Help interpreting shell script

NotionCommotion · 04-29-2017, 11:09 AM

I've gone through the following script, and added comments describing my interpretation or questions of various lines. Note that these are not normal #comments, but just `my comments to you indicated by backticks`. Thank you

Code:

`use shell command language which in my case is a symlink to bash`
#!/bin/sh

`comments which should follow this specific format`
### BEGIN INIT INFO
# Provides:          Test SOAP Simulator
# Required-Start:    $local_fs $network $named $time $syslog
# Required-Stop:     $local_fs $network $named $time $syslog
# Default-Start:     2 3 4 5
# Default-Stop:      0 1 6
# Short-Description: Start SOAP Simulator
# Description:       Initiate a given PHP file every 5 seconds to simulate a SOAP server.
### END INIT INFO

`define some global variables`
SCRIPT= /usr/local/bin/test_soap
RUNAS=Michael
PIDFILE=/var/run/test_soap.pid
LOGFILE=/var/log/test_soap.log

start() {
  `Need a little help on this one`
  if [ -f "$PIDFILE" ] && kill -0 $(cat "$PIDFILE"); then
	`send 'Service already running' to stdErr?  What is with the &amp symbol`
    echo 'Service already running' >&2
	`return 1 indicated an error to the switch statement.  What ultimately happens to this value?`
    return 1
  fi
  echo 'Starting service…' >&2
  `create the local variable CMD for a command which will executes a given script and sends its output to a given file, and also echos the command on the screen?  what is with the ! symbol?`
  local CMD="$SCRIPT &> \"$LOGFILE\" & echo \$!"
  `Using user "Michael", execute the command which will return an integer, and store that integer in the test_soap.pid file`
  su -c "$CMD" $RUNAS > "$PIDFILE"
  echo 'Service started' >&2
}

stop() {
  if [ ! -f "$PIDFILE" ] || ! kill -0 $(cat "$PIDFILE"); then
    echo 'Service not running' >&2
    return 1
  fi
  echo 'Stopping service…' >&2
  `please help with the kill -15 ... part.  rm part is clear`
  kill -15 $(cat "$PIDFILE") && rm -f "$PIDFILE"
  echo 'Service stopped' >&2
}

status() {
    `get info about the process by pid. Is $$ actually a variable name?  Where is it defined?  Send output to bottemless pit, but still make it available to if/else statement?`
    if ps -p $$ > /dev/null
    then
       echo "$prog is running"
    else
       echo "$prog is not running"
    fi
    RETVAL=$?
    return $RETVAL
}

case "$1" in
  start)
    start
    ;;
  stop)
    stop
    ;;
  restart)
    stop
    start
    ;;
  status)
    status
    ;;
  *)
    echo "Usage: $0 {start|stop|restart|status}"
esac

hazel · 04-29-2017, 12:45 PM

Quote:

Originally Posted by NotionCommotion

I've gone through the following script, and added comments describing my interpretation or questions of various lines. Note that these are not normal #comments, but just `my comments to you indicated by backticks`. Thank you

start() {
`Need a little help on this one`
if [ -f "$PIDFILE" ] && kill -0 $(cat "$PIDFILE"); then
`send 'Service already running' to stdErr? What is with the &amp symbol`
echo 'Service already running' >&2

>&2 means "Send this output to where the error stream is currently going". echo normally prints to standard output.

Quote:

`return 1 indicated an error to the switch statement. What ultimately happens to this value?`
return 1
fi

It becomes the return value for the script.

Quote:

echo 'Starting service…' >&2
`create the local variable CMD for a command which will executes a given script and sends its output to a given file, and also echos the command on the screen? what is with the ! symbol?`
local CMD="$SCRIPT &> "$LOGFILE" & echo \$!"

$! is a special variable representing the pid of the job that has been launched

Quote:

`please help with the kill -15 ... part. rm part is clear`
kill -15 $(cat "$PIDFILE") && rm -f "$PIDFILE"
echo 'Service stopped' >&2

15 is the signal number for SIGTERM (terminate). You can find the full list in /usr/include/bits/signum.h.

Quote:

status() {
`get info about the process by pid. Is $$ actually a variable name? Where is it defined? Send output to bottemless pit, but still make it available to if/else statement?`
if ps -p $$ > /dev/null

$$ is another of bash's special variables. It represents the script's own pid. You can get a list of them from the bash man page: just go to Special Parameters.

NotionCommotion · 04-29-2017, 01:29 PM

Thank you Hazel, Your remarks were very helpful.

