C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  Derived constructor throwed, base did not destroy          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message wij@seed.net.tw Guest Posted: Sun Mar 20, 2005 1:59 am    Post subject: Derived constructor throwed, base did not destroy Hi All: I searched the google for this question, the answers found says all constructed objects will be destroyed. But the following codes does not seem to agree. It is the same if ~D throws. The compiler is g++. Thank you. class B { public: B() { std::cout << "B"; } ~B() { std::cout << "~B"; } }; class D : public B { public: D() : B() { std::cout << "D"; // Test showed ~B is not called. Program displays "BD" throw int(1); // ~B() is not called } ~D() { std::cout << "~D"; // Test showed ~B is not called // throw int(0); // ~B() is not called } }; int main(void) { D d; return(0); }; [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Ivan Vecerina Guest Posted: Sun Mar 20, 2005 9:10 am    Post subject: Re: Derived constructor throwed, base did not destroy <wij (AT) seed (DOT) net.tw> wrote Quote: I searched the google for this question, the answers found says all constructed objects will be destroyed. But the following codes does not seem to agree. This is because your sample code triggers a special case: the exception being thrown is not caught, so terminate() will be called and the program will exit immediately. In this situation, it is unspecified whether stack unwinding occurs before the call to terminate or not. If you modify the code as follows to avoid program termination, things will happen as expected: int main(void) { try{ D d; } catch(...) { } return(0); }; Ivan -- http://ivan.vecerina.com/contact/?subject=NG_POST <- email contact form Brainbench MVP for C++ <> http://www.brainbench.com [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top pranjal_srivastava@hotmai Guest Posted: Mon Mar 21, 2005 4:16 pm    Post subject: Re: Derived constructor throwed, base did not destroy Hi, If a exception is thrown while constructing an object...its Dtor is never called. You have to specifically handle all the exceptions in the Ctor. Same is the case here....The ctor of D throws exception so dtor will never be called. -Pranjal [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Alberto Barbati Guest Posted: Tue Mar 22, 2005 1:52 am    Post subject: Re: Derived constructor throwed, base did not destroy [email]pranjal_srivastava (AT) hotmail (DOT) com[/email] wrote: Quote: Hi, If a exception is thrown while constructing an object...its Dtor is never called. You have to specifically handle all the exceptions in the Ctor. Same is the case here....The ctor of D throws exception so dtor will never be called. If the ctor of D throws, then the *dtor of D* will never be called, as you say. But we are talking about the *dtor of B*, here. The dtor or B *will* be called, because B is a fully constructed sub-object once the flow enters the body of D: (§15.2/2). I agree with Ivan Vecerina that the problem must be caused by the lack of a try-block, a fact that allows the program to call terminate() without unwinding the stack (§15.5.1/2). Alberto [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Vinzenz Feenstra Guest Posted: Fri Mar 25, 2005 11:53 am    Post subject: Re: Derived constructor throwed, base did not destroy Hi, the real reason is, that you don't see the complete output, because std::cout is buffered. Try the same test with std::cerr instead of std::cout and you'll see that the D'tors will be called. BR Vinzenz [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top wij@seed.net.tw Guest Posted: Sat Mar 26, 2005 4:52 am    Post subject: Re: Derived constructor throwed, base did not destroy Quote: ...the exception being thrown is not caught, so terminate() will be called and the program will exit immediately. In this situation, it is unspecified whether stack unwinding occurs before the call to terminate or not. Thanks for the answer. I made an experiment to confirm a consequence(see below). Hope this will be fixed, dont know which part is responsible, though. class T { public: T() { std::cerr << "T"; } ~T() { std::cerr << "~T"; } }; void g(void) { D d; } void f(void) { T t; g(); } int main(void) { f(); }; The program output "TBD"(terminated), instead of expected TBD~B~T [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top wij@seed.net.tw Guest Posted: Sat Mar 26, 2005 9:30 am    Post subject: Re: Derived constructor throwed, base did not destroy Quote: Try the same test with std::cerr instead of std::cout and you'll see that the D'tors will be called. Using std::cerr results the same on my machine. This may be valid though. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Ivan Vecerina Guest Posted: Sat Mar 26, 2005 9:47 am    Post subject: Re: Derived constructor throwed, base did not destroy "Vinzenz Feenstra" <evilissimo (AT) web (DOT) de> wrote Quote: the real reason is, that you don't see the complete output, because std::cout is buffered. Not correct. The fact is that, if the program is terminated because the exception was not caught, it is implementation-defined whether any destructors of stack-based objects will be called or not. Here's the Chapter and verse in the C++ standard: 15.3/9 If no matching handler is found in a program, the function terminate() is called; whether or not the stack is unwound before this call to terminate() is implementation defined Quote: Try the same test with std::cerr instead of std::cout and you'll see that the D'tors will be called. Of course, buffering of std::cout also could be causing the symptom, but eventually the previous output would be truncated as well. Regards, Ivan -- http://ivan.vecerina.com/contact/?subject=NG_POST <- email contact form [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Ivan Vecerina Guest Posted: Sat Mar 26, 2005 5:38 pm    Post subject: Re: Derived constructor throwed, base did not destroy <wij (AT) seed (DOT) net.tw> wrote Quote: ...the exception being thrown is not caught, so terminate() will be called and the program will exit immediately. In this situation, it is unspecified whether stack unwinding occurs before the call to terminate or not. Thanks for the answer. I made an experiment to confirm a consequence(see below). Hope this will be fixed, dont know which part is responsible, though. I'm