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

Access violation in boost::condition
by Eric Colleu 22 Jun '06

22 Jun '06

Hi, I'm using boost 1.33.1 on multi-processor plateform (2 Pentium4 3.2 GHz) and Windows XP SP2 (compiled with MinGW). I wrote a MessageQueue class which use boost::mutex and boost::condition to wait for message and notify message reception (see code below) I made a little test that create 2 boost::thread sending and receiving messages in 2 queues. The access violation is produced in the "WaitForSingleObject" method called by "condition_impl::do_wait" Have you ever meet this problem? Eric Colleu ------messageQueue.cpp-----START #include <iostream> #include "messageQueue.hpp" namespace util { MessageQueue::MessageQueue(void) : m_mutex() , m_condition() , m_queue() , m_cancel(false) { } MessageQueue::~MessageQueue(void) throw() { } void MessageQueue::push(MessageHandle message) { boost::mutex::scoped_lock lock(m_mutex); m_queue.push(message); m_condition.notify_all(); } MessageHandle MessageQueue::pop(const boost::xtime& delay) { boost::mutex::scoped_lock lock(m_mutex); while (m_queue.size() == 0) { if (!m_condition.timed_wait(lock, delay)) { throw MsgTimeout("timeout in pop message"); } } if(m_cancel) { m_cancel = false; throw MsgWaitCancelled("Wait Event cancelled"); } MessageHandle msg = m_queue.front(); m_queue.pop(); return msg; } MessageHandle MessageQueue::pop(void) { boost::mutex::scoped_lock lock(m_mutex); while (m_queue.size() == 0 && ! m_cancel) { m_condition.wait(lock); } if(m_cancel) { m_cancel = false; throw MsgWaitCancelled("Wait Event cancelled"); } MessageHandle msg(m_queue.front()); m_queue.pop(); return msg; } int MessageQueue::size(void) { boost::mutex::scoped_lock lock(m_mutex); return m_queue.size(); } void MessageQueue::cancel_wait(void) { boost::mutex::scoped_lock lock(m_mutex); m_cancel = true; m_condition.notify_all(); } } ------messageQueue.cpp-----END ------messageQueue.hpp-----START #if defined(CORE_UTIL_QUEUE_MESSAGEQUEUE_CYCLE) #error Header cyclic inclusion detected. #endif // defined(CORE_UTIL_QUEUE_MESSAGEQUEUE_CYCLE) #define CORE_UTIL_QUEUE_MESSAGEQUEUE_CYCLE #if !defined(CORE_UTIL_QUEUE_MESSAGEQUEUE_HPP) #define CORE_UTIL_QUEUE_MESSAGEQUEUE_HPP // =================================================================== #include <queue> #include <boost/shared_ptr.hpp> #include <boost/thread.hpp> #include <boost/thread/condition.hpp> #include "core/exceptions.hpp" #include "message.hpp" namespace util { /** File d'attente de messages. * Cette classe fonctionne comme une FIFO, et permet de * faire une attente bloquante d'un message. */ class MessageQueue { public: MessageQueue(void); virtual ~MessageQueue(void) throw(); /** Empile un message en fin de liste. * @param message Le message a ajouter. */ void push(boost::shared_ptr<Message> message); /** Attend un message pendant un temps déterminé. * @param delay Temps d'attente du message. * @return Le message si un message a été reçu * @throws MsgTimeout Si aucun message n'a été empilé pendant le temps défini en paramètre. * @throws MsgWaitCancelled Si l'attente a été interrompue par un cancel_wait. */ MessageHandle pop(const boost::xtime& delay); /** Attend un message infiniment. * @return Le message si un message a été reçu * @throws MsgWaitCancelled Si l'attente a été interrompue par un cancel_wait. */ MessageHandle pop(void); /** Retourne la taille de la file d'attente. * @return le nombre d'éléments restant dans la file d'attente. */ int size(void); /** Interrompt une attente de message. */ void cancel_wait(void); private: boost::mutex m_mutex; boost::condition m_condition; std::queue<MessageHandle > m_queue; bool m_cancel; }; } // =================================================================== #endif // !defined CORE_UTIL_QUEUE_MESSAGEQUEUE_HPP #undef CORE_UTIL_QUEUE_MESSAGEQUEUE_CYCLE ------messageQueue.hpp-----END ------message.hpp-----START #if defined(CORE_UTIL_QUEUE_MESSAGE_CYCLE) #error Header cyclic inclusion detected. #endif // defined(CORE_UTIL_QUEUE_MESSAGE_CYCLE) #define CORE_UTIL_QUEUE_MESSAGE_CYCLE #if !defined(CORE_UTIL_QUEUE_MESSAGE_HPP) #define CORE_UTIL_QUEUE_MESSAGE_HPP // =================================================================== #include <string> #include <boost/shared_ptr.hpp> namespace util { /** * La classe Message représente les message transittant dans les MessageQueue, * entre le gestionnaire (device::AbstractDeviceManager) et la pile de communication * (stack_manager::SimpleStack). * Ces messages peuvent être des requêtes, réponses ou événements. */ class Message { public: Message(std::string id, bool is_event) : m_id(id), m_isEvent(is_event) {}; virtual ~Message(void) throw() {}; /** Indique si le message est un événement. * @return * - true : s'il s'agit d'un événement * - false sinon. */ bool isEvent(void) {return m_isEvent;}; /** Retourne l'identifiant du message. * @return une chaîne de caractère représentant l'identifiant du message. */ std::string getId(void) {return m_id;}; /** Encode les données à transmettre à la pile de communication sous forme d'une chaîne. */ virtual std::string encode(void) = 0; virtual void decode(void) = 0; protected: std::string m_id; private: bool m_isEvent; }; /** Pointeur sur un Message. */ typedef boost::shared_ptr<Message> MessageHandle; } // =================================================================== #endif // !defined CORE_UTIL_QUEUE_MESSAGE_HPP #undef CORE_UTIL_QUEUE_MESSAGE_CYCLE ------message.hpp-----END

