Re: Subversion Init script

[Peter Samuelson <pe...@p12n.org>]

[S.H.Mohanjith]
> I have attached an Init script for subversion. The script should be
> placed in /etc/init.d/.

And yet it mentions /etc/rc.d/init.d/.  That is a legacy location.

> I didn't get the chance to test on many distros.  :-( I would like to
> correct any portability issues, let me know if there are any.

There are lots of portability issues.

1) source /etc/rc.d/init.d/functions
   I guess this is for the 'daemon', 'success' and 'failure' functions.
   But in any case, that file doesn't exist on most systems.  Also,
   "source" is a bashism - see 3).

2) [ -x /usr/bin/svnserve ] || exit 1
   Not everybody installs svnserve to /usr/bin.  Use autoconf.

3) SYSCONFIG="/etc/sysconfig/subversion"
   Pretty sure /etc/sysconfig is Red Hat-specific.  Debian, for
   example, uses /etc/default.  I don't know whether there is a
   standard location for this sort of thing.

4) echo -n $"Starting $desc ($prog): "
   echo -n $"Shutting down $desc ($prog): "
   echo $"Usage: $0 {start|stop|forcestop|restart|condrestart|status}"
   I have no idea why you want to use the $"" construct here, but it is
   a bashism.  On other POSIX shells, it won't do what you think.  I
   suppose it's acceptable to use bashisms in a script labeled
   #!/bin/bash, but is there a need to require bash to be installed on
   every Subversion user's system?

5) touch /var/lock/subsys/$prog
   /var/lock/subsys doesn't exist on most systems.

6) daemon $prog -d $OPTIONS --pid-file $pidfile
   [ $RETVAL -eq 0 ] && success || failure
   See 1).

7) pidstr=`pgrep $prog`
   I think pgrep is specific to Linux (in particular, procps).

8) kill -s 3 $pid
   I think "kill -QUIT" is more portable - if you think SIGQUIT is
   appropriate here.  I think the default kill signal, SIGTERM, is
   better.  That is the traditional way to ask a process to die.

9) I know the LSB is a Linux-specific issue, but your script is in
   several ways not LSB-compliant (see
   http://refspecs.freestandards.org/LSB_3.1.0/LSB-Core-generic/LSB-Core-generic/iniscrptact.html).
   Some LSB-related issues I noticed:

   9a) You don't support the required 'force-reload' action.

   9b) The 'status' and 'stop' actions do not return the documented
       error codes.  ('start' may or may not be compliant; I don't know
       the return values for the Red Hat 'daemon' function.)

   9c) The script fails if svnserve is not installed - the LSB says it
       should silently succeed.

   9d) There are no LSB header comments.  (Yes, I know, these are
       optional.)

   9e) You don't use LSB facilities for status messages.  This is a
       problem for LSB compliance, but OTOH, those facilities are not
       portable beyond LSB platforms, so what you're doing now may be
       better.

> Please let me know if there are any other issues in the Init script as well.

Hmmmm.

1) # pidfile: /var/lock/subsys/svnserve
   While this is just a comment, it isn't accurate.  You use
   /var/run/svnserve.pid elsewhere.

2) if [ $pid ]; then
   If you're "fluent in shell scripting", I hope I don't need to point
   out the problem with this line....  (:

3) As the LSB notes, admins may run init scripts with strange values in
   PATH and other shell variables.  You should set your own PATH if you
   need it - which you do.

4) Stylistically, I don't like your mixed $VARIABLE vs. $variable
   names.  And you should probably define a variable for
   /var/lock/subsys/$prog, especially since that's the wrong path on
   some systems anyway.

5) You don't provide a facility for running svnserve as a non-root
   user.  I suspect most admins would prefer to specify a dedicated
   user.  Use 'su' for this.

