On Fri, Dec 14, 2012 at 02:05:27PM -0500, Ian Stakenvicius wrote: > On 14/12/12 01:28 PM, Greg KH wrote: > > On Fri, Dec 14, 2012 at 11:43:41AM +0100, Fabian Groffen wrote: > >> Handling separate /usr support ============================== > >> After the discussion on [1] during the previous meeting, a delay > >> of one month due to a new fork of udev was requested. We need an > >> update on what's happened. > >> > >> Chainsaw reported udev and eudev have moved on, and for both it > >> is now possible to have a separate /usr. The follow-up > >> discussion related to the /usr-merge is necessary. > > > > udev was never the problem of having a separate /usr without an > > initrd. Have all of the other packages been properly fixed to > > resolve this issue correctly? > > > > Also, what's the plan for eudev going forward? > > > > > Eudev's project announcement is coming soon, should answer your questions.
Ok, when is "soon"? I'm guessing that the result of the council meeting ment that things are progressing, right? If so, in what way? > In terms of udev's dependencies, yes, the few dependencies that were > installing only to /usr (ie, kmod and xz-utils) have been switched to > install to /, and then fixed again due to issues with they way they > were done the first time so that they also work. I believe however > they are still ~arch keyworded. I am not referring to udev's dependancies, that was never the real issue with a separate /usr/ partition as those could easily be fixed with a configuration option for the package. > There may of course be other entirely independent packages needed at > boot time prior to localmount, I do not know that status of those. That's the big problem, those need to be fixed. > Once eudev (the gentoo package) fully supports separate-/usr (which it > doesn't at this time as it uses the same init scripts as udev-196), we > will be sure to resolve them. Again, udev itself was never an issue, it could work just fine with a separate /usr/ partition. Now perhaps our ebuild didn't build it in that matter, but that's a configuration/ebuild issue, not a upstream package issue. > It should be noted that sys-fs/udev (the package) since .. 186 I > think? whichever version dropped support for the failed-rules queue > (and whichever package dropped the udev-postmount init script) does > not support booting with a separate /usr. This has more to do with > how the package installs than the upstream code itself, though; as > such (WilliamH please correct me if I'm wrong) the plan is still to > require an initramfs if using sys-fs/udev with a separate-/usr. If the plan is still to require an initramfs (hint, it's the only way it can work), then why was the eudev package forked and created? Please, I'm totally confused now, especially after reading the commits in the eudev repo, I see nothing that fixed any /usr/ problems, what am I missing? Oh, you also slowed the build time of the package down in eudev compared to udev, but I'm sure you realized that already, and did it for a good reason. confused, greg k-h
