From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from stoexhub02.domain01.net ([83.145.59.143]) by linuxtogo.org with esmtp (Exim 4.72) (envelope-from ) id 1RIaJD-0004GX-LW for openembedded-core@lists.openembedded.org; Tue, 25 Oct 2011 08:16:23 +0200 Received: from localhost (85.231.30.150) by stoexhub02.domain01.net (10.12.10.2) with Microsoft SMTP Server id 8.3.83.0; Tue, 25 Oct 2011 08:09:16 +0200 From: Anders Darander Organization: ChargeStorm AB To: Patches and discussions about the oe-core layer Date: Tue, 25 Oct 2011 08:09:34 +0200 User-Agent: KMail/1.13.7 (Linux/3.0.0-2-amd64; KDE/4.6.5; x86_64; ; ) X-Accept-Language: sv, en, de X-GPG-Key-Id: 0x920E938C X-GPG-Keyserver: hkp://keys.gnupg.net MIME-Version: 1.0 Message-ID: <201110250809.34347.anders@chargestorm.se> X-GFI-SMTP-Submission: 1 Subject: Issue with RDEPENDS in module.bbclass for initramfs X-BeenThere: openembedded-core@lists.openembedded.org X-Mailman-Version: 2.1.11 Precedence: list Reply-To: Patches and discussions about the oe-core layer List-Id: Patches and discussions about the oe-core layer List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 25 Oct 2011 06:16:23 -0000 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable We have recently started investigating to make an even smaller build of our local distro & machine. This smaller build should be run exclusively from an initramfs, thus we have a need to redu=C4=87e the size of the rootfs as much as possible. (Especially as we'd still like some "bigger" tools/pieces in the initramfs). However, our first attempt to create a small image turned our to be larger than expected. After a first investigation, it turns out that our external kernel modules have an RDEPENDS on kernel-image (this is from module.bbclass). As we're creating an initramfs-based rootfs, this is obviously not desired... Locally removeing that RDEPENDS reduces the image as expected... I understand that in the general distro case, there is a need to be able to express a dependency on a certain kernel version. However, I really don't think that bringing in the kernel-image is a good idea, not an general embedded platforms at least. (On the x86 this is usually not a problem, as you almost always have the kernel as a part of the rootfs anyway). Could we come up with another way of expressing this dependency? That would definitely be valuable for the small-ish embedded systems. Cheers, Anders =2D-=20 Anders Darander ChargeStorm AB / eStorm AB