C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  Type safe pointers          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message John Potter Guest Posted: Sat Aug 23, 2003 8:13 am    Post subject: Type safe pointers Looking for expectations for an implementation which maintains type information for pointers and reports misuse when dereferenced. long li(42); short si; For each of the following, indicate performs, error, undefined behavior. Note that unknown value is different from undefined behavior. si = li; si = *reinterpret_cast<short*>(&li); si = *static_cast<short*>(static_cast<void*>(&li)); Since this is a discussion group, why is also of interest. Thanks, John [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top John Potter Guest Posted: Sun Aug 24, 2003 11:34 pm    Post subject: Re: Type safe pointers On 23 Aug 2003 13:56:41 -0400, [email]pdimov (AT) mmltd (DOT) net[/email] (Peter Dimov) wrote: Quote: John Potter <jpotter (AT) falcon (DOT) lhup.edu> wrote Looking for expectations for an implementation which maintains type information for pointers and reports misuse when dereferenced. Note the above. Quote: si = *reinterpret_cast<short*>(&li); si = *static_cast<short*>(static_cast<void*>(&li)); Undefined behavior. 3.10/15. You expect the implementation to fail to detect and report either? John [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Shay Guest Posted: Mon Aug 25, 2003 10:11 pm    Post subject: Re: Type safe pointers In article <g02dkvk3s6uikrjnq4ndblc0mhkmlpvju3 (AT) 4ax (DOT) com>, [email]jpotter (AT) penguin (DOT) lhup.edu[/email] wrote: Quote: Looking for expectations for an implementation which maintains type information for pointers and reports misuse when dereferenced. long li(42); short si; [snip] si = li; assert( si == 42 ); Quote: si = *reinterpret_cast<short*>(&li); reinterpret_cast<> says we know what we're doing, so do what's most natural for the implementation. The value depends on the architecture, most likely 0 or 42 depending on endian-ness. Quote: si = *static_cast<short*>(static_cast<void*>(&li)); Here we're lying, so an implementation which maintained type information for void* should give us a runtime error even if there is no dereference. Basically I would expect void* to behave similar to this: class void_ptr { void* p; std::type_info const* type; public: template<typename T> void_ptr( T* t ) : p( t ), type( &typeid (t) ) { } template<typename T> operator T () const { assert( typeid (T) == *type ); return static_cast<T> (p); } }; The alternative, to delay the error until dereference, would require all pointer types to be able to hold a pointer to any type of object. I have something like this in mind: unsigned const seg_size = 0x1000000; class segment_ptr { char seg; public: segment_ptr( void* p ) : seg( reinterpret_cast<unsigned> (p) / seg_size ) { assert( *this == p ); } operator void* () const { return reinterpret_cast<void*> (seg * seg_size); } }; -- Shay [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Peter Dimov Guest Posted: Mon Aug 25, 2003 10:27 pm    Post subject: Re: Type safe pointers John Potter <jpotter (AT) falcon (DOT) lhup.edu> wrote Quote: On 23 Aug 2003 13:56:41 -0400, [email]pdimov (AT) mmltd (DOT) net[/email] (Peter Dimov) wrote: John Potter <jpotter (AT) falcon (DOT) lhup.edu> wrote Looking for expectations for an implementation which maintains type information for pointers and reports misuse when dereferenced. Note the above. Noted, although it's somewhat ambiguous; is the type information maintained by the compiler or are pointers "fat"? Quote: si = *reinterpret_cast<short*>(&li); si = *static_cast<short*>(static_cast<void*>(&li)); Undefined behavior. 3.10/15. You expect the implementation to fail to detect and report either? Hm. Considering "maintains type information for pointers and reports misuse when dereferenced" I would expect a compile-time warning and/or a runtime error in both cases. