From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga04.intel.com (mga04.intel.com [192.55.52.120]) by dpdk.org (Postfix) with ESMTP id 6E5D17CBA for ; Fri, 18 Aug 2017 10:35:49 +0200 (CEST) Received: from fmsmga006.fm.intel.com ([10.253.24.20]) by fmsmga104.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 18 Aug 2017 01:35:47 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.41,391,1498546800"; d="scan'208";a="141797068" Received: from bricha3-mobl3.ger.corp.intel.com ([10.237.221.24]) by fmsmga006.fm.intel.com with SMTP; 18 Aug 2017 01:35:45 -0700 Received: by (sSMTP sendmail emulation); Fri, 18 Aug 2017 09:35:44 +0100 Date: Fri, 18 Aug 2017 09:35:44 +0100 From: Bruce Richardson To: Luca Boccassi Cc: Marco Varlese , dev@dpdk.org, thomas@monjalon.net Message-ID: <20170818083544.GA22884@bricha3-MOBL3.ger.corp.intel.com> References: <20170814095208.166496-1-bruce.richardson@intel.com> <1502979024.15962.10.camel@suse.com> <1502983536.31476.4.camel@gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <1502983536.31476.4.camel@gmail.com> Organization: Intel Research and =?iso-8859-1?Q?De=ACvel?= =?iso-8859-1?Q?opment?= Ireland Ltd. User-Agent: Mutt/1.8.3 (2017-05-23) Subject: Re: [dpdk-dev] [RFCv2 00/40] Building DPDK with meson and ninja X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 18 Aug 2017 08:35:49 -0000 On Thu, Aug 17, 2017 at 04:25:36PM +0100, Luca Boccassi wrote: > On Thu, 2017-08-17 at 16:10 +0200, Marco Varlese wrote: > > Hi Bruce, > > > > On Mon, 2017-08-14 at 10:51 +0100, Bruce Richardson wrote: > > > Following on from previous RFC [http://dpdk.org/dev/patchwork/patch > > > /25104/] > > > here is a second draft implementation for building DPDK with meson > > > and > > > ninja. While still not building all of DPDK, and needing patch > > > cleanup so > > > that patches don't overwrite previous work, it is more complete in > > > many > > > ways than the previous version and includes: > > > > > > * dynamic build configuration e.g. building pcap driver only if > > > pcap is > > >   found, only build af_packet if the target is linux, and only > > > building QAT > > >   and openssl crypto drivers if libcrypto is found > > > * support for pmdinfo inside the PMDs(for shared builds) and > > > binaries (for > > >   static builds) > > > * generalized/standardized way of building libs and drivers, though > > > the > > >   drivers code still needs generalization at the driver, rather > > > than > > >   driver-class level. > > > * support for having a pkgconfig file for DPDK on install, and > > > helloworld > > >   and l2fwd can be built using the pkgconfig info (via make, not > > > ninja) > > > * support for library versions > > > * an implementation for FreeBSD as well as Linux > > > * all libraries are included in the build, as well as a number of > > > NIC, > > >   crypto, and mempool drivers > > > * the igb_uio kernel module is build via the kernel Kbuild system > > > as part > > >   of a meson/ninja DPDK build > > > > This is really great to see. I do have one suggestion. > > Would it be possible to version the libraries built by the build- > > system > > following the same version of the DPDK release? > > > > For instance, in DPDK 17.08 we currently have: > > # objdump -p librte_pmd_ixgbe.so.1 |grep SONAME > >   SONAME               librte_pmd_ixgbe.so.1 > > > > Would it make sense to instead have librte_pmd_ixgbe.so.17.08 > > > > I think it would help to facilitate the installation of multiple DPDK > > library > > versions on the same system.  > > > > For example, we could have the following scenario: > > > > 1) OpenVSwithc linked with version 17.02 of DPDK > > 2) VPP linked with version 17.08 of DPDK > > 3) DPDK 18.xx installed in the system for any cutting-edge > > application > > prototyping. > > > > Is this something which could be incorporated as part of this work? > > Christian sent a patch a while ago, which was merged, to enable this in > the current build system, it's the CONFIG_RTE_MAJOR_ABI option, we use > it in Debian and Ubuntu for the reasons you mentioned. > > And if it's not been translated yet, I agree it's an important one. > No, it's not translated yet - mainly for the reasons that I had forgotten it existed, and that there is a lot yet unported. General question: should this be the default or not? It looks to me that it should probably be, but what do others think? /Bruce