On Sat, 9 Jun 2001, Herbert Xu wrote: > You're right. Compiling raid1.c with -O does seem to generate saner code > compared to -O2. I figured that. If you can, try compiling the same code with gcc-3.0 with both -O and -O2 and see what it generates. We have a better shot at correcting the problem if it still exists in 3.0. C