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top llewelly Guest Posted: Tue Aug 26, 2003 12:39 am    Post subject: Re: Type safe pointers John Potter <jpotter (AT) falcon (DOT) lhup.edu> writes: Quote: Looking for expectations for an implementation which maintains type information for pointers and reports misuse when dereferenced. long li(42); short si; For each of the following, indicate performs, error, undefined behavior. Note that unknown value is different from undefined behavior. si = li; si = *reinterpret_cast<short*>(&li); I would expect this to result in si containing an implementation-defined value, 'unsurprising' to one familar with the implementation's representations of both long and short, endianess, etc. I would expect no warnings or errors from this. Quote: si = *static_cast<short*>(static_cast<void*>(&li)); [snip] I assume this example is intended to represent cases where the two static_cast's occur in seperate hunks of code as well. I would expect a runtime report of misuse for this. Conceptually, I see static_cast as a tool which reverses a conversion. The static_cast to short* does not reverse the original conversion - instead it converts to an unrelated type. Note it is short* and long* I consider unrelated, not short and long, which are obviously related. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Michiel Salters Guest Posted: Tue Aug 26, 2003 12:45 am    Post subject: Re: Type safe pointers John Potter <jpotter (AT) falcon (DOT) lhup.edu> wrote Quote: Looking for expectations for an implementation which maintains type information for pointers and reports misuse when dereferenced. long li(42); short si; For each of the following, indicate performs, error, undefined behavior. Note that unknown value is different from undefined behavior. si = li; // Works si = *reinterpret_cast<short*>(&li); // Error (runtime) si = *static_cast<short*>(static_cast<void*>(&li)); // Error (runtime) Since this is a discussion group, why is also of interest. The error cases are because such a hypothetical implementation will have access to the dynamic type of each object, i.e. given a void* it will be able to determine the type(s) that it points to. These cases are easy, in that a short and a long are sufficiently different. It would have been harder if we had a struct {long}* Regards, -- Michiel Salters [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top llewelly Guest Posted: Tue Aug 26, 2003 7:07 pm    Post subject: Re: Type safe pointers [email]pdimov (AT) mmltd (DOT) net[/email] (Peter Dimov) writes: Quote: John Potter <jpotter (AT) falcon (DOT) lhup.edu> wrote On 23 Aug 2003 13:56:41 -0400, [email]pdimov (AT) mmltd (DOT) net[/email] (Peter Dimov) wrote: John Potter <jpotter (AT) falcon (DOT) lhup.edu> wrote Looking for expectations for an implementation which maintains type information for pointers and reports misuse when dereferenced. Note the above. Noted, although it's somewhat ambiguous; is the type information maintained by the compiler or are pointers "fat"? si = *reinterpret_cast<short*>(&li); si = *static_cast<short*>(static_cast<void*>(&li)); Undefined behavior. 3.10/15. You expect the implementation to fail to detect and report either? Hm. Considering "maintains type information for pointers and reports misuse when dereferenced" I would expect a compile-time warning and/or a runtime error in both cases. What about the cultural notion that a cast is 'a way to tell the implementation I know what I'm doing' ? [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Peter Dimov Guest Posted: Wed Aug 27, 2003 10:12 pm    Post subject: Re: Type safe pointers llewelly <llewelly.at (AT) xmission (DOT) dot.com> wrote Quote: pdimov (AT) mmltd (DOT) net (Peter Dimov) writes: John Potter <jpotter (AT) falcon (DOT) lhup.edu> wrote On 23 Aug 2003 13:56:41 -0400, [email]pdimov (AT) mmltd (DOT) net[/email] (Peter Dimov) wrote: John Potter <jpotter (AT) falcon (DOT) lhup.edu> wrote Looking for expectations for an implementation which maintains type information for pointers and reports misuse when dereferenced. Note the above. Noted, although it's somewhat ambiguous; is the type information maintained by the compiler or are pointers "fat"? si = *reinterpret_cast<short*>(&li); si = *static_cast<short*>(static_cast<void*>(&li)); Undefined behavior. 3.10/15. You expect the implementation to fail to detect and report either? Hm. Considering "maintains type information for pointers and reports misuse when dereferenced" I would expect a compile-time warning and/or a runtime error in both cases. What about the cultural notion that a cast is 'a way to tell the implementation I know what I'm doing' ? A cast is a way to tell the implementation that you know what it does. Testing the code on another implementation, one specifically intended to catch pointer misuse, implies that you want to know whether your code is portable, i.e. does not take advantage of undefined behavior that happens to do the right thing on your platform. That's how I understand it, at least. I may be missing something. