.34.0 introduced new Boost.Test variant usage specifically for this case. It's called "external test runner usage variant". The idea is to build your test module as DLL (define BOOST_TEST_DYN_LINK during compilation). Now you can run the test modult using any external test runner tool. Boost.Test supply simple command line test runner: boost_test_runner utility (it's located in libs/test). What it does essenciallty is it loads DLL dynamically, init the framework and test tree and run all the registerred test cases. > So, my question is this. Is there a provided framework for testing code > within a DLL project? And if not, is there a way to kick start the > testing from a non- > main method (say a custom DLL entry point), and would this strategy work > within the confines of Boost.Test? I believe that other well known unit > testing > framework (which shall remain nameless) supports such a thing, and I would > suspect Boost.Test does also. > > I apologize if this question is often asked or obviously documented > elsewhere. > Thank you for the assistance. The upcoming docs will cover this usage vasriants in more details. HTH, Gennadiy

1 0

[regex] ECMAScript conformance
by Armel Asselin 30 May '07

30 May '07

[boost 1.34.0, with last update to regex sources on 28th february 2007] Hello, first of all I must tell that I am very happy with boost.regex, it's fantastic. though, there are a few inconsistencies though with the ECMAScript spec. 1. it seems that re-matching an atom does not first clear its sub-matches, as "step 4" in page 135 of ECMA 262 3rd edition tells to. that is, it does what is in the 'and not' part here: /(z)((a+)?(b+)?(c))*/.exec("zaacbbbcac") which returns the array ["zaacbbbcac", "z", "ac", "a", undefined, "c"] and not ["zaacbbbcac", "z", "ac", "a", "bbb", "c"] 2. the {n,m} notation should not accept spaces in ECMAScript mode (though it's practical, it's not conformant). I easily admit that its not really important. Best regards Armel

3 2

problem with building program_options (VC8)
by Sergey Skorniakov 30 May '07

30 May '07

I have build boost libraries with command line bjam --toolset=msvc --user-config=config32.jam --stagedir=. debug-symbols=on stage File config32.jam is: using msvc : 8.0 : : <cxxflags>"-wd4996 -wd4103 -Zp4" <define>_CRT_SECURE_NO_DEPRECATE <define>_CRT_NONSTDC_NO_DEPRECATE <define>_SECURE_SCL=0 <define>BOOST_SPIRIT_THREADSAFE <define>BOOST_THREAD_USE_DLL <define>BOOST_THREAD_BUILD_DLL ; The resulting libboost_program_options-vc80-mt-1_34.lib has size 7,420,264 bytes and my program, builded with /D "_CRT_SECURE_NO_DEPRECATE" /D "_CRT_NONSTDC_NO_DEPRECATE" /D "_SECURE_SCL=0" /D "BOOST_THREAD_USE_DLL", crashes. When I changed config32.jam to using msvc : 8.0 : : <cxxflags>"-wd4996 -wd4103 -Zp4 -D_CRT_SECURE_NO_DEPRECATE -D_CRT_NONSTDC_NO_DEPRECATE -D_SECURE_SCL=0 -DBOOST_SPIRIT_THREADSAFE -DBOOST_THREAD_USE_DLL -DBOOST_THREAD_BUILD_DLL " ; The resulting libboost_program_options-vc80-mt-1_34.lib size is 7,233,568 bytes and my program works well. This looks like <define> in config file sometimes has no effect. Furthermore, program options library is alignment-sensitive.

2 2

callbacks to function template (solved)
by Ares Lagae 30 May '07

30 May '07

With the help of the boost lists I finally solved the problem of callbacks to a function template. I will describe what exactly the problem was, and how I have solved it, in the hope that it will be usefull for somebody. I am also open for suggestions and comments. Problem description ------------------- I am building a set of parsers. The parser classes pass data to the application using callbacks. Because there are potentially a lot of callbacks, and some of the callbacks are optional, I wrap each callback into a std::tr1::function object, and then pass it to the parser. The problem I came across can be simplified and abstracted to parsing a format like this one. bool 1 char 2 short 3 int 4 long 5 float 6.0 double 7.0 Each line contains a type and is followed by a value of that type. This data should be parsed and passed to the application. The types are fixed (bool, char, short, int, long, float, double). (In the actual real problem, there are more types, and "declarations" involving several types.) Solution 1 ---------- The first solution (see listing_1.cpp) registers a callback for each data type. Because each data type is handled in a similar way, there is a lot of code repetition. Solution 2 ---------- The second solution (see listing_2.cpp and variable_list.hpp) replaces the callback function by a function template. Instead of having a member variable for the callback for each data type, the member variables are grouped into a "variable_list". This is a class template that, for each type in a type list, keeps a variable. The code repetition in the parser class is eliminated, but the user still has to call "set_callback" to register a callback for each of the data types. Solution 3 ---------- The third solution (see listing_3.cpp) eliminates the similar code at the user side. Instead of calling "set_callback" to register a callback for each of the data types, a functor is constructed, that for each type constructs a std::tr1::function object that wraps the callback. That function object is passed to the parser, and the parser sets all the callbacks. This again results in similar code in the parser "set_callbacks". Solution 4 ---------- The fourth solution (see listing_4.cpp) eliminates the similar code in the parser "set_callbacks" by executing a functor for each type in the type list. This code is approximately equivalent to callbacks to a function template. Best regards, -- Ares Lagae Computer Graphics Research Group, Katholieke Universiteit Leuven http://www.cs.kuleuven.be/~ares/

2 2

filesystem/convenience.hpp build warning
by Os 30 May '07

30 May '07

Having just moved to v1.34.0, I now get a build warning from this file. boost/filesystem/convenience.hpp:46: warning C4245: 'argument' : conversion from 'int' to 'boost::filesystem::system_error_type', signed/unsigned mismatch Is this a bug? The line in question is boost::throw_exception( basic_filesystem_error<Path>( "boost::filesystem::create_directories", ph, -1 ) ); Changing it to boost::throw_exception( basic_filesystem_error<Path>( "boost::filesystem::create_directories", ph, (boost::filesystem::system_error_type)-1 ) ); seems to fix it. cheers, os. -- os(a)collective.co.uk http://www.collective.co.uk/ http://www.myspace.com/darkroomtheband

1 0

Why even use Boost preprocessor library over my own code generator ?
by Keriask 30 May '07

30 May '07

Hi. I find the 2 pros given in the docs to use the preprocessor library not valid! I was about to use the preprocessor library for the 1st time but after seing and studying some of the preprocessed output from various compilers it killed the only apparently valid argument the preprocessor docs present themselves to use it: integration and read to run. Obviously integration can be made easily with the #line and #error directives, wich will even allow my custom code generator to have syntax error checking. The other argument is not really true since a simple text processor can be made portable very easily, hey Linux does it with much more complicated stuff and it works most times. BTW, I'm not trying to bash it by any means(the work and effort put there are amazing), I just want to find a good reason to use it myself which unfortunately I haven't so far. Best regs Jose ____________________________________________________________________________________Be a better Globetrotter. Get better travel answers from someone who knows. Yahoo! Answers - Check it out. http://answers.yahoo.com/dir/?link=list&sid=396545469

8 12