not sure what you meant by "this will be fixed". The behavior you are observing is conform to the standard, and is reasonable IMO. terminate() is an extreme solution used by the runtime when it sees nothing left to do to recover from an error. Not catching an exception that may be thrown is the real bug in the program. If you want to ensure that the destructors are called, just add a try-catch block in your main program, as suggested in my previous post. Again: int main() { try { f(); } catch(...) { return EXIT_FAILURE; } return EXIT_SUCCESS; } Ivan -- http://ivan.vecerina.com/contact/?subject=NG_POST <- email contact form [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Alberto Barbati Guest Posted: Sun Mar 27, 2005 2:19 am    Post subject: Re: Derived constructor throwed, base did not destroy [email]wij (AT) seed (DOT) net.tw[/email] wrote: Quote: int main(void) { f(); }; The program output "TBD"(terminated), instead of expected TBD~B~T What's that going to prove? I repeat, it's the lack of the try-block that causes the problem. In fact if you replace function main with int main() { try { f(); } catch(...) {} } you get the result you expect (tried on VC++7.1, but you should get the same output on any conformant compiler). Alberto [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top wij@seed.net.tw Guest Posted: Mon Mar 28, 2005 1:33 am    Post subject: Re: Derived constructor throwed, base did not destroy Quote: I'm not sure what you meant by "this will be fixed". The behavior you are observing is conform to the standard, and is reasonable IMO. terminate() is an extreme solution used by the runtime when it sees nothing left to do to recover from an error. So, I don't have to implement every class like this: class T { T() try { ... any_function(); } catch(...) { // *** Disclaimer *** // Throwing in the above try block is conditionally caught. // The following is executed depending not on this class. // User of this class should have a catch handler (need not // to be relavent to this internal handler) ... }; }; Quote: IMO. terminate() is an extreme solution used by the runtime terminate() is OK for me (I don't consider it extreme by itself) What does IMO represent? I am new on the internet world, there's lots of these! {"IMO" == "In my opinion". You might want to have a look at the Jargon File for information on acronyms: <http://www.catb.org/~esr/jargon/>, although it only lists the form "IMHO"... -mod/dk} Quote: Not catching an exception that may be thrown is the real bug in the program. Agree. But I did as in the example above. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top wij@seed.net.tw Guest Posted: Mon Mar 28, 2005 1:43 am    Post subject: Re: Derived constructor throwed, base did not destroy Quote: What's that going to prove? I repeat, it's the lack of the try-block ... I can't prove anything but to figure out an consequence from the given answer to work around my encounter. I got it, the consequence is byproduct. Please see also the others in this thread. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Ivan Vecerina Guest Posted: Mon Mar 28, 2005 12:04 pm    Post subject: Re: Derived constructor throwed, base did not destroy <wij (AT) seed (DOT) net.tw> wrote Quote: I'm not sure what you meant by "this will be fixed". The behavior you are observing is conform to the standard, and is reasonable IMO. terminate() is an extreme solution used by the runtime when it sees nothing left to do to recover from an error. So, I don't have to implement every class like this: class T { T() try { ... any_function(); } catch(...) { // *** Disclaimer *** // Throwing in the above try block is conditionally caught. // The following is executed depending not on this class. // User of this class should have a catch handler (need not // to be relavent to this internal handler) ... }; }; What is the above intended to achieve? All you need is 1 exception handler, for example in the main() function (as in the example I wrote). The destructors of all stack-allocated objects will then be called if an exception is thrown. And this is all standard practice and standard behavior - nothing worth a specific comment or disclaimer in the implementation of a class. Quote: IMO. terminate() is an extreme solution used by the runtime terminate() is OK for me (I don't consider it extreme by itself) terminate() is ok if you do not need any destructors to be called. Regards, Ivan -- http://ivan.vecerina.com/contact/?subject=NG_POST <- email contact form [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top wij@seed.net.tw Guest Posted: Mon Mar 28, 2005 11:48 pm    Post subject: Re: Derived constructor throwed, base did not destroy The code sample was a big mistake (out of overlook in mixed evaluation) I apologize the disturbance. Quote: What is the above intended to achieve? A bad respond to "> I'm not sure what you meant by "this will be fixed"" It should be the 'using logic when destructors will be called' which troubled me. (It probably not good out of here without leaving things clear, if this have not been disscussed. I skipped this topic in learning C++.) Quote 15.3/9: If no matching handler is found in a program, the function terminate() is called; whether or not the stack is unwound before this call to terminate() is implementation defined. What about: 1. Rephrasing for use of 15.3/9 If no matching handler is found, destructors in the call chain may or may not be called(stack unwinding), program behavior relying on stack unwinding is in this case implementation-defined, thus not standard conforming. 2. Consequence To be standard conforming, a program should provide handler 'catch(...)' in main(..) to catch all possible object thrown. (not just catch specifics) [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Display posts from previous: All Posts1 Day7 Days2 Weeks1 Month3 Months6 Months1 Year Oldest FirstNewest First         C++Talk.NET Forum Index -> C++ Language (Moderated) All times are GMT Page 1 of 1   Jump to: Select a forum C/C++----------------C LanguageC Language (Moderated)C++ Language (Moderated)C++ language (comp.lang.c++)C++ language, library and standards Non-English C/C++ Newsgroups----------------C++ (German)C++ (French)  You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum Powered by phpBB © 2001, 2006 phpBB Group SEO toolkit © 2004-2006 webmedic.