/etc/profile always sets INPUTRC to /etc/inputrc, thus making it impossible for a user to set its own .inputrc. I for example configure some gdb keybindings in .inputrc among various other stuff. The following code is IMHO better in any case. : if [ -z "$INPUTRC" -a ! -f "$HOME/.inputrc" ]; then : INPUTRC=/etc/inputrc : export INPUTRC : fi The corresponding code for *csh like shells is left as an exercise for the reader :-)
I was just about to report this same bug. The patch that drdisk provided for /etc/profile to conditionally set INPUTRC is a good idea, but a better solution to this problem would be to have the readline package contribute an /etc/skel/.inputrc file that contains this line: $include /etc/inputrc For backwards compatibility with old accounts that weren't created with an ~/.inputrc file, the /etc/profile file could just use the conditional code to set INPUTRC that drdisk provided. Note that if the readline package contributes /etc/skel/.inputrc, it might be a good idea to make it part of the base (mandatory) RPM system. Currently, it isn't; I can "rpm --test -e" it so long as I also remove all of the packages that depend on it. Optionally, you might be able to convince Chet Ramey to make /etc/inputrc a default location for a system-wide inputrc file, and have libreadline cascade them (first read /etc/inputrc, then read ~/.inputrc), the same way bash does for the /etc/profile and ~/.bash_profile files. It seems like that level of customization is what is really needed here, and no backwards-compatibility checks would be required.
BTW, just to clarify, I noticed the issues with INPUTRC for a different reason: the default /etc/inputrc that comes with Red Hat contains: set convert-meta off Under bash2, that setting makes it so that all the cool key bindings I'm used to (M-f for forward-word, M-d for backward-kill-word, etc.) generate Latin-1 characters instead. That's useful in the international scheme of things, to be sure, but that's definitely not the behavior I want; I want the "set convert-meta on" behavior.
Fixed in setup-2.0.9, etcskel-2.1, both of which should be in the next Raw Hide release.