Bug 21222

This is how I fixed it.
(sopic) /etc% diff -u inputrc.orig inputrc
--- inputrc.orig        Fri Nov 17 15:09:53 2000
+++ inputrc     Fri Nov 17 15:10:10 2000
@@ -1,26 +1,7 @@
 # do not bell on tab-completion
-#set bell-style none
+set bell-style none
 
 set meta-flag on
 set input-meta on
 set convert-meta off
 set output-meta on
-
-"\eOd": backward-word
-"\eOc": forward-word
-
-# for linux console
-"\e[1~": beginning-of-line
-"\e[4~": end-of-line
-"\e[5~": beginning-of-history
-"\e[6~": end-of-history
-"\e[3~": delete-char
-"\e[2~": quoted-insert
-
-# for xterm
-"\eOH": beginning-of-line
-"\eOF": end-of-line
-
-#for freebsd console
-"\e[H": beginning-of-line
-"\e[F": end-of-line
(sopic) /etc% 

Sorry.  With all due respect, this is not a "fix."  The shell is supposed to allow command-line editing in its default state, as long as set -o vi has been 
invoked.  I shouldn't have to tweak anything to get this feature to work, and I shouldn't have to set up a work-around.  The editing feature works for root 
and for nobody else, and this is -- plainly and simply -- a bug.

Let me add that I'm appreciative of shane for the information about how he "fixed" the problem by revising the /etc/inputrc file.  I realize I might 
have come across as unappreciative in my previous additional comment.

The point I was trying to make is that root and non-root users are all pointed to the same /etc/inputrc file via the INPUTRC environment variable (I have 
confirmed this on my system).  root and non-root users are using the same instance of bash.  There is no documented difference in expected behavior of 
command-line editing between root and non-root users.

And yet, there IS a difference.  When shane changed a line and deleted some others from /etc/inputrc, this should have changed the behavior 
of the command-line editing for EVERYONE, not just non-root users.  It still doesn't explain why -- given the same /etc/inputrc file to init Readline with -- 
there is a behavioral difference between root and non-root.  In previous versions of RH, the default installation has not had this problem.  The new version 
of bash (2.04) may be the culprit, or perhaps the Readline library it's linking to is the culprit.  Perhaps a permission setting is screwed up somewhere?

I'm investigating this in my spare time, but it would be great if someone could pinpoint the root cause (no pun intended) of this problem.

I've also worked around this very annoying bug for a long time.  What I've
discovered is that
it works for non root users if you don't add set -o vi to .profile or it's
friends.  You log in and
type set -o vi by hand and then it works properly.  I certainly with gnu.org
would fix it.

In response to kitchen's comments:  My experience has not been the same as yours.  This is the first version of bash I've had this 
problem with.  Furthermore, in the current version, it doesn't matter if you put set -o vi in .profile or if you type it in after you log in.  The problem is the 
same.

The fix provided by shane cures the symptom.  The problem is, that it shouldn't have made a difference.  That is, root and the non-root users 
were all looking at /etc/inputrc as installed by the RedHat installer.  set -o vi worked with root, and not with anyone else.  Then, I put shane's 
changes into /etc/inputrc, and now set -o vi works for everybody!  That's great, except that it doesn't explain why there was a difference in behavior 
between root and non-root users to begin with.  There shouldn't have been.  Perhaps root is not actually looking at the contents of inputrc (even though 
the INPUTRC environment variable was set to /etc/inputrc), or perhaps certain inputrc options are ignored by root?  There is no documentation supporting 
a difference in behavior.

*** Bug 23014 has been marked as a duplicate of this bug. ***

FWIW, this seems to have been fixed for me in 7.1, Shane

This is finally fixed in the current set of related packages (bash, readline, 
setup).