On Mon 2 Apr 2018 at 08:14:40 -0700, Ian Zimmerman wrote:
> This pipe is something of a Rube Goldberg device. Why not pass the
> variable directly:
>
> chroot $ROOT /usr/bin/env PS1="(chrooted to $HOST) $PS1" bash
That is of course a lot more elegant, I must have been half-asleep
when I wrote that pipe the first time.
> In fact I think I see a problem with your way: the chrooted shell sees a
> command like
>
> export PS1=You have chrooted into 'eden' from root
>
> which obviously cannot work. (No clue if that is why it breaks for
> Thelma, and no clue why it works for you :P)
What my syntax is doing is to let the $PS1 inside the PS1 definition
be evaluated by the chroot shell. Suppose you run this command with
HOST=eden and ROOT=/mnt/eden:
echo 'export PS1="(chroot '$HOST') $PS1"; exec <dev/tty' | exec chroot
$ROOT /bin/bash -i
The parent shell will translate this into
echo 'export PS1="(chroot 'eden') $PS1"; exec </dev/tty' | exec chroot
/mnt/eden /bin/bash -i
This is where the purpose of the 's around $HOST shows: $HOST is outside
the single quotes, so gets substituted, while the rest of the string,
notably $PS1, remains the same. The child shell will therefore receive
input
export PS1="(chroot eden) $PS1"; exec </dev/tty
which will prepend the desired text to the child shell’s prompt.
But indeed why bother if the $PS1 of the parent shell will do just as
well?
--
Sebastiaan L. Zoutendijk | [email protected]
