[arch-dev-public] [signoff] kbd and klibc-kbd

[Roman Kyrylych]

2008/3/22, Xavier <shiningxc at gmail.com>:
> Roman Kyrylych a écrit :
>
> > 2008/3/21, Tobias Powalowski <t.powa at gmx.de>:
>  >>  somehow we need a depend that also adds klibc-kbd, else systems get broken
>  >>  that have keymap hook activated.
>  >>  probably kbd package itself needs this depend.
>  >
>  > Just an info for others who reads this,
>  > we discussed the issue on IRC and there are actually 2 issues:
>  > 1) during -Syu kbd (which doesn't include keymap hook now) is
>  > installed before kernel,
>  > then mkinitcpio is run during kernel install and it fails because of
>  > no hook available.
>  > 2) klibc-kbd is not installed as a new dependency of to-be-upgraded mknitcpio,
>  > it looks to me like some weird pacman bug. :-/
>  >
>  > #1 can be solved by changing depends=(... mkinitcpio>=0.5.15 ..) in
>  > kernel26 PKGBUILD to the current mkinitcpio version (which depends on
>  > klibc-kbd)
>
>
> I don't understand, what does that change? Though it's probably not a
>  bad idea to do it.

The order was kbd, kernel, klibc-kbd (it should be here, but wasn't!),
mkinitcpio.
New kbd doesn't have keymap hook, so during kernel install mkinitcpio will fail.
When kernel depends on the new mkinitcpio, the order will be:
kbd, klibc-kbd, mkinitcpio, kernel
which effectively solves the issue.

>  > but if #2 is a real bug - this won't work anyway, so #2 should be investigated.
>  > Tobias, could you please resend that part of pacman.log to pacman-dev
>  > ML or at least to Dan?
>  >
>
>
> pacman seems to work fine here. If it didn't pull new dependencies of
>  packages, that would be pretty bad.
>  More details and a part of pacman.log to pacman-dev would indeed be
>  appreciated :)

Ok, I will repost pacman.log there by myself.

-- 
Roman Kyrylych (Роман Кирилич)