2 2

[serialization] XML archive is order-dependent
by Stan Vasilyev 22 Jun '06

22 Jun '06

I'm using boost::archive::xml_iarchive to read data from an XML archive created by hand. It works as long as I keep all the tags in the same order that they are serialized in my code. As soon as I put in an extra XML tag somewhere, de-serialization fails. Since XML is order-independent, is it possible to make boost::serialization also order-independent? Stan

5 6

'standard' free functions for operators, etc
by Gottlob Frege 22 Jun '06

22 Jun '06

In many parts of boost, free functions are used as customization points for various things, better explained with examples: intrusive_ptr_add_ref() advance/increment/etc in boost iterator library etc Are there other examples? Now consider my example: template<T, U> myclass_assign(T *t, U const & u) { t->operator=(u); } That forms the default implementation, but clients can define their own specializations (for cases where T is a simple type and not an object at all). At first I thought that 'myclass_assign' should just be 'assign' since the assignment should be the same, just based on type (not on the circumstances of why 'myclass' is doing the assign), but (just while writing the email) I saw a problem there: in the case of, say: typedef int RecordID; assign(RecordID id, SomeValue const & value) { // do the assignment } But of course, RecordID is just a typedef, not a 'strong typedef', so it could be just one example of many assign(int, SomeValue), completely unrelated to each other. So how should that be handled? Specific naming (using 'myclass_') lowers the chances of a problem, but doesn't remove it. Which means that, in theory at least, intrusive_ptr has the same problem: intrusive_ptr_add_ref(some_typedef_that_is_really_an_int) { ... } intrusive_ptr_add_ref(some_OTHER_typedef_that_is_also_an_int) { ... } Should myclass, and also intrusive_ptr instead use a policy-based approach, similar to boost iterator? intrusive_ptr<T, policy = legacy_policy> { ... } Is there some other approach that I'm not considering (ie runtime boost::functions instead of compile time policy functions)? Tony

2 1

Re: [Boost-users] Boost serializer: incomplete type error reported
by Diego Molla 22 Jun '06