Code:

su -c "/usr/local/bin/test_soap &> '/var/log/test_soap.log' & echo \$!" Michael > "/var/run/test_soap.pid"

To confirm, this will create job test_soap as user Michael, send its output to test_soap.log, echo the PID of this recently created job, and send it to test_soap.pid. Correct?

Code:

kill -15 $(cat "$PIDFILE")

The 15 is redundant, correct?

Code:

if [ -f "$PIDFILE" ] && kill -0 $(cat "$PIDFILE"); then

If the file exists and we error checking indicated that it can be killed?

hazel · 04-29-2017, 01:57 PM

Quote:

Originally Posted by NotionCommotion

Thank you Hazel, Your remarks were very helpful.

Code:

su -c "/usr/local/bin/test_soap &> '/var/log/test_soap.log' & echo \$!" Michael > "/var/run/test_soap.pid"

That doesn't look quite right to me. I think it should be "&& echo", in other words a boolean "and" separating the two commands.

Quote:

To confirm, this will create job test_soap as user Michael, send its output to test_soap.log, echo the PID of this recently created job, and send it to test_soap.pid. Correct?

I think so. But I wouldn't like to testify to it in court! The syntax is a bit too complicated for my knowledge of bash.

Quote:

Code:

kill -15 $(cat "$PIDFILE")

The 15 is redundant, correct?

Yes, I think it is. According to the kill man page, SIGTERM is the default kill signal.

Quote:

Code:

if [ -f "$PIDFILE" ] && kill -0 $(cat "$PIDFILE"); then

If the file exists and we error checking indicated that it can be killed?

If the file exists, it could either be because the process is already running or because the pidfile was not deleted properly when the job last ended. These two situations can be distinguished by sending an empty message to that pid via kill; success means that the process is running and doesn't have to be started.

NotionCommotion · 04-29-2017, 05:28 PM

Quote:

Originally Posted by hazel

That doesn't look quite right to me. I think it should be "&& echo", in other words a boolean "and" separating the two commands.

I think so. But I wouldn't like to testify to it in court! The syntax is a bit too complicated for my knowledge of bash.

Code:

  local CMD="$SCRIPT &> \"$LOGFILE\" & echo \$!"
  su -c "$CMD" $RUNAS > "$PIDFILE"

This is the line in question before I inserted the various variables in an attempt to make it more readable, but likely making it less.

The second line seems clear. Have given user run command and send output to pidfile.

The first line, however, is less clear.

Run a script, send its stdout to a file. Okay...

& echo or && echo? It is not a ternary, so why do we want Boolean here? But if not Boolean, what is it?

Lastly, echo \$!. Okay, I understand $! now. Is the backslash needed to escape something?

rknichols · 04-29-2017, 07:57 PM

Quote:

Originally Posted by NotionCommotion

Code:

  local CMD="$SCRIPT &> \"$LOGFILE\" & echo \$!"

"&>": Redirect both stdout and stderr.

"&": Run the preceding command in the background

"echo \$!": Print the PID of the background job that was just created. The backslash is necessary because you want to delay the expansion of "$!" until the command is actually run, not at the time that string is being assigned to the CMD variable.

[EDIT] I don't think that PID is going to be printed correctly, because that string gets parsed again when "$CMD" is passed as an argument to the su command. That probably needs to be

Code:

  local CMD="$SCRIPT &> \"$LOGFILE\" & echo \\\$!"

BW-userx · 04-29-2017, 08:04 PM

is this

Code:

if [ -f "$PIDFILE" ] && kill -0 $(cat "$PIDFILE"); then
	`send 'Service already running' to stdErr?  What is with the &amp symbol`
    echo 'Service already running' >&2
	`return 1 indicated an error to the switch statement.  What ultimately happens to this value?`
    return 1
fi

just to check if both pid number and pid file are present? if yes then echo is running else return 1 for error?

in other words if true (has PID ) and true (has PID-FILE) then don't run script?

off the command line

Code:

userx%slackwhere ⚡ ~ ⚡> 

if [ true ]  && [ true ] ; then echo "what" ; fi
what

maybe written like this?

Code:

if [ -f "$PIDFILE" ] && [ kill -0 $(cat "$PIDFILE") ] ; then
     echo 'Service already running' > /dev/null 2>>/dev/null

    # if this is ran off the command line might 
    # it not be a good idea to send message to term instead of this?

    echo " already running bye " 

   #stop script from running any further. 
   exit 0
fi

is it really an error just because it is already running?

