- Boost-users - lists.boost.org

Issues on getting the reference of the held value of boost::any
by Seishi Onishi <looper2k＠yahoo.com> 17 Dec '02

17 Dec '02

I'm using a collection of boost::any for the variable container class which has this map member. std::map<std::string, boost::any> m_collection; As you imagine, the container class can keep anything even the container class itself recursively. You know, an boost::any may hold copy expensive object. So I need a method to get the reference of the value held in boost::any not to make the expensive copy. But I found that the current any_cast implementation was not well support the reference. Issue 1: The following code causes compiler error on VC6&7. boost::any dany = double(1.2); double& rd = boost::any_cast<double>(dany); Issue 2: The following code works but one copy constructer called. class Foo foo; boost::any fany( foo ); class Foo& rfoo = boost::any_cast<class Foo>(fany); Actually, I found the way to sneak out of the issues. double& rd = *boost::any_cast<double>(&dany); class Foo& rfoo = *boost::any_cast<class Foo>(&fany); But the problem is these codes won't throw bad_any_cast. And I also wonder this *any_cast usage may not be the way the developer supposed. So I expect that the current any_cast code will be changed to well support the reference access. It can be, template<typename ValueType> const ValueType& any_cast(const any & operand) { const ValueType * result = any_cast<ValueType>(&operand); if(!result) throw bad_any_cast(); return *result; } template<typename ValueType> ValueType& any_cast(any & operand) { ValueType * result = any_cast<ValueType>(&operand); if(!result) throw bad_any_cast(); return *result; } -Seishi

1 0

[Boost-Users][BGL] Permutation of a graph
by Paolo Fosser 17 Dec '02

17 Dec '02

Hi all, I'm using BGL in some graph matching algorithms. Now I have to shuffle a bit one of the graphs I'm comparing. I think it all ends up in considering the nodes in a different order when cycling on them. (BTW am I right?). So my question: is there an elegant way to do this in BGL? The only solution that comes into my mind is to build up a map from the range of vertex descriptors to a permutation of them. TIA, Paolo -- Do you want to live forever? Alex Chiu has invented a device which will give you physical immortality. Click here! http://www.alexchiu.com/affiliates/clickthru.cgi?id=pfosser

2 3

testing if shared_ptr has been assigned
by TEA Hartmann, Steven 16 Dec '02

16 Dec '02

I need some help in testing for the condition of whether a shared_ptr has been assigned. In the following example, the "if( b != NULL ) does not work. How would I do this test? typedef boost::shared_ptr<int> IPtr; int i( 0 ); IPtr a( i ); IPtr b; if( some condition ) b = a; if( b != NULL ) { // How should this really be coded? } Best Regards, Steve [Non-text portions of this message have been removed]

5 4

Borland: Cannot generate template specialization
by Felix E. Klee 16 Dec '02

16 Dec '02

Hi, I'm trying to compile C++ code that uses the boost/numeric/ublas matrix library with the Kylix C++ Compiler (BC++ 5.7 for LINUX). However, I get several error messages, among them: Error E2299 boost_1_29_0/boost/numeric/ublas/matrix.hpp 522: Cannot generate template specialization from 'container_reference<C>' in function matrix<complex<double>,row_major,unbounded_array<complex<double> > >::iterator1::iterator1(matrix<complex<double>,row_major,unbounded_array<complex<double> > > &,complex<double> * const &) Any ideas how to work around this problem? Felix

3 4

Use vc7 toolset for VC7.1?
by Dick Bridges 16 Dec '02

16 Dec '02

Is there a VC7.1-specific toolset or should I continue to use the vc7 toolset? I'd like to avoid the VC6 and VC7.0 workarounds if possible.

