Bash：从文件读取命令

Question

I'm using a minimal netinstall image of Debian. Here is my script to start my user processes and daemons at the first login. It reads a file consisting of commands line by line, checks if they are running already and launches them, if not:

#!/bin/bash
## This script reads a list of user startup daemons and launches them after the first login

STAT=`who | grep "$USER" | wc -l` ## Get login count
if [ "$STAT" -eq 1 ] ; then ## If login shell
  while IFS=' ' read line  ## First attempt to separate fields fails
  do
    CMD=`echo "$line" | grep -v ^\#` ## Use non comment lines only
    if [ -n "$CMD" ] ; then
      eval set "$CMD" ## The actual field separation
      APPID=$(pgrep "$1")&& ## Check if process $1 of $CMD is already active and launch it if not
        (echo "$1 already active as "$APPID) || ("$@" 1>/dev/null&& echo "Started $1 as $(pgrep "$1")")
    fi
  done < ~/.scripts/daemons
  echo "start-daemons: Finished"
else
  echo "start-daemons: Not a login shell"
fi

The file of commands ~/.scripts/daemons looks as following:

## List of processes / daemons to start at user login
## <command> <argument1> <argument2> ..
mpd
pulseaudio -D
gpg-agent --daemon --enable-ssh-support --write-env-file /tmp/.gpg-agent-info

The script works for mpd . The problem is, though, that lines consisting of a command and (multiple) options (and thus, spaces) will not be splitted into fields. Setting field separator IFS to ' ' makes no difference, which I find strange. What is the matter in this case? Thank you!

Edit: As suggested by alvits, a field separation through evaluating set $line within the while loop, provides an easy workaround. But why does read line no separation in the first place?

Answer 1

I think you should be checking as "pgrep pulseaudio" . That is, use just the process/daemon name rather than giving the process/daemon name along with the options. I don't have a machine where i can test this now. This you can do by taking the first field from each line(that will be the process/daemon name). Similarly for the second one. One more thing you can try is the "-f" option of pgrep. Try n let me know

Answer 2

Got it: read needs to know the fields a line should be splitted into. By changing line 6 to

 while IFS=' ' read field1 field2

field1 will contain the command, while field2 has the options, which can contain spaces again, since anything after field1 belongs to field2.