Chapter 6. Exit and Exit Status

it also looks like your mixing methods here Boolean operators with conditionals
yours

Code:

 if [ ! -f "$PIDFILE" ] || ! kill -0 $(cat "$PIDFILE"); then
    echo 'Service not running' >&2
    return 1

should be one or the other

Code:

[ test ] && do something if return is 0

[ test ] || do something if  return is anything other than 0

# all of these fire if true

if [ condition true ] ; then
do something
fi

#yes even this one is checking for a true statement

if [ ! true ] ; then
do something
fi

# if this or that
if [ something ] || [ something ] ; then
do something
fi

# if both true
if [[ true && true ]] ; then
do something 
fi

ect..

[[ ]] and [ ] get interpreted differently as well.

example
off one of my scripts nested Boolean with conditionals

Code:

#conditional if [ test condition ] ; then fi
 
 if [[ "$ext" == 'flac' ]] ; then
 
# Boolean expression or logical expresion

  [[ && -n "$ALBUM" && -n "$ARTIST" ]] && ( NewFile="$title"."mp3" ; flac -cd "$FILENAME" | lame -b 320 - "$NewFile" ; printf "num 3 script_dir - NewFile -> $script_dir/$NewFile\n $ArtistAlbum\n" ;  mv -v "$script_dir/$NewFile" "$ArtistAlbum" ; \
  id3v2 -A "$ALBUM" "$ArtistAlbum/$NewFile" ; id3v2 -a "$ARTIST" "$ArtistAlbum/$NewFile" ; id3v2 -t "$TITLE" "$ArtistAlbum/$NewFile" ; id3v2 -g "$GENRE" "$ArtistAlbum/$NewFile" )

# Boolean expression or logical expresion

  [[ -f "$ArtistAlbum/$NewFile" ]] && id3v2 -T "$track" "$ArtistAlbum/$NewFile"  

# Boolean expression or logical expresion

  [[ ! -n "$ALBUM" && -n "$ARTIST" ]] && ( Artist="$move_to/$ARTIST" ; mkdir -pv "$Artist" )

  Artist="$move_to/$ARTIST"
# Boolean expression or logical expresion

 [[ -n "$ARTIST" &&  ! -n "$ALBUM" ]] && ( NewFile="$title"."mp3" ; flac -cd "$FILENAME" | lame -b 320 - "$NewFile" ;  printf "script_dir - NewFile -> $script_dir/$NewFile\n" ; mv -v "$script_dir/$NewFile" "$Artist" ; \
id3v2 -A "$ALBUM" "$Artist/$NewFile" ; id3v2 -a "$ARTIST" "$Artist/$NewFile" ; id3v2 -t "$TITLE" "$Artist/$NewFile" ; id3v2 -g "$GENRE" "$Artist/$NewFile"  )
# Boolean expression or logical expresion


[[ -f "$Artist/$NewFile"  ]] &&  id3v2 -T "$track" "$Artist/$NewFile" 

# end if [ condistion ]  ; then fi
fi

you can see the use of both together and the use of

[[ true && true]] && if both true then do something (returned 0 )

and the use of

[[ true && not "true" ]] && if both truth then do something (returned 0 )

this ! true still evaluates to true because you are checking to be sure it is not there. if it is not then that is a 'yes it is not there' or true

do not quote me on the proper names for these operators

grail · 04-30-2017, 02:44 AM

I think most of your answers have been supplied, so I will ask a question ... Did you give us the original init script?

The reason for my question is because it refers to a variable ($prog) which is never assigned and hence will be empty, so your status output will be "is running" or "is not running"

I do have some small input on some of your comments:

use shell command language which in my case is a symlink to bash :- The line in question does not have to do with the shell, it is the interpreter line, hence you can change it to refer to perl (which is not a shell). Also, important note, when referring to bash via the symlink sh actually tells bash to run in POSIX mode, hence certain bash specific constructs will not be available, but makes the overall script
more usable in more environments which might not have bash as the default

comments which should follow this specific format :- more than just the format, the comments you are referring to give specific details about this particular init script

create the local variable CMD for a command which will executes a given script and sends its output to a given file, and also echos the command on the screen? :- just clarifying, it will NOT echo to the screen

Send output to bottemless pit, but still make it available to if/else statement? :- No there are no details being made available to the if/else. The 'if' is a command and checks the return value of the command presented to it (ps in this case)

One last point, whilst the kill commands default may be SIGTERM, the inclusion of the switch '-15' makes it unambiguous. Also, on other systems this may not necessarily be the default action (thinking AIX or Solaris)

pan64 · 04-30-2017, 04:59 AM

to BW-userx:
you still do not understand how it works (in general). the if/then/else structure works with commands:

Code:

if command1;
then command2;
fi

#and not on 
if [ command ]
# or
if [[ command ]]

actually [ and [[ are commands themselves and if only evaluates their (or any other commands') exit code. Nothing else, nothing more.

