February 2013 - Boost-users - lists.preview.boost.org

upgrading lock
by Igor R 21 Feb '13

21 Feb '13

Hello, Just wanted to be sure that I'm doing this correctly. typedef boost::upgrade_lock<boost::shared_mutex> upgrade_lock; typedef boost::upgrade_to_unique_lock<boost::shared_mutex> upgrade_to_unique_lock; upgrade_lock readLock(mutex_); // accessing shared resources in read-only mode //... upgrade_to_unique_lock writeLock(readLock); // modifying shared resources Is the above correct? In particular, upgrade_lock acquires "read" lock in raii style (similarly to shared_lock), right? Thanks.

2 2

Re: [Boost-users] Can boost help me organize/parse lots of binary data?
by Gabriel Bizzotto 21 Feb '13

21 Feb '13

Chris, > class Item > { > virtual void Decode(istream& Stream) = 0; > } > class Double : public Item > class String : public Item > class EmployeePacket > { > std::vector<Item*> Items; ... Maybe boost::any will do better here ? http://www.boost.org/doc/libs/1_53_0/doc/html/any.html

2 1

[lockfree] compile-time capacity vs. run-time capacity
by Akira Takahashi 21 Feb '13

21 Feb '13

Hi, I confused with boost::lockfree::queue behavior. Fist, run-time capacity: #include <iostream> #include <boost/lockfree/queue.hpp> int main() { boost::lockfree::queue<int, boost::lockfree::fixed_sized<true>> que(3); std::cout << std::boolalpha; std::cout << que.push(1) << std::endl; // true std::cout << que.push(2) << std::endl; // true std::cout << que.push(3) << std::endl; // true std::cout << que.push(4) << std::endl; // false } I think this behavior is correct. But, compile-time capacity: #include <iostream> #include <boost/lockfree/queue.hpp> int main() { boost::lockfree::queue<int, boost::lockfree::capacity<3>> que; std::cout << std::boolalpha; std::cout << que.push(1) << std::endl; // true std::cout << que.push(2) << std::endl; // true std::cout << que.push(3) << std::endl; // false : why? std::cout << que.push(4) << std::endl; // false } Is this bug or specification? Thanks, Akira >>======================== Akira Takahashi mailto:faithandbrave@gmail.com site: https://sites.google.com/site/faithandbrave/about/en

1 1

Coroutine documentation question
by Brian Wood 21 Feb '13

21 Feb '13

I have a question about a sentence on this page: http://www.boost.org/doc/libs/1_53_0/libs/coroutine/doc/html/coroutine/intr… "The principal difference between coroutines and routines is that a coroutine enables explicit suspend and resume their progress via additional operations by preserving local state ..." I'm wondering if it would be accurate/helpful to change that to this: "The principal difference between coroutines and routines is that coroutines enable explicit suspend and resume of their progress via additional operations by preserving local state ..." Shalom Brian Wood Ebenezer Enterprises - so far G-d has helped us. http://webEbenezer.net (651) 251-9384

1 0

Boost.Geometry: Correct use of multi-part polygons
by John Lilley 20 Feb '13

20 Feb '13

Greetings, I am looking for some guidance around the correct creation and use of multi-part polygons. I am a bit confused about when to use each of the following: * Polygon with multiple rings * Vector of polygons * Polygon set * Set/vector of polygons, each potentially containing multiple rings I gather from this example picture: http://geometrylibrary.geodan.nl/03__polygon__example_8cpp-example.html That if one wants holes to be opened as a result of intersection, they need to be rings as opposed to polygon sets or vectors of polygons? An example would be polygon(s) representing an island chain. Some of the islands are atolls and as such have a hole in the middle. What would a representation of this set of islands look like? And, what If I don't know if I have "real" rings (interior) vs just a collection of polygons (like I got this from an external source that doesn't have the same model). If I just throw a pile of points at a polygon and call clean(), will that magic everything up for me? john

