From mboxrd@z Thu Jan 1 00:00:00 1970 From: Peter Korsgaard Date: Wed, 05 Feb 2014 13:52:55 +0100 Subject: [Buildroot] [PATCH v6 3/5] udev: convert to virtual package. In-Reply-To: (Thomas De Schampheleire's message of "Wed, 5 Feb 2014 13:31:14 +0100") References: <1389627907-7821-1-git-send-email-eric.le.bihan.dev@free.fr> <1389627907-7821-4-git-send-email-eric.le.bihan.dev@free.fr> <20140204121030.GA8125@pc-eric> Message-ID: <87d2j1202g.fsf@dell.be.48ers.dk> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net >>>>> "Thomas" == Thomas De Schampheleire writes: Hi, >> I find it odd that some packages depend on >> BR2_ROOTFS_DEVICE_CREATION_DYNAMIC_UDEV whereas others depend on >> BR2_PACKAGE_UDEV. To me, both symbols are interpreted as "the package needs a >> udev daemon to talk to at runtime" (whether via D-Bus or libudev). So I >> replaced both with BR2_PACKAGE_HAS_UDEV to be homogeneous. It is true that if >> a package depends on BR2_PACKAGE_HAS_UDEV, it is not explicitly said if it is >> a runtime or a build dependency. But, AFAIK, programs that want to communicate >> with the udev daemon do it via libudev/libgudev, so we end up with a build >> dependency. >> >> In system/Config.in, I removed BR2_ROOTFS_DEVICE_CREATION_DYNAMIC_UDEV because >> I want the user to understand the change: he/she should not think of 'udev' as >> a package, but as a feature which, at this stage, is only available via eudev. >> I could have kept BR2_ROOTFS_DEVICE_CREATION_DYNAMIC_UDEV and have it select >> BR2_PACKAGE_EUDEV, but I've found it confusing. >> >> But I should have added BR2_ROOTFS_DEVICE_CREATION_DYNAMIC_UDEV in >> Config.in.legacy to migrate to BR2_ROOTFS_DEVICE_CREATION_DYNAMIC_EUDEV. >> >> In the end, the user selects BR2_ROOTFS_DEVICE_CREATION_DYNAMIC_EUDEV, which >> selects BR2_PACKAGE_EUDEV, which in turn selects BR2_PACKAGE_HAS_UDEV. All the >> packages depending on it are now available. > Based on your additional input, and rechecking the changes, I > understand your reasoning and accept it. > You still should add the legacy symbol, though. > Peter, what about you? With the legacy handling added I'm fine with the rename. -- Bye, Peter Korsgaard