From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Justin T. Gibbs" Subject: Re: Aic7xxx v6.2.22 and Aic79xx v1.3.0Alpha2 Released Date: Wed, 11 Dec 2002 09:08:26 -0700 Sender: linux-scsi-owner@vger.kernel.org Message-ID: <1313340000.1039622906@aslan.scsiguy.com> References: <200212101602.gBAG2Hi02930@localhost.localdomain> <20021211135855.A19325@infradead.org> <1266570000.1039619906@aslan.scsiguy.com> <20021211153935.A23704@infradead.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20021211153935.A23704@infradead.org> Content-Disposition: inline List-Id: linux-scsi@vger.kernel.org To: Christoph Hellwig Cc: James Bottomley , linux-scsi@vger.kernel.org >> Why is this based on Alpha1 and not Alpha2. > > Because that's a) what James put in the BK tree and b) that's what I > downloaded from your website today for reference. Where did you download it from. This file is Alpha2 based: http://people.FreeBSD.org/~gibbs/linux/SRC/aic79xx-linux-2.5.tar.gz >> driver has to build all the way back to 2.4.7 (RedHat 7.2 support). > > RedHat only supports kernel 2.4.18 for RH7.2/i386 and 2.4.9 with tons of > hacks for the other arches, so your argument is void. Tell that to the OEMs that Adaptec has to support. They still test and require 7.2 drivers. > Not to mention the > only support that is dropped by my changes is for builtin support < > 2.2.18. I'll have to verify that. >> Removing ifdefs just makes it harder for me to merge in changes from >> external trees. Yes, the ifdefs are ugly, but so is the Linux SCSI layer >> and the unmanaged way that interfaces have been changed without any >> consideration to backwards compatibility (eg. the HIGHIO stuff could have >> been done with 0 impact to drivers, but wasn't for some strange reason). > > linux driver interface change between stable series, face it. I have no problem with interfaces changing for good reason, but, for example, a driver that alread sets unchecked_isa_dma to 0 and uses the PCI dma mask shouldn't have to set addition flags (with different names in different vendor's trees) to enable HIGHIO. It's yet-another *stupid* interface change. > And a driver with tons of ifdefs is utter crap. I agree with that, but since we have to build for something like 35 different kernel versions (major vendor releases plus all of their errata), there is not much choice as a vendor. Without the ifdefs, one of those builds invariably gets broken by a minor driver update because the person doing the update doesn't understand how their change will effect the support for the other kernels. > Interestingly only vendor driver use that shitty scheme. Yes, because the vendor actually has to support all of those versions unlike some guy who hacks this stuff in his spare time and could care less about anyone else's requirements but his own. So your choice becomes either accept the ifdefs so that the vendor can sell its products by meeting OEM/Channel demands, or have the vendor exit the Linux market and not even attempt to support their product. -- Justin