2 4

Boost.geometry intersection / difference question
by John Lilley 20 Feb '13

20 Feb '13

What types are the intersection and difference algorithms implemented for? In particular: Can polygons intersect with non-polygons (e.g. point sets to produce clipped point sets, or linestrings to produce clipped linestrings)? Similar question for difference, where polygon is the second input. Thanks John

2 1

Re: [Boost-users] [Atomic] [Tread] Proposal for Synchronized<T>
by Gabriel Bizzotto 20 Feb '13

20 Feb '13

Hi, > I've been working already on a synchronized_value class that is in trunk [1]. I plan to merge it to release branch as soon as I have finished the tests and the documentation. You can see some examples in [2] and [3]. > I've take a look to your class and I think my class provides everything yours provide. Surely your version has been given more thought. > My class allows in addition to lock several synchronized values using the deadlock free lock algorithm as e.g. I did not understand the unique_synchronize part, though. > Please could you compare your class to mine and let me know what do you think? I can't see any flaw right now, but I'm still playing with it. If anything shows up I'll let you know. Thanks !

2 1

Re: [Boost-users] [Atomic] [Tread] Proposal for, Synchronized<T>
by Gabriel Bizzotto 20 Feb '13

20 Feb '13

> I have one that does reader/writer locking. I'll pass it along for comparison along with design notes (it was done partially for teaching), if it's OK to send a zip file to you. > ?John Sure, let's take a look ! You also want to carbon-copy to vicente.botet(a)wanadoo.fr, which is putting the finishing touch on the future boost::synchronized_value.

2 1

Directed cycle detection in a DAG between two nodes
by Amanullah Yasin 20 Feb '13

20 Feb '13

Hi, In Hill Climbing algorithm, before applying an operation "add_edge", "reverse_edge" or "remove_edge", I need to check whether the graph will be a Directed Acyclic Graph (DAG) or not. So there is recursive function but in some cases it goes in infinite loop and gives "stack overflow" error. //======= Function ======= bool myAlgo::remainsADAG(const Node_type& nodeInit, const Node_type& node, const Graph_type& graph) { graph_traits<Graph_type>::adjacency_iterator vi, vi_end; tie(vi, vi_end) = adjacent_vertices(node, graph); for (; vi != vi_end; ++vi) { if ((*vi == nodeInit) || (!remainsADAG(nodeInit, *vi, graph))) return false; } return true; } ============== e.g. for Graph 12->19->20->23 19->24->31->18>19 18->32 To test if there is a directed path between v12 and v32 it goes as: 12->19->20->23 24->31->18>19 *There is an infinite loop{18->19->20->23, 24->31->18}.* ================= Can some one help me or is there any other easy way to find a directed path between two nodes in boost graph? Thanks in advance. YASIN

4 3

[Preprocessor] Generate #include directive with preprocessor
by Adam Badura 20 Feb '13

20 Feb '13

Is it possible to use preprocessor to generate #include directives (or in any other way include the files)? For example if I would have a sequence of strings (being file paths) what to do to include all those files? It is possible at all?

4 3

Sperical / geo coordinates question: crossing anti-meridian
by John Lilley 19 Feb '13

19 Feb '13

I am trying to determine if boost.geometry handles certain cases of spherical coordinates. Suppose you have county outlines of Alaska. The Aleutian Islands chain crosses the anti-meridian, and so contains latitudes that are in the ranges [ 180 - delta ... 180 ] and [ -180, -180 + delta ] where delta is < 20 degrees or so. Simple (2D planar) point-in-polygon and polygon intersection algorithms fall down on these cases due to anti-meridian crossings and flipping between +/- values. Does Boost geometry deal with these issues "correctly"? I put correctly in quotes because there is some real ambiguity in the coordinate system (maybe the polygon really does wrap around the world?), but one could apply a "minimum perimeter" heuristic and normalize the polygons before performing the computation by moving the [ 180 - delta ... 180 ] range over to [ -180 - delta ... -180 ] or vice-versa. Thanks, john

