From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:60069) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1gPTee-0006rd-Cp for qemu-devel@nongnu.org; Wed, 21 Nov 2018 09:39:01 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1gPTed-0005Rz-HR for qemu-devel@nongnu.org; Wed, 21 Nov 2018 09:39:00 -0500 Date: Wed, 21 Nov 2018 15:38:16 +0100 From: Samuel Ortiz Message-ID: <20181121143816.GD4426@caravaggio> References: <20181105014047.26447-1-sameo@linux.intel.com> <20181116172919.43f3e27d@redhat.com> <20181119163110.2f357f40@redhat.com> <20181121072954-mutt-send-email-mst@kernel.org> <20181121151526.5785b43f@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20181121151526.5785b43f@redhat.com> Subject: Re: [Qemu-devel] [PATCH v5 00/24] ACPI reorganization for hardware-reduced API addition List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Igor Mammedov Cc: "Michael S. Tsirkin" , Peter Maydell , Stefano Stabellini , qemu-devel@nongnu.org, Shannon Zhao , qemu-arm@nongnu.org, xen-devel@lists.xenproject.org, Anthony Perard , Paolo Bonzini , Richard Henderson , Eduardo Habkost Igor, On Wed, Nov 21, 2018 at 03:15:26PM +0100, Igor Mammedov wrote: > On Wed, 21 Nov 2018 07:35:47 -0500 > "Michael S. Tsirkin" wrote: > > > On Mon, Nov 19, 2018 at 04:31:10PM +0100, Igor Mammedov wrote: > > > On Fri, 16 Nov 2018 17:37:54 +0100 > > > Paolo Bonzini wrote: > > > > > > > On 16/11/18 17:29, Igor Mammedov wrote: > > > > > General suggestions for this series: > > > > > 1. Preferably don't do multiple changes within a patch > > > > > neither post huge patches (unless it's pure code movement). > > > > > (it's easy to squash patches later it necessary) > > > > > 2. Start small, pick a table generalize it and send as > > > > > one small patchset. Tables are often independent > > > > > and it's much easier on both author/reviewer to agree upon > > > > > changes and rewrite it if necessary. > > > > > > > > How would that be done? This series is on the bigger side, agreed, but > > > > most of it is really just code movement. It's a starting point, having > > > > a generic ACPI library is way beyond what this is trying to do. > > > I've tried to give suggestions how to restructure series > > > on per patch basis. In my opinion it quite possible to split > > > series in several smaller ones and it should really help with > > > making series cleaner and easier/faster to review/amend/merge > > > vs what we have in v5. > > > (it's more frustrating to rework large series vs smaller one) > > > > > > If something isn't clear, it's easy to reach out to me here > > > or directly (email/irc/github) for clarification/feed back. > > > > I assume the #1 goal is to add reduced HW support. So another > > option to speed up merging is to just go ahead and duplicate a > > bunch of code e.g. in pc_virt.c acpi/reduced.c or in any other > > file. > > This way it might be easier to see what's common code and what isn't. > > And I think offline Igor said he might prefer that way. Right Igor? > You mean probably 'x86 reduced hw' support. That's was what I've > already suggested for PCI AML code during patch review. Just don't > call it generic when it's not and place code in hw/i386/ directory beside > acpi-build.c. It might apply to some other tables (i.e. complex cases). > > On per patch review I gave suggestions how to amend series to make > it acceptable without doing complex refactoring and pointed out > places we probably shouldn't refactor now and just duplicate as > it's too complex or not clear how to generalize it yet. > > Problem with duplication is that a random contributor is not > around to clean code up after a feature is merged and we end up > with a bunch of messy code. > > A word to the contributors, > Don't do refactoring in silence, keep discussing approaches here, > suggest alternatives. That way it's easier to reach a compromise > and merge it with less iterations. And if you do split it in smaller > parts, the process should go even faster. > > I'll sent a small RSDP refactoring series for reference. I was already working on the RSDP changes. Let me know if I should drop that work too. Cheers, Samuel.