Re: [gentoo-user] Masking udev to postpone the update

[Alan McKinnon]

On Thu, 22 Mar 2012 21:44:53 -0400
Allan Gottlieb <gottlieb@×××.edu> wrote:

> On Sun, Mar 18 2012, Allan Gottlieb wrote:
>
> > On Sun, Mar 18 2012, Alan McKinnon wrote:
> >
> >> On Sun, 18 Mar 2012 13:14:48 -0700
> >> Allan Gottlieb <gottlieb@×××.edu> wrote:
> >>
> >>> I will update to the new world order, but would very much prefer
> >>> to postpone that for a few weeks.  Is it enough to put
> >>>
> >>> >sys-fs/udev-171-r5
> >>>
> >>> in /etc/portage/package.mask  ?
> >>
> >>>=sys-fs/udev-181
> >>
> >> would be better. Rather mask the first version that causes issues
> >> and all subsequent versions. With your suggestions, there may be
> >> future updates between 171 and 181 (without initrd issues) that you
> >> want, but you can't use them as you masked them.
> >
> > Done, thanks.  Thank you volker as well.
> >
> > allan
>
> I am now unable to update world
>
>     Total: 26 packages (20 upgrades, 3 new, 1 in new slot, 2
> reinstalls, 1 uninstall), Size of downloads: 604,681 kB Conflict: 3
> blocks
>

I see you already have the solution from later in the thread.

Just don't do what I did and act like a dumbass like so:

1. emerge world including a couple days ago including udev-182
2. don't read the elog
3. miss the part where it says CONFIG_DEVTMPFS=y is required (and you
don't have it enabled)
4. don't run conf-update
5. walk away in blissful unawareness never rebooting just suspend/resume

Then:

1. Crawl out of bed one fine morning facing a deadline
2. Approach computer
3. Shit. The bloody thing hit some weird segfault again overnight and
killed X
4. Hmmmmmmmmm, virtual consoles don't seem to work. Machine frozen
5. Press big red power button

Only to find:

1. Nothing starts up properly, machine essentially useless
2. Oh dear. Other than /, nothing mounts. /dev is almost empty
3. That deadline didn't go away

To fix:

1. reboot into maintenance mode
2. <phew>, /usr is part of / so no mounting issues for that
3. find copy of elog, read it
4. configure and build kernel
5. Reboot
6. Didn't work. Scratch head, look around. Oh, hang on, the kernel was
installed to /boot on /, not the real /boot as that couldn't mount
7. Reboot, fiddle with grub menu, desperately trying to remember how
that shit all works again
8. Finally grub finds the new kernel and boots
9. Lots of errors, not quite the same as before. Something
about /sbin/udevd not found. Ohhhhhh shit......
10. Reboot again to maintenance mode
11. Poke around, bang head. Idiot! You forgot to run conf-update
12. conf-update lets everything else know udevd is now
in /lib64/udev/udevd
13. reboot. Finally, everything works again.
14. Realize new kernel is now in a directory underneath the /boot mount
(exactly where it can't be gotten to). Read man mount, search for --bind
15. Curse, swear, now 90 minutes late. Get on motorbike and roar off to
work in a foul mood dicing death by riding between motorcars causing
taking 2 old ladies by surprise and scaring 3 wandering cats witless


Moral of the story:

Read the elog right now and do what it says :-)



-- 
Alan McKinnnon
alan.mckinnon@×××××.com