SIGINT to cancel read in bash script?

Question

I'm writting a bash wrapper to learn some scripting concepts. The idea is to write a script in bash and set it as a user's shell at login.

I made a while loop that reads and evals user's input, and then noticed that, whenever user typed CTRL + C, the script aborted so the user session ends.

To avoid this, I trapped SIGINT, doing nothing in the trap.

Now, the problem is that when you type CTRL + C at half of a command, it doesn't get cancelled as one would do on bash - it just ignores CTRL + C.

So, if I type ping stockoverf^Cping stackoverflow.com, I get ping stockoverfping stackoverflow.com instead of the ping stackoverflow.com that I wanted.

Is there any way to do that?

#!/bin/bash

# let's trap SIGINT (CTRL + C)
trap "" SIGINT

while true
do
    read -e -p "$USER - SHIELD: `pwd`> " command
    history -s $command
    eval $command
done

Answer 1

I know this is old as all heck, but I was struggling to do something like this and came up with this solution. Hopefully it helps someone else out!

#/usr/bin/env bash
# Works ok when it is invoked as a bash script, but not when sourced!
function reset_cursor(){
    echo
}
trap reset_cursor INT
while true; do
    command=$( if read -e -p "> " line ; then echo "$line"; else echo "quit"; fi )
    if [[ "$command" == "quit" ]] ; then
        exit
    else
        history -s $command
        eval "$command"
    fi
done
trap SIGINT

By throwing the read into a subshell, you ensure that it will get killed with a sigint signal. If you trap that sigint as it percolates up to the parent, you can ignore it there and move onto the next while loop. You don't have to have reset_cursor as its own function but I find it nice in case you want to do more complicated stuff.

I had to add the if statement in the subshell because otherwise it would ignore ctrl+d - but we want it to be able 'log us out' without forcing a user to type exit or quit manually.