22 Jun '06

Thanks for your advice, Gennaro. I had to move the following includes to the beginning of the file: #include <boost/archive/text_oarchive.hpp> #include <boost/archive/text_iarchive.hpp> I also realised that the following gives the same error of incomplete type: void SetOfRules::save(const string &fileName) // Save into a file { ofstream fout(fileName.c_str()); boost::archive::text_oarchive oa(fout); oa << *this; fout.close(); } I had to use a helper function like this: void _save(const SetOfRules &sor, const string &fileName); void SetOfRules::save(const string &fileName) // Save into a file { _save(*this,fileName); } void _save(const SetOfRules &sor, const string &fileName) { std::ofstream fout(fileName.c_str()); boost::archive::text_oarchive oa(fout); oa << sor; fout.close(); } // save And similarly to load a class. Diego >>Hello, >> >>I have an error that is puzzling me. I have defined serialization of >>several custom classes, and everything worked perfectly until we've >>changed to a new Linux OS and had to reinstall Boost. Now the program >>doesn't compile and the error given is: >> >> error: incomplete type >>'boost::serialization::extended_type_info_null<SetOfRules>' used in >>nested name specifier >> >>I don't really understand why it says there is an incomplete type when >>there used to be no errors. The same error happens with all the classes >>where I've defined serialization. > > I'm not too much into the latest gcc developments (certainly others > will give a more accurate reply) but it looks like a two-phase lookup > issue: at the time your function is parsed SetOfRules is actually > incomplete, but the old gcc version you had didn't diagnose it and > delayed analisys to instantiation time. > > This is just my feeling. You might try moving SetOfRules' #include > before. > > --Gennaro. -- Do you want to know what is going on in Language Technology in Australia and New Zealand? Join ALTA -- http://www.alta.asn.au/ This message is intended for the addressee named and may contain confidential information. If you are not the intended recipient, please delete it and notify the sender. Views expressed in this message are those of the individual sender, and are not necessarily the views of Macquarie University. --------------------------------------------------------------------- Diego MOLLA ALIOD diego(a)ics.mq.edu.au Department of Computing http://www.ics.mq.edu.au/~diego Macquarie University

1 0

Re: [Boost-users] problem in multithreaded regex_replace on linux
by Ashish Varman 22 Jun '06

22 Jun '06

Dear John Apparantly, I am doing something wrong that is very elementary. I do not have any -mt version of the regex library. I even dispensed with STLport and tried to build boost first for default system libraries only. After giving the command "bjam -sTOOLS=gcc" in the directories boost_1_33_1/libs/regex/build and boost_1_33_1/libs/thread/build, I have the following libraries: $ find | grep "\.a$" ./libs/thread/build/bin-stage/libboost_thread-gcc-mt-d-1_33_1.a ./stage/lib/libboost_regex-gcc-d-1_33_1.a ./stage/lib/libboost_regex-gcc-1_33_1.a ./bin/boost/libs/regex/build/libboost_regex.a ./bin/boost/libs/regex/build/libboost_regex.a/gcc/debug/libboost_regex-gcc-d-1_33_1.a ./bin/boost/libs/regex/build/libboost_regex.a/gcc/release/libboost_regex-gcc-1_33_1.a ./bin/boost/libs/thread/build/libboost_thread.a ./bin/boost/libs/thread/build/libboost_thread.a/gcc/debug/threading-multi/libboost_thread-gcc-mt-d-1_33_1.a I have not been able to find how to build -mt versions of the regex library. Apparantly, the threads library is built so by default. Also, while trying to compile static_mutex_test.cpp, a lot of undefined references crop up even when I link with both boost::regex and boost::thread. The command was $ g++ -I ~/Software/boost_1_33_1 -L ~/Software/boost_1_33_1/libs/thread/build/bin-stage/ -L ~/Software/boost_1_33_1/stage/lib/ -lboost_thread-gcc-mt-d-1_33_1 -lboost_regex-gcc-1_33_1 -pthread static_mutex_test.cpp /tmp/ccvkfjk2.o(.text+0x16): In function `failed_test(char const*, int, char const*)': : undefined reference to `boost::scoped_static_mutex_lock::scoped_static_mutex_lock[in-charge](boost::static_mutex&, bool)' /tmp/ccvkfjk2.o(.text+0xa8): In function `failed_test(char const*, int, char const*)': : undefined reference to `boost::scoped_static_mutex_lock::~scoped_static_mutex_lock [in-charge]()' ----------- snip -------------- /tmp/ccvkfjk2.o(.text+0x5da): In function `t3()': : undefined reference to `boost::scoped_static_mutex_lock::~scoped_static_mutex_lock [in-charge]()' /tmp/ccvkfjk2.o(.text+0x5f4): In function `t3()': : undefined reference to `boost::scoped_static_mutex_lock::~scoped_static_mutex_lock [in-charge]()' collect2: ld returned 1 exit status Since the scoped_static_mutex_lock is defined only in the regex sources, I think the errors are due to the -mt versions not being built. I have verified that -mt versions are being built and used on windows. I tried using the same technique (building using makefiles) but no luck $ make -f gcc.mak $ find | grep "\.a$" ./gcc/libboost_regex-gcc-1_33_1.a ./gcc/libboost_regex-gcc-d-1_33_1.a So now my main problem becomes "How do I build -mt versions of regex on linux". I hope the answer does not make me look too dumb. Thanks in advance Yours Sincerely Ashish Varman

