From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mailman by lists.gnu.org with tmda-scanned (Exim 4.43) id 1Mp2Kd-0006gA-I3 for qemu-devel@nongnu.org; Sat, 19 Sep 2009 11:58:39 -0400 Received: from exim by lists.gnu.org with spam-scanned (Exim 4.43) id 1Mp2KZ-0006fy-V9 for qemu-devel@nongnu.org; Sat, 19 Sep 2009 11:58:39 -0400 Received: from [199.232.76.173] (port=53706 helo=monty-python.gnu.org) by lists.gnu.org with esmtp (Exim 4.43) id 1Mp2KZ-0006fv-S4 for qemu-devel@nongnu.org; Sat, 19 Sep 2009 11:58:35 -0400 Received: from moutng.kundenserver.de ([212.227.17.9]:49244) by monty-python.gnu.org with esmtp (Exim 4.60) (envelope-from ) id 1Mp2KZ-00053x-65 for qemu-devel@nongnu.org; Sat, 19 Sep 2009 11:58:35 -0400 Message-ID: <4AB4FF9C.80005@mail.berlios.de> Date: Sat, 19 Sep 2009 17:58:20 +0200 From: Stefan Weil MIME-Version: 1.0 References: <1253360475-10671-1-git-send-email-weil@mail.berlios.de> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Subject: [Qemu-devel] Re: [PATCH] eepro100: Remove unused device status entries List-Id: qemu-devel.nongnu.org List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Juan Quintela Cc: QEMU Developers Juan Quintela schrieb: > Stefan Weil wrote: >> A lot of entries are unused (they were added by copy + paste >> from other drivers during development of eepro100.c). >> >> Removing them from nic_save, nic_load makes any >> old saved status incompatible, so a new version >> for the virtual machine data was needed, too. >> >> This patch is a step to synchronize my maintainer version >> of eepro100.c (git://repo.or.cz/qemu/ar7.git) with the >> version integrated in QEMU. >> >> +/* Parameters for nic_save, nic_load. */ >> +static const int eepro100_instance = -1; >> +static const int eepro100_version = 20090807; > > > Please, don't make version big numbers, idea is that versions are small > numbers. > > Why do you want to be incompatible with the old saved state? We are > going to big troubles to make everythnig backwards compatible, removing > fields just makes everybody life worse. You can just rename as: > > s/foo/foo_dummy/ or /foo_vmstate/ and let it into the state. > > Later, Juan. hi Juan backward compatibility is ok for widely used features in stable software releases, and for a short moment I had thought about implementing it. Reading dummies would have been sufficient (no need to keep dummies in the state). But I decided against it. I doubt that anybody is using saved machine states with eepro100 driver data, because those network cards are not QEMU's default, and they are still unfinished. So backward compatibility would have been an unnecessary burden here. It is also better not to pretend that there is any backward compatibility when it is nearly impossible to test this (and I am sure that there are severe faults in it). My patch aims at finishing (or at least fixing and completing) this driver. Some more patches are still needed for this goal (see my version on repo.or.cz). Personally, I prefer version numbers which express something, for example the date of a change (as in my case). If it helps, I promise to send a patch which sets the version number to 4 as soon as the rest of eepro100.c is ok. Antony, please take the patch for now as it is, because I won't send a new one just to change the version number. Regards Stefan