From mboxrd@z Thu Jan 1 00:00:00 1970 From: Luc Van Oostenryck Subject: Re: [PATCH 1/6] storage should not be inherited by pointers Date: Mon, 28 Nov 2016 03:02:53 +0100 Message-ID: <20161128020252.GA744@macbook.local> References: <20161124170947.14379-1-luc.vanoostenryck@gmail.com> <20161124170947.14379-2-luc.vanoostenryck@gmail.com> <20161125033817.GA14283@macbook.local> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from mail-wm0-f65.google.com ([74.125.82.65]:33076 "EHLO mail-wm0-f65.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753216AbcK1CDX (ORCPT ); Sun, 27 Nov 2016 21:03:23 -0500 Received: by mail-wm0-f65.google.com with SMTP id u144so16547549wmu.0 for ; Sun, 27 Nov 2016 18:02:56 -0800 (PST) Content-Disposition: inline In-Reply-To: Sender: linux-sparse-owner@vger.kernel.org List-Id: linux-sparse@vger.kernel.org To: Christopher Li Cc: Linux-Sparse On Mon, Nov 28, 2016 at 09:04:56AM +0800, Christopher Li wrote: > Sorry I jump the conclusion. If it is only concern with MOD_STATIC, > MOD_EXTERN & MOD_TOPLEVEL, that is not a bug issue at all. > I thought you are going to extend that logic to other modifier bits as well. Well ... yes and no :) Like it can be seen in the other patches, I mostly just wrote some tests that, I think, clearly expose soem problems. Those problems can be considered as bugs or not, some clearly are, I think, some much less so but trigger questions. I think we should clearly define what is the desired behavior of modifiers like nocast, noderef, bitwise, ... when - taking the address of an object with such modifiers - using typeof() on such object or pointers - assignment with them (but this seems much better done/defined) The case with 'safe' is similar but its working is a bit different. And indeed, like I think you fear, any changes regarding the address_space will create problems in existing code and it's why I think it's better for now to first look at the situation with the modifiers. You certainly can see this seriemore as the start of a reflexion than a try to solve anything. Luc