Re: [PATCH -mm] signals-clear-signal-tty-when-the-last-thread-exits.fix

[Andrew Morton <akpm@linux-foundation.org>]

On Wed, 24 Mar 2010 15:23:48 +0100 Oleg Nesterov <oleg@redhat.com> wrote:

> (fixup for signals-clear-signal-tty-when-the-last-thread-exits.patch)
> 
> I didn't get this warning, but the old gcc complains
> 
> 	kernel/exit.c: In function 'release_task':
> 	kernel/exit.c:85: warning: 'tty' may be used uninitialized in this function
> 
> This clearly wrong, to the point it blames release_task() instead of
> __exit_signal(). But let's make compiler happy anyway, hopefully this
> is what it wants.
> 
> Signed-off-by: Oleg Nesterov <oleg@redhat.com>
> ---
> 
>  kernel/exit.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> --- 34-rc1/kernel/exit.c~FIX_EXIT_SIGNAL_TTY_WARNING	2010-03-21 18:36:44.000000000 +0100
> +++ 34-rc1/kernel/exit.c	2010-03-24 14:59:55.000000000 +0100
> @@ -82,7 +82,7 @@ static void __exit_signal(struct task_st
>  	struct signal_struct *sig = tsk->signal;
>  	bool group_dead = thread_group_leader(tsk);
>  	struct sighand_struct *sighand;
> -	struct tty_struct *tty;
> +	struct tty_struct *tty = NULL;	/* supress gcc warning */

uninitialized_var() is a neater way.

(uninitialized_var() will save a teeny bit of .text on old gcc.  One
suspects that a newer gcc which is capable of working out that this
variable _isn't_ uninitialized would also be capable of eliding the `= 0').