2 1

date_time microsec_clock::univeral_time() implementation
by Brian Neal 22 Jun '06

22 Jun '06

We have an application that was using our own custom functions for getting the OS time. We were basically representing time as the number of milliseconds since the UNIX epoch in an unsigned long, or as the number of nanoseconds since the epoch in a 64-bit type. We later discovered boost, and cut everything over to boost::date_time. This has worked out very, very well and has uncovered several time related bugs in our old code. However, we do have a small percentage of code that is time critical, and we have noticed that calling microsec_clock::univeral_time() is 2.6 times slower than our old function, and for this one bit of code that is significant. Looking at the implementation for UNIX like systems, I see that create_time() in microsec_time_clock.hpp is calling gettimeofday() followed by a call to gmtime() or localtime() as appropriate. I think it's this second call that is the difference between our old code and boost. In our time critical chunk of code we are just time tagging events, and we aren't concerned about timezone. I was thinking about adding a function to our code that called clock_gettime() (which is what our old code did), but then constructing a time_type out of that using the UNIX epoch for the date part and the results of clock_gettime() for the time_duration part, i.e, in pseudo code: time_type our_get_time() { clock_gettime(CLOCK_REALTIME, &timespec); time_duration td = massage(timespec); return time_type(unix_epoch, td); } This still won't be as fast as our old code, because we essentially ignored the date part of the times and simply worked with durations since the epoch. But I think the killer for us is the call to gmtime, which the above avoids. Comments? I'm curious why this approach wasn't taken in boost. Was it the timezone factor? I'm also curious why gettimeofday() was chosen over clock_gettime()? Is one more available than the other? Thanks.

2 2

Re: [Boost-users] warning C4561: '__fastcall' incompatible with the '/clr' option:, converting to '__stdcall'
by Sohail Somani 21 Jun '06

21 Jun '06

> -----Original Message----- > From: boost-users-bounces(a)lists.boost.org > [mailto:boost-users-bounces@lists.boost.org] On Behalf Of > Ovanes Markarian > Sent: Wednesday, June 21, 2006 6:48 AM > To: boost-users(a)lists.boost.org > Subject: Re: [Boost-users] warning C4561: '__fastcall' > incompatible with the '/clr' option:, converting to '__stdcall' > > I don not think you should mix up managed C++ with ISO C++. > Please read this critical reviews by > Danny Kalev about features of CLI and how they are compatible > with ISO C++. > > A Critical Overview of C++/CLI > > Part 1: > http://www.informit.com/guides/content.asp?g=cplusplus&seqNum=274 > > Part 2: > http://www.informit.com/guides/content.asp?g=cplusplus&seqNum=275 I had no idea... Thanks for the link