2 2

[Atomic] [Tread] Proposal for, Synchronized<T>
by John M. Dlugosz 19 Feb '13

19 Feb '13

On 2/19/2013 4:02 PM, boost-users-request-at-lists.boost.org |Boost/Allow to home| wrote: > Subject: Re: [Boost-users] [Atomic] [Tread] Proposal for > Synchronized<T> > Message-ID:<5123C32D.9020700(a)wanadoo.fr> > Content-Type: text/plain; charset="iso-8859-1"; Format="flowed" > > Le 19/02/13 17:58, Gabriel Bizzotto a ?crit : >> >Hi ! >> > >> >I've created two simples classes that helped me avoid data races when >> >sharing objects among threads. >> > >> >The idea is to declare objects that need for synchronization as >> >Synchronized<T>, which holds a T and a mutex. To use the T, a Get >> >function has to be called that returns a proxy, which holds a pointer >> >to the T and a scoped lock on the mutex. The proxy defines operator-> >> >and operator* to give access to the T. To have acces to the T, one has >> >to create the proxy. As long as the proxy exists, the mutex associated >> >with the T is locked, effectively preventing use by other threads. I have one that does reader/writer locking. I'll pass it along for comparison along with design notes (it was done partially for teaching), if it's OK to send a zip file to you. —John

1 0

[property_map] constructor of vector_property_map
by Andreas Fabri 19 Feb '13

19 Feb '13

Hello, Why is initial_size in the constructor __________________________________________________________________ vector_property_map(unsigned initial_size, const IndexMap& index = IndexMap()) This constructor version allows to specify maximum index of element that will be stored. Correct number will improve performance, but semantic is always the same. Source: http://www.boost.org/doc/libs/1_53_0/libs/property_map/doc/vector_property_… __________________________________________________________________ not std::size_t, for the simple reason that the std::vector might be bigger. best regards, andreas -- Andreas Fabri, PhD Chief Officer, GeometryFactory Editor, The CGAL Project phone: +33.492.954.912 skype: andreas.fabri

2 1

compiling error from code using boost::bind code in Vs2010
by Miguel Carvajal 19 Feb '13

19 Feb '13

hello, actually i am using boost 1.48 on vs2010 and the following code does not compile on boost #include<boost/bind.hpp> #include<boost/function.hpp> using namespace boost; class A { public: void evaluate(function<VectorClass(VectorClass,double)> f) }; class B { A a; public; VectorClass f(VectorClass,double); void test() { a.evaluate(bind(f,this,_1,_2)); } }; Any help?

2 1

[Atomic] [Tread] Proposal for Synchronized<T>
by Gabriel Bizzotto 19 Feb '13

19 Feb '13

Hi ! I've created two simples classes that helped me avoid data races when sharing objects among threads. The idea is to declare objects that need for synchronization as Synchronized<T>, which holds a T and a mutex. To use the T, a Get function has to be called that returns a proxy, which holds a pointer to the T and a scoped lock on the mutex. The proxy defines operator-> and operator* to give access to the T. To have acces to the T, one has to create the proxy. As long as the proxy exists, the mutex associated with the T is locked, effectively preventing use by other threads. For more details and examples, I've already published the classes here http://codereview.stackexchange.com/questions/15632/force-locking-for-threa… I find the scheme general purpose enough to fit the Atomic lib, or perhaps Thread. Please manifest yourselves if there is any interest in such device. Thx. Gabriel ++++++[>+++[>++++>++++++<<-]<-]>>-.>---.-------..>++++[<++++>-]<+. WebRep Classificação geral Este site não tem classificação (número de votos insuficientes) WebRep Classificação geral Este site não tem classificação (número de votos insuficientes) WebRep Classificação geral Este site não tem classificação (número de votos insuficientes) WebRep Classificação geral Este site não tem classificação (número de votos insuficientes) WebRep Classificação geral Este site não tem classificação (número de votos insuficientes) WebRep Classificação geral Este site não tem classificação (número de votos insuficientes) WebRep Classificação geral Este site não tem classificação (número de votos insuficientes) WebRep Classificação geral Este site não tem classificação (número de votos insuficientes) WebRep Classificação geral Este site não tem classificação (número de votos insuficientes)