Re: Subversion Init script

["S.H.Mohanjith" <mo...@mohanjith.net>]

Peter Samuelson wrote:
> [S.H.Mohanjith]
>   
>> I have attached an Init script for subversion. The script should be
>> placed in /etc/init.d/.
>>     
>
> And yet it mentions /etc/rc.d/init.d/.  That is a legacy location.
>
>   
>> I didn't get the chance to test on many distros.  :-( I would like to
>> correct any portability issues, let me know if there are any.
>>     
>
> There are lots of portability issues.
>
> 1) source /etc/rc.d/init.d/functions
>    I guess this is for the 'daemon', 'success' and 'failure' functions.
>    But in any case, that file doesn't exist on most systems.  Also,
>    "source" is a bashism - see 3).
>
> 2) [ -x /usr/bin/svnserve ] || exit 1
>    Not everybody installs svnserve to /usr/bin.  Use autoconf.
>
> 3) SYSCONFIG="/etc/sysconfig/subversion"
>    Pretty sure /etc/sysconfig is Red Hat-specific.  Debian, for
>    example, uses /etc/default.  I don't know whether there is a
>    standard location for this sort of thing.
>
> 4) echo -n $"Starting $desc ($prog): "
>    echo -n $"Shutting down $desc ($prog): "
>    echo $"Usage: $0 {start|stop|forcestop|restart|condrestart|status}"
>    I have no idea why you want to use the $"" construct here, but it is
>    a bashism.  On other POSIX shells, it won't do what you think.  I
>    suppose it's acceptable to use bashisms in a script labeled
>    #!/bin/bash, but is there a need to require bash to be installed on
>    every Subversion user's system?
>
> 5) touch /var/lock/subsys/$prog
>    /var/lock/subsys doesn't exist on most systems.
>
> 6) daemon $prog -d $OPTIONS --pid-file $pidfile
>    [ $RETVAL -eq 0 ] && success || failure
>    See 1).
>
> 7) pidstr=`pgrep $prog`
>    I think pgrep is specific to Linux (in particular, procps).
>
> 8) kill -s 3 $pid
>    I think "kill -QUIT" is more portable - if you think SIGQUIT is
>    appropriate here.  I think the default kill signal, SIGTERM, is
>    better.  That is the traditional way to ask a process to die.
>
> 9) I know the LSB is a Linux-specific issue, but your script is in
>    several ways not LSB-compliant (see
>    http://refspecs.freestandards.org/LSB_3.1.0/LSB-Core-generic/LSB-Core-generic/iniscrptact.html).
>    Some LSB-related issues I noticed:
>
>    9a) You don't support the required 'force-reload' action.
>
>    9b) The 'status' and 'stop' actions do not return the documented
>        error codes.  ('start' may or may not be compliant; I don't know
>        the return values for the Red Hat 'daemon' function.)
>
>    9c) The script fails if svnserve is not installed - the LSB says it
>        should silently succeed.
>
>    9d) There are no LSB header comments.  (Yes, I know, these are
>        optional.)
>
>    9e) You don't use LSB facilities for status messages.  This is a
>        problem for LSB compliance, but OTOH, those facilities are not
>        portable beyond LSB platforms, so what you're doing now may be
>        better.
>
>   
>> Please let me know if there are any other issues in the Init script as well.
>>     
>
> Hmmmm.
>
> 1) # pidfile: /var/lock/subsys/svnserve
>    While this is just a comment, it isn't accurate.  You use
>    /var/run/svnserve.pid elsewhere.
>
> 2) if [ $pid ]; then
>    If you're "fluent in shell scripting", I hope I don't need to point
>    out the problem with this line....  (:
>
> 3) As the LSB notes, admins may run init scripts with strange values in
>    PATH and other shell variables.  You should set your own PATH if you
>    need it - which you do.
>
> 4) Stylistically, I don't like your mixed $VARIABLE vs. $variable
>    names.  And you should probably define a variable for
>    /var/lock/subsys/$prog, especially since that's the wrong path on
>    some systems anyway.
>
> 5) You don't provide a facility for running svnserve as a non-root
>    user.  I suspect most admins would prefer to specify a dedicated
>    user.  Use 'su' for this.
>   
Hi Peter,
Thanks for pointing out the problems. I will make the changes necessary 
for the Init script to be LSB and to make it portable.
I'll post it as soon as I complete it.

This was my first Init script. That's why it of such a bad quality. I'm 
sorry for that.

Cheers,
Mohanjith