http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55771
Bug #: 55771 Summary: Negation and type conversion incorrectly exchanged Classification: Unclassified Product: gcc Version: 4.8.0 Status: UNCONFIRMED Keywords: wrong-code Severity: normal Priority: P3 Component: c AssignedTo: unassig...@gcc.gnu.org ReportedBy: i...@airs.com This program should print the same thing twice: #include <stdio.h> void f1() { unsigned long x = 3; float y = 1; printf ("%g\n", (-x) * y); } void f2() { unsigned long x = 3; float y = 1; unsigned long z = - x; printf ("%g\n", z * y); } int main() { f1(); f2(); } However, on x86_64 GNU/Linux with current mainline, it prints -3 1.84467e+19 It is already incorrect in the first GIMPLE dump. f1 has x = 3; y = 1.0e+0; D.2219 = (float) x; D.2220 = -D.2219; f2 has x = 3; y = 1.0e+0; z = -x; D.2223 = (float) z; In other words, in f1, the conversion to float happens before the negation. The bug happens with both the C and C++ frontends.