benson@odi.com (Benson I. Margulies) (01/13/91)
I have a program that works fine -g, and fails -O. Seems that the compiler is making a bad assumption about the contents of r30. It loads up a bad address and takes a SIGSEGV. I called it in, though I couldn't boil the testcase down, and received "xlc emergency release 1.0". Well, I installed it, and now I've a plague of similiar problems. I suspect that I have to compile everything with the new compiler to use it at all. Needless to say, I'm still negotiating with defect support on this, but I wonder if anyone else has been here before . -- Benson I. Margulies