2 1

Building problems with regex on a SuSE 8.1 installation
by beat.schaer＠ruag.com 15 Dec '02

15 Dec '02

Hi I tried to compile boost 1.29.0 on a SuSE 8.1 machine and got the following error: gcc-C++-action libs/regex/build/bin/libboost_regex.so/gcc/release/runtime-link-dynamic/shar ed-linkable-true/cregex.o /home/adtsbuilder/boost_1_29_0/boost/regex/detail/regex_match.hpp: In member function `unsigned int boost::RegEx::GrepFiles(bool (*)(const char*, const boost::RegEx&), const char*, bool, unsigned int)': /home/adtsbuilder/boost_1_29_0/boost/regex/detail/regex_match.hpp:1902: Internal compiler error in expand_call, at calls.c:3049 Please submit a full bug report, with preprocessed source if appropriate. See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions. The used compiler is the one delivered with SuSE 8.1. It is a gcc 3.2. Does anybody knows what the problem might be? Regards Beat

4 3

shared_ptr does not call destructor
by Ani Taggu 15 Dec '02

15 Dec '02

Hi: I am using windows 2000 with VC 6.0 and boost 1.28. In my project, I have an array of pointers to objects: class foo { public: ~foo() {} private: std::vector<boost::shared_ptr<MyType> > myArr_; }; I found that the destructor for MyType is never called. Printing the shared_ptr::use_count = 6 at ~foo() above. Any suggestions on what could be happening and how to approach the debugging? Thanks, -Ani

1 0

Test Library: Program Execution Monitor output
by klapshin <klapshin＠yahoo.com> 13 Dec '02

13 Dec '02

Hello, I just plugged program execution monitor in my code and there are certain things I don't really like. First of all is not it odd to see a contradictory messages like: --------------------------- no errors detected **** error return code 1 ********** errors detected; see standard output for details *********** --------------------------- I don't really need this "no errors detected" message. Looking into code I found a way to turn it off -- I have to define an environment variable. I have a way to control this behavior. This is good. Bad news though that the way to control is quite an awkward one. I would like to have some internal control, probably overwritable via environment variable. Another concern is that I have no control on return codes in case of exceptions and runtime errors. For instance in my company all programs have to return 1 in case of failure (and 0 on success). Other return codes are no-no. I am sure there are zillions of different environments with different, often contradictory requirements for error handling. The point I am trying to make is that it is not that difficult to make Monitor more flexible, and it will increase its value tremendously. The ultimate solution would be giving user an opportunity to supply error-handling function which will print out whatever it want to, and return desired return code. Best regards, Kirill Lapshin

2 1

weak_ptr and threads
by Duane Murphy 12 Dec '02

12 Dec '02

The application I am working on uses a number of threads to do work. In some cases pointers from one thread (usually the "main" thread) are passed to a worker thread. The worker thread will use this pointer to handle processing of the results (in various ways). The problem that I am concerned about is lifetime control. Objects in the main thread could be deleted before the worker thread has a chance to use them. This of course would be a "bad thing." I am considering investigating the use of weak_ptr as a potential solution to this problem. The basic idea would be to pass weak_ptr's to the worker thread. The worker thread can then check the validity of the weak_ptr before accessing it. Can weak_ptr be used without using shared_ptr? It seems like they are closely tied. I was hoping to use weak_ptr without forcing the use of shared_ptr. Are there other solutions to this problem, besides the obvious "Dont let it happen." :-) [ Note that we are not yet using boost::threads so this is a more general question about managing inter-thread object lifetimes. ] Thanks for your tips and suggestions, ...Duane

2 1

weak_ptr and threads
by Duane Murphy 12 Dec '02

12 Dec '02

The application I am working on uses a number of threads to do work. In some cases pointers from one thread (usually the "main" thread) are passed to a worker thread. The worker thread will use this pointer to handle processing of the results (in various ways). The problem that I am concerned about is lifetime control. Objects in the main thread could be deleted before the worker thread has a chance to use them. This of course would be a "bad thing." I am considering investigating the use of weak_ptr as a potential solution to this problem. The basic idea would be to pass weak_ptr's to the worker thread. The worker thread can then check the validity of the weak_ptr before accessing it. Can weak_ptr be used without using shared_ptr? It seems like they are closely tied. I was hoping to use weak_ptr without forcing the use of shared_ptr. Are there other solutions to this problem, besides the obvious "Dont let it happen." :-) [ Note that we are not yet using boost::threads so this is a more general question about managing inter-thread object lifetimes. ] Thanks for your tips and suggestions, ...Duane

1 0