3 2

[Property Tree] special character in XML tag
by Gábor Márton 19 Feb '13

19 Feb '13

Hi, I have the following property tree, dumped in INFO format: CudbAppServiceId=1 { userLabel "" sqlAppSrvPlSchema identities-pl.sql CudbAppServiceId CudbAppServiceId=1 } When I dump it in XML format, I got the following: <?xml version="1.0" encoding="utf-8"?> <CudbAppServiceId=1> <userLabel/> <sqlAppSrvPlSchema>identities-pl.sql</sqlAppSrvPlSchema> <CudbAppServiceId>CudbAppServiceId=1</CudbAppServiceId> </CudbAppServiceId=1> This is not a valid XML, because we have a tag "CudbAppServiceId=1". Of course, any other xml parser like xmllint will fail to parse it, because of the equality sign in the tag. Is there any solution to this problem? Like somehow emitting = instead of the raw '=' inside the tag and in the text? Thanks, Gabor

3 7

chaining functions input output
by MM 19 Feb '13

19 Feb '13

Hello, i have these 2 functions provided to me and I wonder whether they can somehow be chained without using intermediary memory? http://codepad.org/tTwkUQhp Rds, MM

4 7

[Lockfree] Interprocess Support
by Gábor Márton 19 Feb '13

19 Feb '13

Hi, In 1.53 documentation it is said: The boost.lockfree data structures have basic support for Boost.Interprocess. The only problem is the blocking emulation of lock-free atomics, which in the current implementation is not guaranteed to be interprocess-safe. What does that and interprocess-safety exactly mean? Is it possible to use an interprocess allocator (e.g. allocator<int, managed_shared_memory::segment_manager>) together with lockfree::queue? Or shall I expect unexpected behaviour? Thanks, Gabor

1 0

Thread safety in Boost msm
by Ukil, Soumya P 18 Feb '13

18 Feb '13

Hi, I am new to boost meta state library. I am building one app, where I have to use msm. My application has two threads, where they use same msm object as a shared resources. Now will the msm be thread safe if both threads call process_event on the same msm object? Any idea is welcome. Thanks, Soumya P. Ukil Morgan Stanley | ISG Technology Building #5. Sector 30 | Mindspace, Goregaon (West) Mumbai (Ex Bombay), 400 090 Phone: +91 22 6641-0038 Soumya.Ukil(a)morganstanley.com<mailto:Soumya.Ukil@morganstanley.com> Be carbon conscious. Please consider our environment before printing this email. ________________________________ NOTICE: Morgan Stanley is not acting as a municipal advisor and the opinions or views contained herein are not intended to be, and do not constitute, advice within the meaning of Section 975 of the Dodd-Frank Wall Street Reform and Consumer Protection Act. If you have received this communication in error, please destroy all electronic and paper copies and notify the sender immediately. Mistransmission is not intended to waive confidentiality or privilege. Morgan Stanley reserves the right, to the extent permitted under applicable law, to monitor electronic communications. This message is subject to terms available at the following link: http://www.morganstanley.com/disclaimers If you cannot access these links, please notify us by reply message and we will send the contents to you. By messaging with Morgan Stanley you consent to the foregoing.

2 1

[Boost.Coroutine] sample in docs
by Lukas Czaja 18 Feb '13

18 Feb '13

Hi, i would like to ask about sample in docs http://www.boost.org/doc/libs/1_53_0/libs/coroutine/doc/html/coroutine/intr… what if several threads run asio::io_service::run(), so is it safe to resume execution state in CompletionHandler in diffrent thread from when execution state was saved ? And what will happen if io_service completes before saving contex (look code below)? s_.async_read_some( boost::asio::buffer( buffer_ + pb_size, bf_size - pb_size), boost::bind( & coro_t::operator(), & coro_, _1, _2) ); // Q: what will happen if handler completes here, before contex switch ca_();

