Hi! The following code snippet leads to an ICE, but I am not sure whether the stuff I am trying to do is valid. Any pointer to info about that also appreciated. Target: i586-suse-linux Configured with: ../configure --enable-threads=posix --prefix=/usr --with-local-prefix=/usr/local --infodir=/usr/share/info --mandir=/usr/share/man --libdir=/usr/lib --libexecdir=/usr/lib --enable-languages=c,c++,objc,fortran,obj-c++,java,ada --enable-checking=release --with-gxx-include-dir=/usr/include/c++/4.1.2 --enable-ssp --disable-libssp --disable-libgcj --with-slibdir=/lib --with-system-zlib --enable-shared --enable-__cxa_atexit --enable-libstdcxx-allocator=new --program-suffix=-4.1 --enable-version-specific-runtime-libs --without-system-libunwind --with-cpu=generic --host=i586-suse-linux Thread model: posix gcc version 4.1.2 20061115 (prerelease) (SUSE Linux) enum Outer { a, b }; template <class T> class C { public: enum Inner { c, d }; template<Inner I, int dummy> struct Dispatcher; template<int dummy> struct Dispatcher<c, dummy> { static const Outer Show = a; }; template<int dummy> struct Dispatcher<d, dummy> { static const Outer Show = b; }; public: template <Inner I> void DoSomething(int j) { typedef Dispatcher<I, 1> dispatcher_t; static const Outer o = dispatcher_t::Show; } }; int main() { C<double> Test; Test.DoSomething<C<double>::c>(1); }
The trunk gives: t.cc: In member function 'void C<T>::DoSomething(int) [with C<T>::Inner I = c, T = double]': t.cc:32: instantiated from here t.cc:24: error: incomplete type 'C<double>::Dispatcher<c, 1>' used in nested name specifier Which seems likes the correct error message.
The compiler at http://www.comeaucomputing.com/tryitout/ has absolutely no problems with the code. M$-VC8 says: error C2754: 'C<T>::Dispatcher<c,dummy>' : a partial specialization cannot have a dependent non-type template parameter which somehow is an error message that clearly states what might be wrong. The "incomplete type" statement is somehow misleading - IMHO Could you change the error message in this case? Of course I disagree with gcc and VC8 here. Any pointer to a TR on this subject? Best Regards, Markus
I started a discussion about this code on comp.lang.c++.moderated, subject "enum in template memfun in class template - valid C++?" see <http://groups.google.com/groups/profile?enc_user=IUBmLhUAAACyipKMT4DNptGfylSI-1gK9h3i3SmjGmAJbX05nZ-8fQ> Your turn. Markus
Sorry, the link was wrong, must be http://groups.google.de/group/comp.lang.c++.moderated/browse_thread/thread/8c3b8a84ed78b003/4d9603171894a75d?hl=de#4d9603171894a75d
compiles without error using 4.4.3 or 4.6.0
Hi! Though I love the fact that this code now compiles, I am still unsure whether this is the right thing to have. From http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#502 I see that the proposed resolution (February, 2010) stands against what David Vandevoorde said in http://groups.google.de/group/comp.lang.c++.moderated/browse_thread/thread/8c3b8a84ed78b003/4d9603171894a75d?hl=de#4d9603171894a75d and the enum now is dependent and so IMHO the code should not compile. Maybe I missed something. Could you please shed some more light on this? Markus Werle
I didn't realise there was a DR about this, confirm ...
... and suspend until the issue is ready
Hi! Looks like http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#502 now has a resolution, namely that a type is dependent if it is a nested class or enumeration that is a member of the current instantiation. So though the voting might not be through, the compilers could already adopt the proposed behavior and give some informational error message. -> Change Status back to new? Markus
unsuspending, DR 502 is part of the FDIS
Summary changed, as there's no ICE now. This is 'accepts-invalid' because DR502 means C<T>::Inner is dependent, and [temp.class.spec]/8 says specialized non-type arguments shall not have dependent types. Reduced: template <class T> struct C { enum Inner { c }; template<Inner I, int dummy> struct Dispatcher; template<int dummy> struct Dispatcher<c, dummy> { }; }; int main() { C<double> Test; } expected error would be something like: error: type 'C<T>::Inner' of template argument 'c' depends on a template parameter
GCC, ICC, clang and MSVC all accept the testcase in comment #11.
(In reply to Andrew Pinski from comment #12) > GCC, ICC, clang and MSVC all accept the testcase in comment #11. I believe the testcase is valid. The instantiation of `C<double>` looks like this: struct C<double> { enum Inner { c }; template<Inner I, int dummy> struct Dispatcher; template<int dummy> struct Dispatcher<c, dummy> { }; }; ... and the template argument `c` here refers to the non-dependent enumeration constant `C<double>::c`. The rule governing whether the original template `C` is valid is [temp.res.general]/6 (https://eel.is/c++draft/temp.res.general#6), and in particular: > no diagnostic shall be issued for a template for which a valid specialization can be generated Because `C` has valid specializations, it's valid.
So there's no bug now then. The error in comment 1 was probably fixed by: commit g:dc28490d05f5ef25c5ff24ef946f48f402a20be8 Author: Jason Merrill Date: Tue Sep 4 13:27:21 2007 re PR c++/14032 (Specialization of inner template using outer template argument doesn't work) PR c++/14032 * pt.c (most_specialized_class): Substitute outer template arguments into the arguments of a member template partial specialization. (strip_innermost_template_args): New fn. From-SVN: r128076 And so this was fixed for GCC 4.3.0