Linux – ps wide output for a single process

ps -C looks for an exact match of the basename of the executable at the time if was executed. So, in this case, ps -C perl or ps -C processname.pl would more likely be what you want. What matters is what you find in /proc/<pid>/stat.

If you want to match on the command line (the concatenation of the arguments of the command including argv[0]), you can use pgrep -f as found on Solaris, recent Linux and most BSDs.

Maybe something like:

#! /bin/bash -
pid=${1?Please give a pid}
sep=
while
  [ "$pid" -gt 0 ] &&
    read -r ppid name < <(ps -o ppid= -o args= -p "$pid")
do
  printf "$sep%$((3+${#pid}/2))d%$((4-${#pid}/2))s%s\n" "$pid" "" "$name"
  sep=$'  |\n'
  pid=$ppid
done

Which gives:

$ that-script "$$"
13612  /bin/zsh
  |
 4137  SCREEN screen -xRS main
  |
 4136  screen screen -xRS main
  |
 4130  zsh
  |
 4128  xterm
  |
  1    /sbin/init splash

That's a bit sluggish as it runs one ps per process in the ancestry. You can improve it by running ps to retrieve all the information at once and post process it with awk for instance:

#! /bin/sh -
pid=${1?Please give a pid}
ps -Ao pid= -o ppid= -o args= |
  awk -v p="$pid" '
    {
      pid = $1; ppid[pid] = $2
      sub(/([[:space:]]*[[:digit:]]+){2}[[:space:]]*/, "")
      args[pid] = $0
    }
    END {
      while (p) {
        printf "%s%*d%*s%s\n", sep, 3+length(p)/2, p, 4-length(p)/2, "", args[p]
        p = ppid[p]
        sep = "  |\n"
      }
    }'