dealing with RH initscripts backward compatibility

[Pekka Savola]

On Sun, 18 Feb 2001, Jim Knoble wrote:

> Circa 2001-Feb-17 22:34:26 +0200 dixit Pekka Savola:
>
> : Hello all,
> :
> : Continuing the thread:
> :
> : Re: PATCH: make contrib/redhat/sshd.init work with older RH releases
> :
> : Attached are newer versions of initscripts.  These are smaller and
> : probably more readable than patches.  Backward compability features
> : haven't been tested that extensively.
> :
> : I think the issue of legacy initscripts support should be handled like
> : with these patches (sshd-functions could be refined, of course), or in
> : addition:
> :
> : * in openssh.spec, there would be a %define to enable "backward
> : compability".  There might even be autodetection for this using
> : /etc/redhat-release.
> :
> : * with this defined, sshd-functions would be taken from contrib and
> : installed in /etc/rc.d/init.d/.
> :
> : * this would give the implementor of sshd-functions more liberty at how he
> : could redefine echo/failure/success/action/etc., because he would know
> : that the changes would only kick in for users using RHL5.2 or earlier
> : [currently]
> :
> : With this, there might be no need for the "do we require this" -checks
> : (~30 first lines of sshd-functions).
> :
> : What do you think?  IMO, I think the new idea is probably better because
> : it allows for more freedom when it comes to the implementation.  Also,
> : there are other issues that will be version-specific (pam, ...).

I'd like comments about this approach.

> : # source initscripts backward compatibility functions if they exist
> : if [ -r /etc/rc.d/init.d/sshd-functions ]; then
> :   . /etc/rc.d/init.d/sshd-functions
>
> I don't like this at all.  If this file goes away due to filesystem or
> operator error, things will unexplainably break.  Having them break
> with an immediate message such as:
>
>   bash: /etc/rc.d/init.d/sshd-functions: No such file or directory
>
> is much easier to diagnose.

The new idea was that sshd-functions would not need to be present on
recent systems.  With this approach, you could cut out the additional
checks (below).

>
> : fi
> :
> : RETVAL=0
> : prog="sshd"
> :
> : # Some functions to make the below more readable
>    /^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>   /
> You probably mean to move this line down to where the functions are.

Sorry, typo.  I meant definitions. :-)

> --------[file: sshd-functions]--------
>
> : # Backward compability functions for initscripts, parts by Red Hat.
> :
> : # Find out whether we need to use the local functions
> : # Unnecessary use should be avoided.
> :
> : if [ ! "`type -type success`" = "function" ]; then
> :   success() {
> :     my_success "$*"
>                /^^^^^^
> 	      /
> No, no no.  Use "$@", not "$*".  "$*" turns multiple arguments into one
> argument---this is almost never what you want, and certainly isn't here.

Ok.

> : # Required for old initscripts < 4.16 or so (RHL5.2)
> : my_success() {
> :   local msg
> :   if [ $# -gt 1 ]; then
> :     msg="$2"
> :   else
> :     msg="done"
> :   fi
> :   case "`type -type success`" in
>
> If you're going to go about things this way, why do you check whether
> 'success' is a function twice?  Do it above, or here, but not both.

Well, I'm not sure which approach if preferable:
 1) include sshd-functions for everyone
 2) include sshd-functions for old os versions only

I think changes to newer systems to newer systems should be weeded off as
soon as possible, hence the checks in 30 first lines.  Which to remove
depends on the approach.

> : # Required for old initscripts < 4.16 or so (RHL5.2)
> : my_action() {
> :   STRING=$1
> :   echo -n "$STRING "
>    /^^^^^^^^^^^^^^^^^^^^
>   /
> my_success() and my_failure() already echo their string .  Why do you
> do it again here?

True.

> :    1.*)
> :       echo ${args} "$@"
>       /^^^^^^
>      /
> This won't work at all they way you've done things.  echo() is defined
> as my_echo(), which calls echo(), which calls ... ad finitum memoriae.

Changing to /bin/echo would fix this, not sure if that'd take advantage of
$".." locales though.

-- 
Pekka Savola                  "Tell me of difficulties surmounted,
Netcore Oy                    not those you stumble over and fall"
Systems. Networks. Security.   -- Robert Jordan: A Crown of Swords