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

Re: [Boost-users] [boost][msm] std::vector<MyStateMachine> generate stack overflow
by Christophe Henry 18 May '11

18 May '11

>Hi, > > >while playing with MSM using VS2010+SP1 in a very simple test project (I'm really just trying to test how to organize states classes in files), I've tried to see if I was able to create as many >state machines as I wanted, so I started to write a simple > > std::vector<MyStateMachine> state_machines(42); > > >Here MyStateMachine is an empty state machine class with default front-end and back-end, with only two empty states. > > >VC10.1 produce a warning on this line saying that there will be a stack overflow because there is a runtime recursion. I checked, it does crash (in debug at least). I'd be quite surprised that it has much to do with the vector. I tried for security on a tutorial and all works ok. I'd more likely suspect an incorrect state machine definition, but without code I cannot help you much. Regards, Christophe

2 2

Boost serialization archive stream error exception
by Marco Piacentini 18 May '11

18 May '11

Hi guys..I've the following problem.. I have a file called A.h and a file B.h. each contains some struct like this: (the structs inside the two classes are different) struct Base { friend class access; template <typename Archive> void serialize(Archive& ar,const unsigned int version) { ar & fieldLength; ar & fieldMD; ar & fieldTime_Stamp; } public: unsigned int fieldLength; unsigned int fieldMD; unsigned int fieldTime_Stamp; virtual void f(){} //to be polymorphic the struct }; struct Derived:public Base { . ... } So i serialize the struct in the classic manner: .... std::ostringstream archive_stream; boost::archive::text_oarchive archive(archive_stream); archive.register_type(static_cast<Derived*>(NULL)) archive <<p; // where p is a pointer to Base NOW THE PROBLEM... on the deserialization side, I follow the same (inverse) procedure...if I deserialize singularly the structs in A.h (without include in the project B.h) and the structs in B.h (without include in the project A.h) all works....but If I include in the project both the classes, the deserialization works for one class, but throws the "Stream error exception" in the instruction " archive >> m;" for the other...it seems a conflict in the registration class or something like this... Any ideas?thanks...

2 1

Link errors building against Boost Python
by Chris Weisiger 18 May '11

18 May '11

I'm trying to build PyFTGL against 64-bit Python2.7. PyFTGL uses Boost, so I have to build Boost as well (the downloadable installers are all 32-bit). I created a user-config.jam in my home directory with this line: using python : 2.7 : C:\\Python27x64\\python.exe ; and then built Boost's Python libraries with > bjam.exe --toolset=msvc-9.0 --address-model=64 --with-python PyFTGL uses Python's distutils to build, which doesn't know how to find libraries whose names start with "lib", so I renamed the .lib files in the stage/lib directory: "libboost_python-vc90-mt-1_46_1.lib" to "boost_python-vc90-mt-1_46_1.lib" and so on. Then I tried to build PyFTGL with this: > C:\Python27x64\python.exe setup.py build I get 41 link errors. Here's the first: ftgl.obj : error LNK2019: unresolved external symbol "__declspec(dllimport) protected: __cdecl boost::python::detail::tuple_base::tuple_base(struct boost::python::detail::new_reference_t *)" (__imp_??0tuple_base@detail @python@boost@@IEAA@PEAUnew_reference_t@123@@Z) referenced in function "public: __cdecl boost::python::tuple::tuple(struct boost::python::detail::new_reference_t *)" (??0tuple@python@boost@ @QEAA@PEAUnew_reference_t@detail@12@@Z) I've put the entire output up on a pastebin here: http://pastebin.com/iuPzKdKs Any ideas what's going wrong here? Thanks for any suggestions! -Chris

3 6

To store Boost graph object in a file
by Amanullah Yasin 18 May '11

18 May '11

Hi, I have to store Boost Graph object in a text file and later read it and compare it with other graph. If any one have an idea or implementation please share it. It will be save my time. Thanks in advance. Best Regards YASIN *Always have your stuff when you need it with @Dropbox. 2GB account is free! http://db.tt/xOEJ2Ag*

3 2

[msm] State machine in vector producing recursive call.
by Klaim - Joël Lamotte 18 May '11

