September 2004 - Boost-users

statically linking to Boost.Threads with msvc7
by Simmons, Aaron 03 Sep '04

03 Sep '04

I've built and installed boost on a WinXp system with VS.Net 2003. I'm trying to use it in a static CRT project. I get the following error whenever I compile: > error C1189: #error : "Mixing a dll boost library with > a static runtime is a really bad idea..." I noticed that this is happening because BOOST_DYN_LINK is #defined in config.hpp. Apparently dynamic linking is the default for Boost? Anyway, I've commented out this #define, but now I get a ton of warnings like the following > warning C4275: non dll-interface class 'std::logic_error' > used as base for dll-interface class 'boost::lock_error' and the project won't link > error LNK1104: cannot open file 'libboost_thread-vc71-mt-s-1_31.lib' I looked in Boost's lib folder and there aren't any static thread binaries (eg, "libboost_thread_xxx.lib). There are only the dynamic libraries (boost_thread_xxx.lib). Apparently something failed in the build process. I have to admit I didn't sit and watch the build process for the half-hour it took to build, so something may have gone wrong. Is there a build log somewhere? Is it possible to have bjam only build Boost.Threads so that I can see if there's an error without having to rebuild the entire library? Thanks, Aaron Simmons

2 1

boost::signals with Qt
by Alain Cormier 03 Sep '04

03 Sep '04

Hi, Just want to know *exactly* how to use boost::signals in a Qt project. I read this : http://www.boost.org/doc/html/ch06s04.html but I don't understand how to do it. I try to recompile boost with : "-sBUILD=<define>BOOST_SIGNALS_NAMESPACE=signalslib". After that, I try to directly change BOOST_SIGNALS_NAMESPACE in signals_common.hpp but didn't work neither. I use gcc 3.3.2 under Linux Mandrake 10 Thank you for any suggestion Al

3 3

RE: [Boost-users] statically linking to Boost.Threads with msvc7
by Ben Hutchings 03 Sep '04

03 Sep '04

Aaron Simmons <a-simmons1(a)ti.com> wrote: I've built and installed boost on a WinXp system with VS.Net 2003. I'm trying to use it in a static CRT project. I get the following error whenever I compile: > error C1189: #error : "Mixing a dll boost library with a static runtime is a really bad idea..." I noticed that this is happening because BOOST_DYN_LINK is #defined in config.hpp. Apparently dynamic linking is the default for Boost? Windows does not support registration of clean-up functions for thread-specific storage (TSS, or TLS), but it does notify DLLs when a thread exits and this gives DLLs a chance to clean-up their thread-specific storage. So in order to make TSS work correctly, Boost.Threads is designed to be built only as a DLL on Windows. This means, as the error message says, that you need to use the DLL version of the CRT. Ben.

1 0

nullary functor returning predefined value
by Yuri Muhitov 03 Sep '04

03 Sep '04

Hello All ! Please tell, can I create nullary functor returning predefined value, using boost::function & boost::bind ? Example uses "portable" syntax of boost::function : #include <boost/bind.hpp> #include <boost/function.hpp> struct InterfaceA; InterfaceA *getInterface (); template<typename T> T forward ( T t ) { return t; } int main ( int, char ** ) { typedef boost::function0<InterfaceA *,void> functor_type; // Value of InterfaceA * supplied by getInterface () function. Usage is straightforward : functor_type f1 = boost::bind ( getInterface ); InterfaceA *p = NULL; // Value of InterfaceA * is known here. How to create functor, returning it ? // functor_type f2 = boost::bind ( ??? ); // Possible workaround. Can I do this without forward<> template function ? functor_type f3 = boost::bind ( &forward<InterfaceA *>, p ); return 0; } Thanks! Yuri.

