Re: [Bug 13012] 2.6.28.9 causes init to segfault on Debian etch; 2.6.28.8 OK

From: Linus Torvalds <torvalds@linux-foundation.org>
To: Dave Jones <davej@redhat.com>
Cc: Ian Lance Taylor <iant@google.com>, Frans Pop <elendil@planet.nl>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	linux-kbuild@vger.kernel.org, barryn@pobox.com,
	bugme-daemon@bugzilla.kernel.org
Subject: Re: [Bug 13012] 2.6.28.9 causes init to segfault on Debian etch; 2.6.28.8 OK
Date: Sat, 25 Jul 2009 09:49:41 -0700 (PDT)	[thread overview]
Message-ID: <alpine.LFD.2.01.0907250944570.3960@localhost.localdomain> (raw)
In-Reply-To: <20090725032343.GA20841@redhat.com>

On Fri, 24 Jul 2009, Dave Jones wrote:
> 
> Is it worth fixing these up, with diffs like the below ?

Historically, when we fix up bugs like this, it causes more bugs than it 
fixes.

It needs _really_ careful people, and people who really understand how C 
type rules work. Stuff that looks obvious often is not, and basing a large 
part of the patch on a compiler warning is a _very_ weak reason for 
something that is more than just syntactic.

And the problem is, nobody can judge the patch from the diff. So it gets 
absolutely zero review, until the day when somebody notices that the 
unsigned version is a bug.

I'd suggest that the rule should be:

 - if you can use -U30 and show all uses, so that people can actually look 
   at the patch and see what it _causes_ (ie not just the "we change 'i' 
   to 'unsigned'", but also the "this is where 'i' gets used, and 
   'unsigned' is right"), then we can apply it.

 - none of the patches go through a 'trivial' tree, or come from newbies 
   that think this is a good way to get involved.

Is it worth it at that point? I dunno.

		Linus