[systemd-devel] [PATCH] core: make parsing of chkconfig headers conditional

[Lennart Poettering]

On Mon, 24.03.14 17:17, Michael Biebl (mbiebl at gmail.com) wrote:

> 
> 2014-03-24 17:11 GMT+01:00 Lennart Poettering <lennart at poettering.net>:
> > On Mon, 24.03.14 14:00, Michael Biebl (mbiebl at gmail.com) wrote:
> >
> >> Don't parse Red Hat style chkconfig headers if chkconfig support is
> >> disabled via --disable-chkconfig.
> >
> > Hmm, I am not overly hapy with adding even more ifdefs to the
> > code... i'd rather see less ifdefs...
> 
> I can rip out the chkconfig parsing completely, if that is what you
> want.

Nah, we need to keep it in for fedora/rhel I figure, for now.

> 
> > Thus, what's the rationale here? Why shouldn't the chkconfig header be
> > read when it exists? I mean, stuff like the description string really
> > shouldn't hurt on debian either, should an init script carry the
> > header...
> 
> The reasoning here is, that on distributions which don't use chkconfig
> (like Debian), there might be SysV init scripts with chkconfig headers
> which are completely untested and actually break the boot process. See
> [0] for an example.
> We really shouldn't parse chkconfig headers on distros which don't
> support chkconfig and decided to build systemd with
> --disable-chkconfig.

Well, but we are a lot more dependent on correct headers than other
parsers anyway... For example, the LSB deps have traditionally been much
less accurately followed than they are now... 

I mean, we don't claim compatibility with scripts with broken headers,
anyway...

Lennart

-- 
Lennart Poettering, Red Hat