this means for example:

Code:

if [ true ]  && [ true ] ; then echo "what" ; fi
# you should write instead:
if true && true; then ....
#(true and false are both commands again)

or

Code:

if [ -f "$PIDFILE" ] && [ kill -0 $(cat "$PIDFILE") ] ; then
# is incorrect, or at least I do not know what does [ kill ] do

furthermore

Code:

  [[ && -n "$ALBUM" && -n "$ARTIST" ]]
# looks strange (from your script) - at least for me

NotionCommotion · 04-30-2017, 09:42 AM

Thanks rknichols, Ah, I now see how "&>" does as you say. bash has such a long manual, it is difficult to remember everything. Couple more specific questions regarding this line alone.

Were escaped double quotes required, or can single quotes be used around $LOGFILE?
Does the order of redirecting or the background &amp make a difference?
Is the backslash to delay expanding $! a bash or echo arguments? I can't find anything on it.
Guess this last one relates to two lines. Are the following two sections the same?

Code:

local CMD="$SCRIPT &> "$LOGFILE" & echo \$!"
su -c "$CMD" $RUNAS > "$PIDFILE"

Code:

local CMD="$SCRIPT &> "$LOGFILE" &"
su -c "$CMD" $RUNAS
echo \$! > "$PIDFILE"

Thanks grail, No, it wasn't quite the original script as I tried to add the status portion. You are correct that $prog isn't defined, and I needed to do so in my original declaration of global variables.

rknichols · 04-30-2017, 10:22 AM

Quote:

Originally Posted by NotionCommotion

Thanks rknichols, Ah, I now see how "&>" does as you say. bash has such a long manual, it is difficult to remember everything. Couple more specific questions regarding this line alone.

Were escaped double quotes required, or can single quotes be used around $LOGFILE?
Does the order of redirecting or the background &amp make a difference?
Is the backslash to delay expanding $! a bash or echo arguments? I can't find anything on it.
Guess this last one relates to two lines. Are the following two sections the same?

Code:

local CMD="$SCRIPT &> "$LOGFILE" & echo \$!"
su -c "$CMD" $RUNAS > "$PIDFILE"

Code:

local CMD="$SCRIPT &> "$LOGFILE" &"
su -c "$CMD" $RUNAS
echo \$! > "$PIDFILE"

The escaped double quotes are required. You want those quotes to be present literally in the string assigned to CMD. Quotes do not nest. If you don't escape those interior quotes, what you have are two quoted strings, "$SCRIPT &> " and " & echo \$!" with the unquoted expansion of LOGFILE between them. Really, as long as LOGFILE doesn't contain any characters that need to be quoted, it doesn't matter. But in that case, those escaped double quotes could have been omitted entirely.
That "&" putting the command in the background acts as a command separator, just like ";". I don't think you would be asking if the placement of a ";" was significant.
The backslash escape on "\$!" simply tells bash to take that "$" character literally and not as the start of a parameter expansion. The two-character sequence "$!" will thus be included literally in the CMD string. If you don't do that, the "$!" will be expanded as part of the variable assignment, i.e., to the PID of the last background process (of which there probably was none) started prior to doing the assignment. It's like saying "X=$Y" before assigning any value to Y.
No, they are not the same. In the first, the echo command is part of the string passed to su to be executed under the new UID. In the second, the echo command is run by the current shell under the current UID. Since the backgrounded process is not a direct child of the current shell, a "$!" in the current shell would not expand to its PID.

Note also the addendum to my previous post. There are not enough backslashes in that "\$!" to get it expanded at the proper time. This is the problem with using shell variables to hold command strings. The strings get evaluated multiple times by the shell, and it is very hard to get quoting and escaping to work the way you want. In most cases, shell functions are a much preferable alternative, but that just doesn't work in the context of passing a command string to su for execution.

It is useful to run your script with the shell's "-x" option. That way, you can see how the strings change each time they are parsed and what actually gets passed to each command.

NotionCommotion · 04-30-2017, 11:37 AM

Thanks again rknichols,

Yes, I understand quotes cannot be nested. I was talking about 'single quotes' and "escaped double quotes". Granted, I just noticed on my last post, I didn't show it this way, so see how my question was misleading.
Regarding position, I was asking whether "$SCRIPT &> "$LOGFILE" & echo \$!" is different than "$SCRIPT & &> "$LOGFILE" echo \$!"
Thanks
I should have asked if the following are the same:

Code:

local CMD="$SCRIPT &> "$LOGFILE" & echo \$!"
su -c "$CMD" $RUNAS > "$PIDFILE"

Code:

local CMD="$SCRIPT &> "$LOGFILE" &"
su -c "$CMD" $RUNAS
su -c echo \$! $RUNAS > "$PIDFILE"

Yes, I saw your addendum, and will experiment to ensure I get the correct results. Also, thanks for pointing out the shell's "-x" option as it will make testing much easier.

rknichols · 04-30-2017, 01:30 PM

Quote:

Originally Posted by NotionCommotion

Yes, I understand quotes cannot be nested. I was talking about 'single quotes' and "escaped double quotes". Granted, I just noticed on my last post, I didn't show it this way, so see how my question was misleading.

