In sysdeps/unix/sysv/linux/ptrace.c, on line 102, if __BOUNDED__POINTERS__ is true, checks are done that pointers point at appropriate locations. Such a check is done on the "data" argument to ptrace() if the "request" argument is PTRACE_SETOPTIONS. However, according to ptrace(2) (and, indeed, according to the gdb sources and the kernel sources), the PTRACE_SETOPTIONS request accepts a bitmask of PTRACE_O_* values as its "data" argument, *not* a pointer.
__BOUNDED__POINTERS__ is not supported and never really worked. The code is only left in place in case there is some future technology coming which can use that information.
No big deal, it didn't actually impact meβI just saw it and thought I'd point out that it looks like it shouldn't be there.