Comment 3 for bug 161795

git-blame on kernel/exit.c shows that the offending line of code comes from the following commit:

commit 14bf01bb0599c89fc7f426d20353b76e12555308
Author: Linus Torvalds <email address hidden>
Date: Sat Oct 1 11:04:18 2005 -0700

    Fix inequality comparison against "task->state"

    We should always use bitmask ops, rather than depend on some ordering of
    the different states. With the TASK_NONINTERACTIVE flag, the inequality
    doesn't really work.

    Oleg Nesterov argues (likely correctly) that this test is unnecessary in
    the first place. However, the minimal fix for now is to at least make
    it work in the presense of TASK_NONINTERACTIVE. Waiting for consensus
    from Roland & co on potential bigger cleanups.

    Signed-off-by: Linus Torvalds <email address hidden>

diff --git a/kernel/exit.c b/kernel/exit.c
index ee6d8b8..4307773 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -1203,7 +1203,7 @@ static int wait_task_stopped(task_t *p, int delayed_group_

                exit_code = p->exit_code;
                if (unlikely(!exit_code) ||
- unlikely(p->state > TASK_STOPPED))
+ unlikely(p->state & TASK_TRACED))
                        goto bail_ref;
                return wait_noreap_copyout(p, pid, uid,
                                           why, (exit_code << 8) | 0x7f,

This is just a style-changed caused by the addition of a new task; the behaviour should have been consistent with the previous code (currently building a due-diligence kernel to test this assumption).

Unfortunately the previous code exists in the original import of the source tree into git, so there is no annotation history for why that line was introduced.