Re: potential mass bug filing: sysvinit dependency
On Tue, 05 Sep 2006, martin f krafft wrote:
> also sprach Henrique de Moraes Holschuh <hmh@debian.org> [2006.09.05.1526 +0200]:
> > This one needs to depend on sysv-like link farm functionality (as
> > opposed to, say, file-rc style). If upstart provides symlinks,
> > then we need a virtual package. If it doesn't, sysv-rc-conf needs
> > to keep depending on sysvinit.
>
> Well, upstart does get rid of SysV as we know it. The upstart
> package itself gets rid of the need for invoke-rc.d, providing
> /sbin/start and /sbin/stop.
invoke-rc.d is a maintainer script compatibility layer to interface to the
initscript subsystem (that happens to guarantee some functionality that
some initscript subsystems don't provide natively while at it).
If that layer doesn't allow for everything we need, then it is time to
revise it, of course. Just like it is well past time to fix the update-rc.d
crap.
I still think we should have renamed these utilities to "initscript-invoke"
and "initscript-register" while implementing at the very least as much
namespace control as we have in dpkg-divert for initscript-register (and a
command line interface that could be better extended for out-of-band
dependency information too).
> However, upstart-compat-sysv provides compatibility.
Right answer. Just make sure to always have that package installed. Even
if we eventually switch to upstart as the default initscript subsystem, we
will still need to support alternate initscript subsystems, and that means
invoke-rc.d or something else that does its job for *all* supported
initscript subsystems.
Ah, there is also all required telinit functionality, too (we should
identify this subset and write it down as policy, as a matter of fact).
--
"One disk to rule them all, One disk to find them. One disk to bring
them all and in the darkness grind them. In the Land of Redmond
where the shadows lie." -- The Silicon Valley Tarot
Henrique Holschuh
Reply to: