Currently the situation is as follow virtual/udev-197-r2 DEPs are || ( >=sys-fs/udev-197-r8[gudev?,hwdb?,introspection?,keymap?,kmod?,selinux?,static-libs?] >=sys-apps/systemd-198-r5[gudev?,introspection?,kmod?,selinux?,static-libs?] kmod? ( >=sys-fs/eudev-1_beta2-r2[modutils,gudev?,hwdb?,introspection?,keymap?,selinux?,static-libs?] ) !kmod? ( >=sys-fs/eudev-1_beta2-r2[gudev?,hwdb?,introspection?,keymap?,selinux?,static-libs?] ) ) sys-fs/udev post-DEPs are openrc? ( >=sys-fs/udev-init-scripts-25 ) This post-DEP is missing in the systemd ebuild. A user may favor systemd over udev, which is fine since systemd includes all of udev, that's what the virtual says. However, if a user decideds to continue to use openrc instead of systemd, udev-init-scripts is not depended upon by any package and will get uninstalled with the next purge of world. Either one of virtual/udev sys-apps/systemd should depend on openrc? ( >=sys-fs/udev-init-scripts-25 ) just as is now for sys-fs/udev. Thanks.. Reproducible: Always
I am not sure my following comment fits the issue here, but I am against the indirect method. For the simple reason: Gentoo developers might make mistakes! Why not having one and for all ebuild gentoo-boot-helpers USE flags: linux freebsd openrc systemd udev eudev ...etc When incompatible to install in one root these flags should exclude each other. Might be there is even a watchout implemented to have the correct profile.
I'm fine either way. @udev, do you think udev-gentoo-scripts fit more into virtual/udev or all providers?
the virtual should only have coverage of USE flags that other packages can depend on, USE="openrc" is not one of them futher we can't guarantee the compability of udev-init-scripts with the installation layout provided sys-apps/systemd so i'd prefer if we don't add such USE flag to the virtual
Well, I believe that it all depends on how we define the virtual. If 'virtual/udev' should mean *any package providing udev implementation compatible with openrc*, then the dep should be there, even unconditional (I don't mind having it that way). In any case, to avoid delaying the fix, I've added the dep on >=25 to systemd. When new udev-init-scripts is released, I will bump it again to have proper 'udevadm' call. @udev, feel free to close the bug after you decide on whether the virtual needs the dep or not.
(In reply to comment #4) > Well, I believe that it all depends on how we define the virtual. > > If 'virtual/udev' should mean *any package providing udev implementation > compatible with openrc*, then the dep should be there, even unconditional (I > don't mind having it that way). @mgorny: If someone wants to run a pure systemd system, they might not want openrc on their box at all, and if they don't want it, they shouldn't get the init scripts either. This might be for a separate bug, but I think we should have more packages support the openrc use flag, not less. This is different from the logrotate use flag, because these are not configuration files that just sit there, but they are actually scripts. Would you please put this dependency behind an "openrc" use flag in systemd? Thanks, William
(In reply to comment #5) udev-init-scripts installs lib/udev/rules.d/90-network.rules, which I definitely don't want on my systemd system. + 06 Apr 2013; Mike Gilbert <floppym@gentoo.org> metadata.xml, + systemd-200-r1.ebuild, systemd-9999.ebuild: + Add openrc use flag to control dependency on sys-fs/udev-init-scripts, bug + 464502.