On 11/10, Roland McGrath wrote: > > > This change affects ia64, microblaze, parisc, powerpc, sh. They pass > > nonzero "step" argument to tracehook but since it was ignored the tracee > > reports via ptrace_notify(), this is not right and not consistent. > > This could be more explicit about the details of the change. Ideally > we should point to an actual test program whose results are changed by > fixing the kernel. > > I think the best thing to say is that previously using purely > PTRACE_SINGLESTEP without regard to what the user instructions were > could sometimes swallow a provided signal, whereas now it treats the > syscall instruction like any other user instruction.
Well, ptrace_report_syscall() checks current->exit_code after ptrace_notify() and does send_sig(). But yes, PTRACE_SETSIGINFO is doesn't work. Also. If PT_TRACESYSGOOD is set ptrace_report_syscall() adds 0x80 which is then reported by wait(&status), this doesn't look right. OK, I assume otherwise you acked these changes, I am going to update the changelog and send them to akpm with your acked-by included. You still have a chance to stop me ;) Oleg.