1 0

[Interprocess] managed_shared_memory between 32 & 64bit
by leunm 17 Feb '13

17 Feb '13

Hi, I am new to Boost and am looking for a way to pass xyz coordinates from a 32-bit application to a 64-bit one. I see two recent posts on this same topic, but the explanation, as well as the Boost documentation (on offset_ptr) is a bit sparse. Below is a simple example I modified from the Boost documentations <http://www.boost.org/doc/libs/1_53_0/doc/html/interprocess/quick_guide.html…> , which works for either 32-bit or 64-bit but not when the parent is 32-bit and child is 64-bit, and vice versa. Can someone tell me what I need to do enable this function? Thank you in advance, Mike ----------------------------------------------------------------- #define BOOST_DATE_TIME_NO_LIB #include <boost/interprocess/managed_shared_memory.hpp> #include <boost/interprocess/containers/vector.hpp> #include <boost/interprocess/allocators/allocator.hpp> #include <string> #include <cstdlib> //std::system #include <conio.h> #include <iostream> #include <windows.h> using namespace boost::interprocess; //Define an STL compatible allocator of ints that allocates from the managed_shared_memory. //This allocator will allow placing containers in the segment typedef allocator<float, managed_shared_memory::segment_manager> ShmemAllocator; //Alias a vector that uses the previous STL-like allocator so that allocates //its values from the segment typedef vector<float, ShmemAllocator> MyVector; //Main function. For parent process argc == 1, for child process argc == 2 int main(int argc, char *argv[]) { if(argc == 1){ //Parent process //Remove shared memory on construction and destruction struct shm_remove { shm_remove() { shared_memory_object::remove("MySharedMemory"); } ~shm_remove(){ shared_memory_object::remove("MySharedMemory"); } } remover; //Create a new segment with given name and size managed_shared_memory segment(create_only, "MySharedMemory", 65536); //Initialize shared memory STL-compatible allocator const ShmemAllocator alloc_inst (segment.get_segment_manager()); //Construct a vector named "MyVector" in shared memory with argument alloc_inst MyVector *myvector = segment.construct<MyVector>("MyVector")(alloc_inst); for(int i = 0; i < 100; ++i) //Insert data in the vector myvector->push_back(0.0); // Loop bool exitFlag = false; while (1) { // Store 3 values myvector->at(0) = rand() % 100; myvector->at(1) = rand() % 100; myvector->at(2) = rand() % 100; std::cout << "Parent: " << myvector->at(0) << ", " << myvector->at(1) << ", " << myvector->at(2) << std::endl; if (kbhit()) { switch (getch()) { default: exitFlag = true; break; } if (exitFlag) break; } Sleep(500); } //When done, destroy the vector from the segment segment.destroy<MyVector>("MyVector"); } else { //Child process //Open the managed segment managed_shared_memory segment(open_only, "MySharedMemory"); //Find the vector using the c-string name MyVector *myvector = segment.find<MyVector>("MyVector").first; //Loop bool exitFlag = false; while (1) { try { std::cout << "Child: " << myvector->at(0) << ", " << myvector->at(1) << ", " << myvector->at(2) << std::endl; } catch (...) { std::cout << "Error accessing vector, exiting..." << std::endl; exitFlag = true; } if (kbhit()) { switch (getch()) { default: exitFlag = true; break; } } if (exitFlag) break; Sleep(500); } } return 0; }; -- View this message in context: http://boost.2283326.n4.nabble.com/Interprocess-managed-shared-memory-betwe… Sent from the Boost - Users mailing list archive at Nabble.com.

