Summary: | after upgrade to syslog-ng - Service 'sysklogd' already provide 'logger'!; | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Aben Siatris <aben> |
Component: | New packages | Assignee: | Gentoo Linux bug wranglers <bug-wranglers> |
Status: | RESOLVED INVALID | ||
Severity: | normal | CC: | chutz+bugs.gentoo.org |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- |
Description
Aben Siatris
2004-04-05 02:39:37 UTC
You can fix this by removing /etc/init.d/sysklogd, even if you unmerge a package, it's initscript will not be wiped out (and it has a provide=logger). ok, thanx :) Comment #1 is actually a workaround and not a solution to the problem, isn't it? Isn't this bug a real shortcoming of the current implementation of the initscripts? If a runscript requires 'logger' (as this example is), shouldn't it instead of requiring the first one it finds (which usually is the first one in alphabetical order), first check if there is a script in some runlevel that provides the feature. So that someone can do # rc-update del metalog # rc-update del sysklogd # rc-update add syslog-ng default # depscan.sh And then at boot, not have metalog started by iptables (for example), but syslog-ng. If you cannot have more than one logger installed, because always the first one is going to be started at boot, then it should be reflected in the .ebuild-s themselves, i.e. RDEPEND="!virtual/logger" or similar for every logger. But that would make changing the logger rather hard, as one would have to remove the old one first. |