[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: 2.3.1-14



At Wed, 26 Feb 2003 21:00:26 +1000,
Anthony Towns wrote:
> 
> On Wed, Feb 26, 2003 at 07:00:16PM +0900, GOTO Masanori wrote:
> > > Anthony Towns wrote:
> > > > Ugh. Surely the postinst should be changed to do something like:
> > > > 	if [ -x /usr/sbin/update-rc.d ]
> > > > 	then
> > > > 		update-rc.d "$@" >/dev/null
> > > > 		return $?
> > > > 	fi
> > "the former".  The latter is no meaning to put updatercd().
> 
> So why do you have updatercd anyway? The only reason I can think of
> is that you might have to run libc6's postinst when update-rc.d isn't
> available/working, but I think bootstrapping tools should cope with that
> fine anyway.

	> dpkg -s sysvinit| grep Pre-Depends
	Pre-Depends: libc6 (>= 2.2.4-4)

Sysvinit pre-depends libc6.  Libc6 should be installed and _correctly_
_configured_ before sysvinit is unpacked.  Your "bootstrapping tools
should cope with" really confirms this rule?
This is the reason which I think updatercd() is needed.

But, existing /usr/sbin/update-rc.d means at least either sysvinit or
file-rc is installed.  So, following your suggestion "the latter" is
something good way.  In this case, updatercd is for only sysvinit
Pre-Depends requirement.

However, updatercd() is already existed, and it has same function as
update-rc.d, so I decide to choose "the former".

> If it's just meant to be "simpler & faster", it ought to be dropped
> since it's less reliable.

Would you teach me why you think "it's less reliable"?

Regards,
-- gotom



Reply to: