Re: [Ksummit-discuss] checkkpatch (in)sanity ?

From: Joe Perches <joe@perches.com>
To: Josh Triplett <josh@joshtriplett.org>,
	Dan Carpenter <dan.carpenter@oracle.com>
Cc: ksummit-discuss@lists.linuxfoundation.org,
	Greg KH <gregkh@linuxfoundation.org>,
	Sasha Levin <levinsasha928@gmail.com>,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: [Ksummit-discuss] checkkpatch (in)sanity ?
Date: Mon, 29 Aug 2016 12:21:37 -0700	[thread overview]
Message-ID: <1472498497.3425.87.camel@perches.com> (raw)
In-Reply-To: <20160829191020.GB27600@cloud>

On Mon, 2016-08-29 at 12:10 -0700, Josh Triplett wrote:
> On Mon, Aug 29, 2016 at 10:06:18PM +0300, Dan Carpenter wrote:
> > 
> > I would like a couple changes which you know already:
> > 
> > 1) Get rid of PREFER_ETHER_ADDR_COPY and similar because the people who
> > send checkpatch.pl fixes aren't qualified to say when it's legal or not
> > so they sometimes introduce bugs.
> I do think we should have *something* that catches such things.
> Perhaps not checkpatch.pl, though.  Perhaps a compiler plugin that
> generates additional warnings, and can perhaps use more global
> information to determine legality?

nit: validity rather than legality.

There are still rather a lot of these.

$ git grep -E "\bmem.*,\s*(ETH_ALEN|6)\s*\);" | wc -l
1776

Dunno if any of them are in performance sensitive
areas where it actually matters.

Someone, I forget who, had a concern about the
object being set possibly being in a struct where
it's possible for the alignment of the set object
to be altered by another change like adding a new
member.