1 0

newbie: bjam with mwcw, #include "user" and <system> paths
by Conrad Weyns 03 Sep '04

03 Sep '04

Hi, I am a newbie and have a few questions. 1. Re: bjam suspicious output with Metrowerks CodeWarrior 9.2 My command line tools times goes back to mpw on the Mac pre 1990... The boost documentation looks very good but it still assumes a knowledge of command line tools. I have managed to build 2 Win targets: One that I hope I will be able to drop into a large ms vc7.2 solution and an other for a mw cw 9.2 workspace. (I sure wish boost deployed ready made .vcproj and .mcp files...) I have extracted the boost source to: C:\boost_1_31_0 I have also downloaded bjam.exe (3.1.10-1-ntc86) and copied it to the boost_1_31_0 root. So, I launched a cmd window: cd C:\boost_1_31_0 bjam "-sVC71_ROOT=C:\Program Files\Microsoft Visual Studio .NET 2003\Vc7" install. This generates: C:\Boost\ - include\ - lib\ include\ is basically a copy of the source without a few make specific files but with all the platform specific source intact. I assume this is a "basic installation" for use with ms vc7. I have managed to use it making a few tests with boost::shared_ptr. So far so good. I proceeded to make the mwcw target after renaming C:\Boost\ to C:Boost_vc7\. bjam "-sCWPRO8_ROOT=C:\Program Files\Metrowerks\CodeWarrior" install This process is much much faster and the new files I get in C:\Boost\include and C:Boost\lib are exactly the same as those generated a while ago by the vc7 compiler (the libs have exactly the same timestamp). In fact it doesn't look as if anything is ever compiled. This is making me suspicious. However I have managed to make a similar test with boost::shared_ptr. Off course, none of the generated libraries are involved so I am probably just kidding myself. 2. The boost source is mixing "user" and <system> paths. For instance in boost\any.hpp: #include "boost/config.hpp" #include <boost/throw_exception.hpp> Is this intentional? If so, why? It is forcing me to add the boost root to both the UserPaths and System Paths settings. This is very unfortunate. I can"clean this up" but that's going to cause work when upgrading. 3. Does anyone here have experience adding boost to a large vc7.1 solution in such a way as to cause zero noise. In other words, next time your collegues synchronize, everything should be automagic - no need to have any one change local settings or read long and boring installation e-mails which is bound to fail anyway. Mirroring the boost hiearchy in a CodeWarrior project is child's play. In msvc it's more like a nightmare because one has to create the sub directories manually, one at a time :-). I can deal with the #include paths issues by adding ".." to projects Additional Include Directories". Msvc does not seem to care whether it's a "user" or <system> path. I'd be interested to know how people deal deal with this in general. regards, Conrad Weyns

1 0

question about ublas
by yanyaqin 03 Sep '04

03 Sep '04

2 1

Re:[Boost-users] Re: There's someone using boost lu_substitute?
by yanyaqin 03 Sep '04

03 Sep '04

2 1

Does ublas::lu_* use the character of sparse_matrix/compressed_matrix?
by yanyaqin 03 Sep '04

03 Sep '04

1 0

exec_mon_example.cpp
by Steven T. Hatton 02 Sep '04

02 Sep '04

On SuSE Linux 9.1 using both my own build of boost 1.31, and SuSE provided rpms, as well as both my own build of gcc 3.3.3 and the one from SuSE, I was not able to build this file found in the boost source tree boost_1_31_0/libs/test/example/exec_mon_example.cpp It can also be found here: http://www.boost.org/libs/test/example/exec_mon_example.cpp This is what happens: $ g++ -o exec exec_mon_example.cpp /tmp/ccGtnzc0.o(.text+0x107): In function `main': : undefined reference to `boost::execution_monitor::execute(bool, int)' collect2: ld returned 1 exit status I've also tried variation of the following: g++ -o exec exec_mon_example.cpp -L /org/boost/lib -l \ boost_test_exec_monitor-gcc /opt/org/boost/lib/libboost_test_exec_monitor-gcc.so: undefined reference to `test_main(int, char**)' collect2: ld returned 1 exit status Does this compile for other Linux users? Any suggestions? -- Regards, Steven

1 0

boost::bind questions
by Matt S Trentini 02 Sep '04

02 Sep '04

Heya Boosters, I ran into a problem using boost::bind yesterday and, although I found a workaround, I've got a few questions. For the record I'm using boost 1.30. I stumbled onto the issue that boost::bind doesn't define the argument_type typedef (I was trying to use std::not1). After trawling the web I found a discussion, and useful workaround, about this here: http://tinyurl.com/5bl9m Questions that arose from this: o Why doesn't boost::bind support argument_type? o What other standard library function adapters does this affect? - Is it all that are derived from unary_function? o Has the operator! addition been incorporated into later versions of boost? - If not, why not? Thanks! Matt

4 6