Re: exit: reparent: cleanup the changing of ->parent

[Oleg Nesterov <oleg@redhat.com>]

On 11/14, Dan Carpenter wrote:
>
> The patch eb6d8479b73d: "exit: reparent: cleanup the changing of
> ->parent" from Nov 13, 2014, leads to the following static checker
> warning:
>
> 	kernel/exit.c:543 forget_original_parent()
> 	warn: add some parenthesis here?
>
> kernel/exit.c
>    538          /* Can drop and reacquire tasklist_lock */
>    539          reaper = find_new_reaper(father);
>    540          list_for_each_entry(p, &father->children, sibling) {
>    541                  for_each_thread(p, t) {
>    542                          t->real_parent = reaper;
>    543                          BUG_ON(!t->ptrace != (t->parent = father));
>                                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>
> The reason for this warning is that many people forget that ! is higher
> precedence than =.

Do you really think we should try to shut up this warning?

IMO this warning is wrong, "!A = B" or "!A != B" looks fine to me...


> This is a complicated condition however you write
> it, but it might be more clear to say:
>
> 				BUG_ON((!!t->ptrace) = (t->parent = father));

This is subjective, but to me it looks more confusing.

If we really want to "fix" it, I'd suggest

				BUG_ON((!t->ptrace) != (t->parent = father));

Oleg.