http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49988
Summary: constexpr on ctor invokes improper initialization Product: gcc Version: 4.6.1 Status: UNCONFIRMED Severity: normal Priority: P3 Component: c++ AssignedTo: unassig...@gcc.gnu.org ReportedBy: fais...@yahoo.com CC: ja...@gcc.gnu.org The following code compiles fine, but displays incorrect (or at the very least inconsistent) behavior if constexpr is applied to A's ctor: template<int ... I> struct X { }; struct A { char data[3]; template<int ... I> constexpr // works if you comment this constexpr out A(const char (&s)[3], X<I...> x) : data{ s[I]...} { } // above also works if you leave constexpr in but replace // const char (&s)[3] with const char* s. }; struct B { A a; B(const char (&s)[3]) : a{s,X<0,1,2>{}} { } }; int main() { B b{"12"}; assert(b.a.data[0] == '1'); // assertion fails at run time } Compiled with gcc4.6.1 with *JUST* the following recent patch MANUALLY applied: http://gcc.gnu.org/viewcvs/branches/gcc-4_6-branch/gcc/cp/semantics.c?r1=177188&r2=177187&pathrev=177188 (please refer to PR c++/49924 for details) i.e gcc -v still reads (compared to PR c++/49924): Using built-in specs. COLLECT_GCC=g++-4.6.1 COLLECT_LTO_WRAPPER=/usr/local/gcc4.6.1/libexec/gcc/i686-pc-linux-gnu/4.6.1/lto-wrapper Target: i686-pc-linux-gnu Configured with: ../gcc-4.6.1/configure --prefix=/usr/local/gcc4.6.1 --enable-generated-files-in-srcdir --enable-threads=posix --with-system-zlib --enable-__cxa_atexit --disable-checking --disable-nls --enable-bootstrap --with-gcc --with-gnu-as --with-gnu-ld --disable-multilib Thread model: posix gcc version 4.6.1 (GCC) I wonder if just applying that one patch manually, while fixing one bug, could have introduced another? Thanks