From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:59552) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1VELs8-0004UD-Tr for qemu-devel@nongnu.org; Tue, 27 Aug 2013 12:12:06 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1VELs2-00007f-Ts for qemu-devel@nongnu.org; Tue, 27 Aug 2013 12:12:00 -0400 Received: from mx1.redhat.com ([209.132.183.28]:44369) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1VELs2-00007R-L6 for qemu-devel@nongnu.org; Tue, 27 Aug 2013 12:11:54 -0400 Date: Tue, 27 Aug 2013 19:13:49 +0300 From: "Michael S. Tsirkin" Message-ID: <20130827161349.GA29285@redhat.com> References: <20130827070110.GA16338@redhat.com> <87zjs3o9l1.fsf@blackfin.pond.sub.org> <20130827160923.GY2631@otherpad.lan.raisama.net> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20130827160923.GY2631@otherpad.lan.raisama.net> Subject: Re: [Qemu-devel] [PATCH] pc: reduce duplication, fix PIIX descriptions List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Eduardo Habkost Cc: Markus Armbruster , qemu-devel@nongnu.org, Anthony Liguori , Paolo Bonzini , Laszlo Ersek , Andreas =?iso-8859-1?Q?F=E4rber?= On Tue, Aug 27, 2013 at 01:09:23PM -0300, Eduardo Habkost wrote: > On Tue, Aug 27, 2013 at 09:51:22AM +0200, Markus Armbruster wrote: > > "Michael S. Tsirkin" writes: > > > > > We have a lot of code duplication between machine types, > > > this increases with each new machine type > > > and each new field. > > > > > > This has already introduced a minor bug: description > > > for pc-1.3 says "Standard PC" while description for > > > pc-1.4 is "Standard PC (i440FX + PIIX, 1996)" > > > which makes you think 1.3 is somehow more standard, > > > or newer, while in fact it's a revision of the same PC. > > > > I wouldn't call it a bug. We're in the habit of never changing old > > machine types, so when we created pc-i440fx-1.4 and pc-q35-1.4 with a > > .desc that clearly explains the difference, we didn't change the older > > versions the PC machine types as well. > > > > That may have been overly cautious. As long as .desc is not exposed to > > the guest, it's not ABI, thus can be changed. > > I just confirmed that on qemu-system-x86_64 the only place where .desc > is used is on the machine_parse() help/error code at vl.c. I don't know > about other architectures. Generally, one would hope we won't expose random internal stuff to guests :). > I suggest we explicitly document the field as a human-readable > machine-type description that should be never exposed to the guest (just > in case somebody decides to use .desc on some guest-visible table one > day). Where would you document this? > -- > Eduardo