Re: linearize bug?

From: Linus Torvalds <torvalds@linux-foundation.org>
To: Kamil Dudka <kdudka@redhat.com>
Cc: Jeff Garzik <jeff@garzik.org>,
	Sparse Mailing-list <linux-sparse@vger.kernel.org>,
	Pekka J Enberg <penberg@cs.helsinki.fi>
Subject: Re: linearize bug?
Date: Sat, 27 Aug 2011 10:27:45 -0700	[thread overview]
Message-ID: <CA+55aFxK95i6AaDhgEteFsL8W_jRkxY93FQ-Z6FVJXix1g5DkA@mail.gmail.com> (raw)
In-Reply-To: <CA+55aFyD7d3hNP4JJL+0yE3ZB-M-65iYT_ay0uH_i-wrBR49Ww@mail.gmail.com>

On Sat, Aug 27, 2011 at 10:13 AM, Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> But your patch is fine too.

Hmm. There's badness going on with that patch: I can't check the
kernel source tree, I get a

  CHECK   arch/x86/kernel/process_64.c
  sparse: simplify.c:82: if_convert_phi: Assertion `br->cond' failed.

so clearly something is screwed up. It happens with your version too,
it seems to have gotten triggered by the loop rewriting change.

Which *should* have been semantically no difference what-so-ever. But
it's been so long, I can't recall all the rules. It may be that trying
to linearize the same expression twice (once for the entry condition,
once for the loop end) was just something invalid.

So the patch is just broken. Never mind.

                        Linus