On Thu, Apr 21, 2011 at 1:03 PM, Tom Gundersen <[email protected]> wrote: > On Thu, Apr 21, 2011 at 7:24 PM, Dan McGee <[email protected]> wrote: >> On Thu, Apr 21, 2011 at 12:10 PM, Tom Gundersen <[email protected]> wrote: >>> * We no longer call mdadm, as this is dealt with by udev > > The relevant rule file is shipped with mdadm and has been for a long > time (2008). The rule file gained support for IMSM arrays in mdadm > version 3.2, so it might be best to use a relatively up-to-date mdadm. > Can we assume that people are not holding back mdadm and upgrading > initscripts? One would think so, and 2008 is in fact quite old so probably not something to fret about. (Note: our binaries require a kernel >= 2.6.27 anyway, which was released in late 2008).
>>> * We no longer copy rule files from /dev/.udev on boot, as this is >>> dealt with by udev > > Handling of persistent net/cd links has been in udev for a long time > (the exact udev version escapes me at the moment). However, as a > general rule I would be in favor of forcing an up-to-date udev. (In > this particular case the problem was the opposite: the newest udev > broke our persistent rule handling). > > The only package I have assumed might be anything but the most recent > is the kernel. Would anyone object to making this the official policy > for initscript, as I don't think we have the manpower to support other > combination of packages? If no one objects, I'll put a note about this > in the release notes. +1 from me. Noting in the release notes that "this version of initscripts requires udev >= xxx" seems appropriate. However, enforcing policy via provided mechanisms (versioned depends) is a good thing as far as I can see. -Dan