1 0

Behaviour of asio
by ustulation 17 Feb '13

17 Feb '13

By going through the doc. i understand that async completion handlers will be called whenever io_service::run() reads posted events from the queue for which the corresponding handlers were registered by us. Now if i have a registered handler for some event (say async_wait() of deadline_timer) which does something time consuming, then event loop (io_service::run()) does not get a chance to read the event queue until the handler returns thus preventing calls to other handlers for which events are queued. I've tested this behaviour with timers and it works as expected. However i could not understand the behaviour of async_read_from() of udp::socket. void A::read() { udpSoc.async_receive_from( boost::asio::buffer(m_cRxData, 1024), m_endpoint, boost::bind(&A::readHandler, this, _1, _2) ); } ////////////////////////////// <1> /////////////////////////// void A::readHandler(......) { std::string szDatagram(m_cRxData, bytesReceived); std::cerr << "Rxd Datagram: " << szDatagram << std::endl; read(); //............ } ////////////////////////////// <2> /////////////////////////// void A::readHandler(......) { read(); std::string szDatagram(m_cRxData, bytesReceived); std::cerr << "Rxd Datagram: " << szDatagram << std::endl; //............ } in above when 2 datagrams arrive simultaneously, readHandler <1> works fine but <2> does not. In <1> std::cerr write 1st and 2nd Datagrams distinctly, but in <2> data gets corrupted and std::cerr writes a string that would be formed if the 1st datagram in it were overwritten by the 2nd datagram (ie., if 1st one was 12345 and 2nd one xyz then <2> above prints "xyz45".) This means that clearly m_cRxData (which is a private member) is getting overwritten. Why does this happen? async_read_from should return immediately without reading into m_cRxData, readHandler(..) should continue and complete, and control should reach event loop, and then finally async read should be done calling the completion handler. Is this explanation wrong? -- View this message in context: http://boost.2283326.n4.nabble.com/Behaviour-of-asio-tp4642911.html Sent from the Boost - Users mailing list archive at Nabble.com.

3 2

Re: [Boost-users] Question about installing and using boost library.
by Giovanni 16 Feb '13

16 Feb '13

Greetings Pan, I am using Ubuntu 12.04 with one of my machines. I used the Ubuntu Software Center to install the Boost C++ Libraries, which makes things so much easier. You can enter "Boost" into the search box on the top-right corner, and find the Boost C++ Libraries to install. There are a number of them, including versions that conflict with each other. I chose the default version, and packages that don't conflict with that. I suggest that you install Boost C++ that way. After you have installed Boost C++ with Ubuntu Software Center, you should/"should" (it depends on how you set things up and if you have root access) see the following. The collection of header files for the Boost C++ Libraries in the location: /usr/lib. The collection of dynamic libraries (to be dynamically linked) for the Boost C++ Libraries in the location: /usr/include/boost. To test if you have installed the Boost C++ Libraries correctly, run a very simple C++ program that includes one of the Boost C++ Libraries and calls a function from that library. For example, you can try my attached toy program. To compile and run the program, try: g++ -o static2.out get_sys_info.cpp -lboost_filesystem -lboost_system ./static2.out As for your specific problem, I am guessing that it is a linking error that I also had. You need to include the Boost C++ Libraries at the end of the compilation command (for GCC/g++). Hence, I placed the filename (C++ source file) before including the libraries. I suggest that you post your compilation/linking commands, or makefile or script for build automation, so that people can help you. In addition, I would include the relevant source code (make it a simple C++ program, like my example) to test your Boost C++ installation. Furthermore, I would include the transcript from the Terminal. Just copy and paste what appears on the Terminal as you compile the program. Good luck, and have a great weekend! Regards On Fri, Feb 15, 2013 at 8:47 AM, pan yi <outlook_eye(a)yahoo.com.cn> wrote: > Hi all, > > I'm a beginner to use boost library for my c++ coding project. After I > include the <boost/algorithm/string.hpp> in my cpp file and compile the > code, the compiling error indicates that: > fatal error: boost/algorithm/string.hpp: No such file or directory > I search on internet and knows that I need to install some boost library. > But which library should I install? (I tried to install the whole library > with "sudo apt-get install libboost*", but it doesn't work.) > > Can any one help me? I'm compiling and executing c++ on ubuntu Version12.04 > environment > > Thank you so much. > > Best Regards > Pan

