https://bugzilla.novell.com/show_bug.cgi?id=433516 User matz@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=433516#c2 Michael Matz changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |ASSIGNED --- Comment #2 from Michael Matz 2008-10-29 09:11:18 MDT --- Doesn't work in 11.0 for me (i.e. gives no warning), not even with gcc4.2. But it does work when I additionally specify -fno-builtin-memset. The problem is generally that if the built-in is active it comes before expanding the inline function from glibc's string3.h. And that builtin directly ignores when it sees the length of 0. What _does_ work, though, is that if the length is constant but in a variable, so that the constness is only found out during optimization and not already when expanding the builtin: int main(int argc, char **x) { int l = 0; memset(buf,20,l); return 0; } Reading also older GCC sources I don't think this ever worked with literal zeros when the built-in was still active. Do you still have a 10.3 installation somewhere, where I could check? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=433516 User matz@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=433516#c4 Michael Matz changed: What |Removed |Added ---------------------------------------------------------------------------- Priority|P5 - None |P4 - Low --- Comment #4 from Michael Matz 2008-10-30 07:22:08 MDT --- Ahaaa! The difference are the libc headers. On 10.3 memset() is a macro(!), on 11.0 (and later) it's an inline function. This was changed for C++ compatibility, so won't be changed back. The difference then is, that on 10.3 the compiler doesn't even see the memset call from the original source, but directly the conditional, whereas on 11.0 that conditional is only there when the inline expander runs, which comes after built-in handling. This is harder to fix: either we would have to expand inline functions before built-ins, which I think is not going to take place, or the warning has to be implemented in GCC :-/ Lowering priority. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.