There is no bit of magic that can make this simple. The string that you write is going to be parsed by a shell 3 times: (1) when the string is assigned to the CMD variable, (2) when the contents of $CMD are expanded and passed to su, and (3) when the entire command is evaluated by the final shell that will execute it. If the value of LOGFILE contains spaces or other characters that are special to the shell, you have to consider what that whole string will look like at each step along the way and how each shell will interpret it. And even the escaped double quotes you have aren't really sufficient. That final shell is still going to see that LOGFILE string unquoted.

Quote:

Regarding position, I was asking whether "$SCRIPT &> "$LOGFILE" & echo \$!" is different than "$SCRIPT & &> "$LOGFILE" echo \$!"

Those are very different. The "&" is a command separator, like <newline> or ";". Ignoring the quoting problems, the first is equivalent to these two command lines

Code:

$SCRIPT &> "$LOGFILE" &
echo \$!

The second is equivalent to

Code:

$SCRIPT &
&> "$LOGFILE" echo \$!"

which runs $SCRIPT in the background without the $LOGFILE redirection and then directs just the echoed PID to $LOGFILE, which is not anything like what you want.

Quote:

I should have asked if the following are the same:

Code:

local CMD="$SCRIPT &> "$LOGFILE" & echo \$!"
su -c "$CMD" $RUNAS > "$PIDFILE"

Code:

local CMD="$SCRIPT &> "$LOGFILE" &"
su -c "$CMD" $RUNAS
su -c echo \$! $RUNAS > "$PIDFILE"

No, that second one is running two separate shells, and the second shell will have no knowledge of the background process launched by the other shell.

NotionCommotion · 04-30-2017, 01:51 PM

Need to spend a bit more time, but makes sense.

Last one (I think!).

Consider the following. Maybe I don't want Michael to have the ability to write to /var/run/ or /var/log/. But as the command is executed by Michael, he needs to have that ability. How can I have my cake and eat it too?

Also, my status doesn't make sense but cause this script will also have a PID when it is being run. What I need to do is test the pid using the value stored in test_soap.pid. I've tried cat'n the value to no success. How is this accomplished?

Thank you for all your help!

Code:

RUNAS=Michael
PIDFILE=/var/run/test_soap.pid
LOGFILE=/var/log/test_soap.log

status() {
  #...
  local CMD="$SCRIPT &> "$LOGFILE" & echo \$!"
  su -c "$CMD" $RUNAS > "$PIDFILE"
  #...
}
status() {
    if ps -p $$ > /dev/null
    then
       echo "is running"
    else
       echo "is not running"
    fi
    RETVAL=$?
    return $RETVAL
}

rknichols · 04-30-2017, 02:39 PM

Quote:

Originally Posted by NotionCommotion

Consider the following. Maybe I don't want Michael to have the ability to write to /var/run/ or /var/log/. But as the command is executed by Michael, he needs to have that ability. How can I have my cake and eat it too?

You could play games with file descriptors and move the output redirection outside the su command:

Code:

RUNAS=Michael
PIDFILE=/var/run/test_soap.pid
LOGFILE=/var/log/test_soap.log

start() {
  #...
  local CMD="$SCRIPT &> &3 & echo \$!"
  su -c "$CMD" $RUNAS  3>"$LOGFILE" >"$PIDFILE"
  #...
}

That opens $LOGFILE on file descriptor 3 before invoking su, then redirects the command's stdout and stderr to that opened fd3. The redirection onto "$PIDFILE" is already outside the su command. Of course the current user (the one invoking this script) must have write permission in those directories.

Quote:

Also, my status doesn't make sense but cause this script will also have a PID when it is being run. What I need to do is test the pid using the value stored in test_soap.pid. I've tried cat'n the value to no success. How is this accomplished?

Should be simple enough:

Code:

if ps -p $(cat $PIDFILE) >/dev/null; then

Or, a bit simpler

Code:

if [ -d /proc/$(<$PIDFILE) ]; then ...

Note that "$(<$PIDFILE)" is just a convenient shorthand for "$(cat $PIDFILE)" that avoids a "useless use of cat".