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Stefan Heinzmann Guest Posted: Fri Aug 29, 2003 10:57 am    Post subject: Re: Type safe pointers John Potter wrote: Quote: Looking for expectations for an implementation which maintains type information for pointers and reports misuse when dereferenced. Even after reading some other postings I'm not sure whether you want the compiler or the runtime system to keep type info for pointers. I'm leaning towards the latter, but maybe you should make it a bit clearer what your aim is. Quote: long li(42); short si; For each of the following, indicate performs, error, undefined behavior. Note that unknown value is different from undefined behavior. si = li; Performs, but I would expect a compiler warning (as is the case with many a compiler) Quote: si = *reinterpret_cast<short*>(&li); Neither compiler error nor runtime error, but undefined behaviour. The reinterpret_cast is the trust-me-blindly cast for me that defeats any type checking except for const qualification. I would expect the content of si to depend on the machine's endianness, but for odd machines I would think that worse effects are possible (such as access violations). Quote: si = *static_cast<short*>(static_cast<void*>(&li)); Given that you're really checking at runtime, I'd report misuse here. I wouldn't report misuse if the outer cast were to char* instead of short*, since I would see this as a legitimate way to get at the machine representation of the long. But I'm dithering here. Runtime checking of void* in C++ is a dodgy notion to me. Cheers Stefan [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Anthony Williams Guest Posted: Fri Aug 29, 2003 3:17 pm    Post subject: Re: Type safe pointers llewelly <llewelly.at (AT) xmission (DOT) dot.com> writes: Quote: John Potter <jpotter (AT) falcon (DOT) lhup.edu> writes: Looking for expectations for an implementation which maintains type information for pointers and reports misuse when dereferenced. long li(42); short si; For each of the following, indicate performs, error, undefined behavior. Note that unknown value is different from undefined behavior. si = li; si = *reinterpret_cast<short*>(&li); I would expect this to result in si containing an implementation-defined value, 'unsurprising' to one familar with the implementation's representations of both long and short, endianess, etc. I would expect no warnings or errors from this. Agreed. Quote: si = *static_cast<short*>(static_cast<void*>(&li)); [snip] I assume this example is intended to represent cases where the two static_cast's occur in seperate hunks of code as well. I would expect a runtime report of misuse for this. I would expect the runtime error to occur with the dereference, not the cast. Quote: Conceptually, I see static_cast as a tool which reverses a conversion. The static_cast to short* does not reverse the original conversion - instead it converts to an unrelated type. Note it is short* and long* I consider unrelated, not short and long, which are obviously related. I am inclined to think that the following is valid: assert(static_cast<long*>( static_cast<void*>( static_cast<short*>( static_cast<void*>(&li))))==&li); as round-tripping through void* is permitted, and short has lesser alignment requirements than long on most (all?) platforms. However, I would expect assert(static_cast<short*>( static_cast<void*>( static_cast<long*>( static_cast<void*>(&si))))==&si); To fail some of the time, on some implementations, where long has stricter alignment requirements, because the cast to long* will essentially lose the information, e.g. by masking the value. Indeed, I would be unsurprised by a runtime error if the actual short was at an address that couldn't be cast to a long*. The code should compile, though. Anthony -- Anthony Williams Senior Software Engineer, Beran Instruments Ltd. Remove NOSPAM when replying, for timely response. [ 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.