18 May '11

Hi, while playing with MSM using VS2010+SP1 in a very simple test project (I'm really just trying to test how to organize states classes in files), I've tried to see if I was able to create as many state machines as I wanted, so I started to write a simple std::vector<MyStateMachine> state_machines(42); Here MyStateMachine is an empty state machine class with default front-end and back-end, with only two empty states. VC10.1 produce a warning on this line saying that there will be a stack overflow because there is a runtime recursion. I checked, it does crash (in debug at least). Is it a bug? I've not found time yet to follow the code see why that produce a recursion (??!!!), but in doubt I verified simple cases : MyStateMachine state_machine ; // no problem MyStateMachine state_machines[42]; // no problem I'll post the full code here once I get home, can't access my test project at my dayjob. I think it might be a well known problem so I ask here first. Joël Lamotte

1 0

Fwd: [MPL] How can I make this occupy less space on disk?
by Pedro d'Aquino 18 May '11

18 May '11

Hello, I'm using Boost.MPL to obfuscate sensitive strings during compilation. I wrote a class called obf_string which can be used in the following way: obf_string<'my_d','atab','ase_','pass','word'> password; connect_to_db( password.unobfuscate() ); If someone runs strings.exe on this executable, they will not see "my_database_password" listed there, because it has been XORed with a binary mask. This approach has several benefits over e.g. preprocessing all the source files with a tool that encrypts all strings. The most significant is that I can easily see and change the string. The problem is that my implementation of obf_string is occupying too much space on disk. I have devised a test to compare it with a normal std::string. The test files, which are attached (test_stdstring.cpp and test_obfstring.cpp), contain 50 different strings each. In one file they are used to construct std::strings, in the other one they are obfuscated through obf_string. On average, each new std::string added 188 bytes, while each obf_string occupied 690 bytes. Here's a plot that shows the difference: http://img59.imageshack.us/img59/6417/obfstring.png. Everything was compiled on VC++10 with /O1 (minimize size). I'm using Boost 1.46.1. I would like to understand why obf_strings take up so much space, and figure out ways I can minimize this. This is really important to me - our application currently instantiates obf_string over 600 times, so the overhead is starting to add up. Naturally, I have also attached the full source code of obf_string. Any help would be greatly appreciated! Thanks, Pedro d'Aquino

3 4

[asio] iPad: resolver fails on a "fresh" 3G connection
by Igor R 18 May '11

18 May '11

Hello, I admit this is probably not a pure ASIO question, but perhaps someone already encountered the following issue (boost 1.46, iOS4.3). When trying to resolve a host name just after 3G connection is established (eg., switch "airplane mode" off and on), I get the following error: "Host not found (authoritative)". The query looks like this: tcp::resolver::query query(tcp::v4(), "my.host.com", "9988"); On the other hand, if prior to the above query I open the browser or some other application that connect to the Internet, then the query works correctly. Moreover, it seems that if I try the above query several times, it succeedes after a while. Another interesting point: all the above occurs with 3G connection only, while with WiFi everything works well. What's could be the reason of such a weird behavior? Thanks.

1 0

[boost][msm] std::vector<MyStateMachine> generate stack overflow
by Klaim - Joël Lamotte 18 May '11

18 May '11

Hi, while playing with MSM using VS2010+SP1 in a very simple test project (I'm really just trying to test how to organize states classes in files), I've tried to see if I was able to create as many state machines as I wanted, so I started to write a simple std::vector<MyStateMachine> state_machines(42); Here MyStateMachine is an empty state machine class with default front-end and back-end, with only two empty states. VC10.1 produce a warning on this line saying that there will be a stack overflow because there is a runtime recursion. I checked, it does crash (in debug at least). Is it a bug? I've not found time yet to follow the code see why that produce a recursion (??!!!), but in doubt I verified simple cases : MyStateMachine state_machine ; // no problem MyStateMachine state_machines[42]; // no problem I'll post the full code here once I get home, can't access my test project at my dayjob. I think it might be a well known problem so I ask here first. Joël Lamotte

1 0

MPI rank() & size() type
by Kraus Philipp 17 May '11

17 May '11

Hello, I use Boost MPI for some project, but I have a question of the return datatype of the communication rank() and size(). size() returns a unsigned datatype, but rank() a signed, so if I compare rank() with size() I will get a warning. Is there a reason for using different return types? Thanks Phil

1 0

Segmentation fault when throwing exception?
by Germán Diago 17 May '11

17 May '11

Hello, I have some code that throws an exception through BOOST_THROW_EXCEPTION. The class derives from std::runtime_error. The thing is that my program crashes and doesn't catch the exception if I use this macro, but if I use a language throw it behaves correctly. Does anyone have a clue? My code doesn't spawn threads. Thanks in advance.

4 6