From mboxrd@z Thu Jan 1 00:00:00 1970 From: Avi Kivity Subject: Re: buildbot for kvm.git Date: Wed, 24 Nov 2010 12:23:37 +0200 Message-ID: <4CECE7A9.3090509@redhat.com> References: <4CDA9E9A.9040708@redhat.com> <201011242315.23682.gollub@b1-systems.de> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: KVM list , Christian Berendt To: Daniel Gollub Return-path: Received: from mx1.redhat.com ([209.132.183.28]:18668 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753211Ab0KXKXo (ORCPT ); Wed, 24 Nov 2010 05:23:44 -0500 In-Reply-To: <201011242315.23682.gollub@b1-systems.de> Sender: kvm-owner@vger.kernel.org List-ID: On 11/24/2010 12:15 PM, Daniel Gollub wrote: > On Thursday, November 11, 2010 02:31:06 am Avi Kivity wrote: > > Daniel, the buildbot has been fairly effective in keeping qemu-kvm.git > > building. I'd like to extend that to kvm.git, especially for non-x86 > > architectures. > > First test builds succeed for x86 and ia64: > http://buildbot.b1-systems.de/kvm/grid > > with gcc-4.5.0-nolibc toolchain from > http://www.kernel.org/pub/tools/crosstool/files/bin/x86_64/4.5.0/ > > powerpc not ready yet - trivial $ARCH change required in the buildbot > master configuration. > > s390 compilation seems to run fine and hit the first compilation issue: > http://buildbot.b1-systems.de/kvm/builders/s390/builds/0/steps/compile/logs/stdio Will work on fixing it - so it's already useful! I think buildbot has a mechanism where you can send it a patch to be build-tested. This allows developers to build-test on architectures they don't have. Can we activate this? > Currently buildbot notification mails don't get send to the kvm list. I would like > to do some more testing before spamming with wrong-testing results. > > Some more questions inline: > > > > I've now a virtual machine with a cross compile environment for s390, > > ppc, and ia64. I'd like buildbot to build for the following archs and > > configs: > > > > x86_64: defconfig + KVM=m KVM_INTEL=m KVM_AMD=m > > i386: defconfig + KVM=m KVM_INTEL=m KVM_AMD=m > > Without virtio? intentionally? No, that's my old configuration I used for build testing. While virtio isn't part of kvm, and is arch-neutral, it would be good to enable it. > > > > The recipe for a build is: > > - check out the source > > Any specific branches beside the master branch? master and next. > > > - MAKE defconfig > > where MAKE="make ARCH=$ARCH > > CROSS_COMPILE=/opt/gcc-4.5.0-nolibc/$ARCH-linux/bin/$ARCH-linux-' > > The buildbot master configuration requires exactly that path for CROSS_COMPILE on > the buildslave. The buildsaves need to be setup the toolchain before hand manually. > > Just let me and Christian know if we should add another toolchain or upgrade to a newer one. I think this is fine. Hopefully we won't upgrade too often, this should last us a year or two. > > - append the extra options to .config > > Done this by invoking: ./scripts/config (proposal/attempt from Christian) > > (See http://buildbot.b1-systems.de/kvm/builders/x86_64/builds/11/steps/configure_2/logs/stdio ) Ah, better than the hackery I used before. > > - MAKE oldconfig > > What about make target oldnoconfig instead? > To avoid that the build get stuck with every new config option? > > Demo of the issue with "make oldconfig": > http://buildbot.b1-systems.de/kvm/builders/x86_64/builds/10/steps/configure_3/logs/stdio I used 'make oldconfig < /dev/null', wasn't aware of oldnoconfig. Thanks for doing this! -- error compiling committee.c: too many arguments to function