From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Henrik Rydberg" Subject: Re: [PATCH v3] Input: Add EVIOC mechanism for MT slots Date: Sun, 5 Feb 2012 20:40:36 +0100 Message-ID: <20120205194036.GA2566@polaris.bitmath.org> References: <1328036403-2827-1-git-send-email-rydberg@euromail.se> <4F2B24AC.1020408@gmail.com> <20120203072725.GA6914@polaris.bitmath.org> <4F2D7730.3090704@gmail.com> <20120205075906.GA7698@polaris.bitmath.org> <4F2EB8C8.9090102@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from smtprelay-b12.telenor.se ([62.127.194.21]:35674 "EHLO smtprelay-b12.telenor.se" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753266Ab2BETjT (ORCPT ); Sun, 5 Feb 2012 14:39:19 -0500 Content-Disposition: inline In-Reply-To: <4F2EB8C8.9090102@gmail.com> Sender: linux-input-owner@vger.kernel.org List-Id: linux-input@vger.kernel.org To: Chase Douglas Cc: Dmitry Torokhov , linux-input@vger.kernel.org, linux-kernel@vger.kernel.org > > Besides leaving a possible giant stack crash in your code, it assumes > > memory is somehow magically allocated. Not good practise in low-level > > programming. You wouldn't use a template this way, would you? > > No, which is why I think this interface is bad. I should be able to > dynamically set the size of the array, but it's not possible with this > interface. It is possible (using num_slots == 0 or creating your own struct), but not ideal, granted. The patch serves the purpose of definining the binary interface, the rest is up to userland. > I think the implementation is fine in terms of how the plumbing works. I > just don't think this macro should be included. Make the user create the > struct themselves: > > In linux/input.h: > > struct input_mt_request { > __u32 code; > __s32 values[]; > }; The above (the first) version is not ideal either, since it cannot be used as it is. > It could be argued that we should leave the macro around for people who > want to statically define the size of the request, but I think that is > leading them down the wrong path. It's easier, but it will lead to > broken code if you pick the wrong size. Rather than creating both a suboptimal static and a suboptimal dynamic version, removing the struct altogether is tempting. All that is really needed is a clear definition of the binary interface. The rest can easily be set up in userland, using whatever method is preferred. Thanks. Henrik