Re: [Qemu-devel] [PATCH 10/10] linux-user: remove unused variables

[Michael S. Tsirkin]

On Wed, Jun 15, 2011 at 07:32:50AM -0700, Richard Henderson wrote:
> On 06/15/2011 01:35 AM, Alexander Graf wrote:
> >> -                abi_ulong arg5 = 0, arg6 = 0, arg7 = 0, arg8 = 0;
> >> +                abi_ulong arg5 = 0, arg6 = 0;
> >>
> >>                 nb_args = mips_syscall_args[syscall_num];
> >>                 sp_reg = env->active_tc.gpr[29];
> >>                 switch (nb_args) {
> >>                 /* these arguments are taken from the stack */
> >>                 /* FIXME - what to do if get_user() fails? */
> >> -                case 8: get_user_ual(arg8, sp_reg + 28);
> >> -                case 7: get_user_ual(arg7, sp_reg + 24);
> >> +                case 8: /* get_user_ual(arg8, sp_reg + 28); */
> >> +                case 7: /* get_user_ual(arg7, sp_reg + 24); */
> > 
> > I'd prefer to see these and the respective variable definitions #if
> > 0'd with a comment, stating that they're currently unused.
> 
> I'd prefer not to see if 0 code.  Better, I think, to mark the
> variables as __attribute__((unused)) with that same comment.

Why keep dead code around? If it's for documentation pruposes
comment or if 0 seems more appropriate than attributes.

> >> @@ -7058,18 +7056,14 @@ abi_long do_syscall(void *cpu_env, int num, 
> >> abi_long arg1,
> >>     case TARGET_NR_osf_sigprocmask:
> >>         {
> >>             abi_ulong mask;
> >> -            int how = arg1;
> >>             sigset_t set, oldset;
> >>
> >>             switch(arg1) {
> >>             case TARGET_SIG_BLOCK:
> >> -                how = SIG_BLOCK;
> >>                 break;
> >>             case TARGET_SIG_UNBLOCK:
> >> -                how = SIG_UNBLOCK;
> >>                 break;
> >>             case TARGET_SIG_SETMASK:
> >> -                how = SIG_SETMASK;
> > 
> > why go through the effort of setting "how" and then not using it? I'm
> > pretty sure this is a bug as well. A few lines down is the following
> > code:
> > 
> >    sigprocmask(arg1, &set, &oldset);
> > 
> > which in TARGET_NR_sigprocmask would be:
> > 
> >   ret = get_errno(sigprocmask(how, &set, &oldset));
> > 
> > So we end up sending guest masks to the host. Richard, this is Alpha
> > specific code. Mind to double-check?
> 
> I remember fixing this before.  Perhaps it was in a patch tree that 
> never got pulled...
> 
> 
> r~