Recently (~10 days ago), 26_numerics/complex/13450.cc started failing for *many* targets, among which powerpc-darwin. On the latter I'm seeing a miscompilation at -O1 and above of this reduced C++ snippet:
#include <math.h> #include <stdio.h> int main() { __complex__ double x; __real__ x = -1.0; __imag__ x = 0.0; __complex__ double t = __builtin_clog(x); double tmp = 0.5 * (__imag__ t); printf("%g %g\n", cos(tmp), sin(tmp)); } ///////////////// Note how the output changes at -O1 and above. Also note that the very same snippet, as C is fine, thus I'm categorizing as C++, not as middle-end or something else. -- Summary: [4.3 Regression] Recent complex miscompilation Product: gcc Version: 4.3.0 Status: UNCONFIRMED Severity: critical Priority: P3 Component: c++ AssignedTo: unassigned at gcc dot gnu dot org ReportedBy: pcarlini at suse dot de GCC target triplet: powerpc-apple-darwin8.8.0 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30980