1 0

memory leak in VC6
by Elisabeth Benoit 21 Jun '06

21 Jun '06

Hello, I'm using boost::bind and boost::signal with VC6. VC6 is detecting some memory leak for a certain type of signal.connect(). I've done a little test program to check if the problem was due to some bug or special configuration in my main program, but the memory leaks are still detected in the test situation. Let me explain the structure of my test program. I have a main, and in the main, an object CCloneAppRecipe named clone. The class CCloneAppRecipe contains a member m_PtrTestMem (a shared pointer, but a raw point or an object does the same thing) of the class CTestMem. CTestMem contains 3 instances of the class CTestTab, and the class CTestTab contains a boost signal typedef boost::signal0<void> TypeSignalTest; TypeSignalTest m_SignalTest; In my main, I invoke clone.connectPtrTestMem(). clone.connectPtrTestMem() then invokes CTestMem::ConnectSignalTriplePtrTab(boost::bind(&CloneAppRecipe::test, this)); ConnectSignalTriplePtrTab() then invokes ConnectSignal(TypeSignalTest::slot_type slot) for each of the 3 instances of the class it contains. ConnectSignal then calls m_SignalTest.connect(slot). And I have a memory leak: 2 blocks of 16 bytes. If I connect only one instance of CTestTab, no leak. Two instances, one leak of 16 bytes. Three, two leaks of 16 bytes, and so on. I connect the m_SignalTest with a CCloneAppRecipe's slot. If I connect it to a CTestMem slot, I have no leak. Or if I place 3 instances of CTestTab directly in the CCloneAppRecipe class (so I don't use CTestMem anymore), I have no leak. But that's not what I need to do. I've tested plenty of variations (shared_pointer or object, etc.) Does anybody have a clue. Thanks elisabeth benoit

1 0

Re: [Boost-users] warning C4561: '__fastcall' incompatible with the '/clr' option:, converting to '__stdcall'
by Ovanes Markarian 21 Jun '06

21 Jun '06

Hello, > >When using the /clr option for generating managed code with Visual C++ >one gets the warning > >C4561: '__fastcall' incompatible with the '/clr' option:,converting to '__stdcall' > >Is there a way to compile files that include boost header files >such that __stdcall is the default, or is ignorng or switching >off the warning the "solution"? > >Thanks in advance, > >andreas I don not think you should mix up managed C++ with ISO C++. Please read this critical reviews by Danny Kalev about features of CLI and how they are compatible with ISO C++. A Critical Overview of C++/CLI Part 1: http://www.informit.com/guides/content.asp?g=cplusplus&seqNum=274 Part 2: http://www.informit.com/guides/content.asp?g=cplusplus&seqNum=275 Regards, Ovanes Markarian

1 0

strange compiler behavior...
by Ovanes Markarian 21 Jun '06

21 Jun '06

Dear Boosters, I have some strange problem and don't understand if this is really intended so by the C++ standard or is this a bug in the compiler of MS Visual C++ 2003 & 2005. I have a template class which receives an mpl::vector with possible storage types. This class contains an instance of boost::variant but is intended to wrap it with a little bit different functionality. //Solution A //A body of a _working__ solution: template<class Seq_> class my_union { public: template<class T> my_union(const T& rhs) {...} //... }; //Solution B //not working more general solution //this struct compiles fine template<T> struct const_ref : boost::add_const<typename boost::add_reference<T>::type> {}; template<class Seq_> class my_union { public: template<class T> my_union(typename const_ref<T>::type rhs) {...} //... }; now call of the class: //assume my_pair is defined earlier //assume classes A,B,C,D are defined upon typedef my_pair<0, A> t1; typedef my_pair<1, B> t2; typedef my_pair<2, C> t3; typedef my_pair<3, A> t4; typedef mpl::vector<t1,t2,t3,t4>::type types_vector; typedef my_union<types_vector> union_t; t1 t(A()); union_t u(t); //error! with Solution B, but fine with Solution A in the my_union I get an error stating: error C2664: 'my_union<Seq_>::my_union(const my_union<Seq_> &)' : cannot convert parameter 1 from 't1' to 'const my_union<Seq_> &' with [ Seq_=types_vector ] Reason: cannot convert from 't1' to 'const my_union<Seq_>' with [ Seq_=types_vector ] No user-defined-conversion operator available that can perform this conversion, or the operator cannot be called Why does Solution A work fine and Solution B does not? Is this really intended by a standard? With Kind Regards, Ovanes Markarian

1 0