2 1

Re: [Boost-users] [Chrono] Memory leak in Chrono?
by Don 16 Feb '13

16 Feb '13

On 2/16/2013 Vincent wrote: > Please, could you create a Trac ticket so this issue not lost? Submitted as ticket #8079 https://svn.boost.org/trac/boost/ticket/8079 Thanks, Don

1 0

Re: [Boost-users] [Chrono] Memory leak in Chrono?
by Don 16 Feb '13

16 Feb '13

-------- Original Message -------- Subject: Boost-users Digest, Vol 3358, Issue 4 From: boost-users-request(a)lists.boost.org To: boost-users(a)lists.boost.org Date: 2/13/2013 1:33 PM > Subject: Re: [Boost-users] [Chrono] Memory leak in Chrono? > From: "Vicente J. Botet Escriba" <vicente.botet(a)wanadoo.fr> > Date: 2/13/2013 1:11 PM > > Le 13/02/13 14:30, Don a écrit : >>> Subject: Re: [Boost-users] [Chrono] Memory leak in Chrono? >>> From: "Vicente J. Botet Escriba" <vicente.botet(a)wanadoo.fr> >>> Date: 2/12/2013 1:17 PM >>> >>> Le 12/02/13 18:13, gast128 a écrit : >>>> we use Boost.Chrono too. MFC applications incorrectly report memory leaks, which can easily be solved. >>> Hi, >>> >>> if you have some patches to propose to avoids this incorrect reports I would try to include them. >> >> Vincente, >> >> I am using MFC, but I don't believe this is a MFC false-leak report. >> >> I added a destructor for duration_units_default_initializer_t, like this (near the end of chrono/io/duration_units.hpp): >> >> ~duration_units_default_initializer_t() >> { >> if (duration_units_default_holder<CharT>::initialized_) >> { >> delete[] duration_units_default_holder<CharT>::n_d_valid_units_; >> duration_units_default_holder<CharT>::n_d_valid_units_ = 0; >> delete[] duration_units_default_holder<CharT>::valid_units_; >> duration_units_default_holder<CharT>::valid_units_ = 0; >> duration_units_default_holder<CharT>::initialized_ = false; >> } >> } >> >> This solves the memory leak, and seems to work fine for my simple test program. > I could add this is this helps on the MFC report. Thanks, that would be useful. It would probably help with other memory leak detectors, as well. Since you are allocating heap that you are never freeing, most leak detectors are going to report this. >> But n_d_valid_units_, valid_units_, and initialized_ are static >> variables, so I assume you want to only create these variables >> once. > > Right. > >> When run with my test program and step through it in the debugger, >> only one instantiation of duration_units_default_initializer_t >> occurs. (That is, once for each char type, single-byte and wide.) > This is the goal of the type (implement a singleton). This technique > is used in some Boost libraries already > >> This occurs when the duration_units_default_initializer is >> instantiated (at the very end of duration_units.hpp). Since this is >> a global variable, this occurs during application startup and the >> destructor runs during application shuttdown. > > Right, this was the reason the destructor didn't do anything. >> >> But presumably this can occur in some other situations, otherwise >> why would you have all the logic of making valid_units_ etc. >> class-static variables.> > On which other situation do you think this could occur? I have no idea, that's what confused me. If duration_units_default_initializer_t is only instantiated as a global (static) variable), then I don't see any reason for making n_d_valid_units_, valid_units_, and initialized_ static variables -- they could be regular member variables. Making them static doesn't hurt, but it's confusing (to me at least) because it looks like you want to share those variables among multiple instantiations, and there is only ever one instantiation. If one can somehow instantiate multiple duration_units_default_initializer_t objects, then making the member variables static does make this behave as a singleton. In that case the simple destructor I wrote will not work, since it will delete the member variables when the first object destructs, not the last. In that case I would use a reference count or smart pointers to only delete when the final object destructs. Thanks for your help, I appreciate your spending the time to answer my questions. Don

2 1