On Fri, 24 Jun 2005, ov wrote: > * What are the values of x, y, and mod when the assertion is hit? > 32, -1, 64
This is quite strange because these values are nominal -- they shouldn't have triggered the assertion. The assertion statement says: ASSERT (0 <= x && x < mod && -mod <= y && y <= mod); substitution gives: 0 <= 32 && 32 < 64 && -64 <= -1 && -1 <= 64 which is a true statement when x, y, and mod are integers, as they are declared to be. For example, when I do the same test on x86 with modulo_add instrumentation added, I see the same 32/-1/64 parameter set. Fri Jun 24 23:58:21 2005 TESTING ENCRYPT/DECRYPT of packet length=97 Fri Jun 24 23:58:21 2005 modulo_add(int x=92, int y=1, int mod=100) Fri Jun 24 23:58:21 2005 modulo_add(int x=32, int y=-1, int mod=64) Fri Jun 24 23:58:21 2005 modulo_add(int x=93, int y=1, int mod=100) Fri Jun 24 23:58:21 2005 modulo_add(int x=31, int y=0, int mod=64) Fri Jun 24 23:58:21 2005 modulo_add(int x=94, int y=1, int mod=100) Combine this with your observation that making the function inline or turning off -O2 optimization fixes the problem, and it's starting to smell like a compiler bug. James
