[Pkg-sysvinit-devel] Bug#806438: Bug#806438: update-rc.d: Do not fail when initscripts is not installed

[Felipe Sateler]

On 27 November 2015 at 19:59, Petter Reinholdtsen <pere at hungry.com> wrote:
> [Felipe Sateler]
>> Do we want to break installs when an init script is broken but the
>> user is using systemd and not sysvinit? I think this is the question
>> we should be asking. The patch I proposed essentially answers "no".
>
> As long as some of the archtectures in Debian are using sysvinit, I
> suspect the best answer to that question is "yes", to ensure as many
> eyeballs as possible can discover, report and fix such problems early.

I disagree that "fail to install" is an appropriate action when there
is no damage done to the system in question.

>> That we check for mountkernfs.sh, which lives in package initscripts,
>> I consider only a proxy for the real question: is this system likely
>> to ever boot into sysvinit?
>
> To me that is not really the question.  To me it is "is the package
> broken, and how should it be discovered".  As long as systemd only work
> on Linux, and Debian provide kFreeBSD and Hurd kernels, we should do our
> best to fix bugs in the init.d scripts.  After all, incorrect init.d
> script dependencies are trivial to fix, but hard to find.

Well, its hard to disagree with fixing broken packages ;).

However, failure to install is too great a cost to pay for detecting
such bugs, especially when the bug will not affect the current system.
We do not cause install failures when packages have RC bugs, or fail
to build on other architectures. We don't even fail to install on file
conflicts when the conflicting package is not installed. I think we
should stick to that logic here: if we know that sysv init is never
going to be booted (because a critical file is missing), then don't
bother with ordering problems, because they never will be problems for
the current system.

>
>> I'm afraid I don't see any other way than triggering error at the
>> point where we detect this might be a problem: when initscripts is to
>> be installed again.
>
> That is one scenario, but not the one I had in mind, as I mention above.
>
> I would like to ensure the package maintainer and all the package users
> detect incorrect boot script dependencies as early as possible,
> preferably before the package is uploaded or at least as soon as it is
> unloaded to unstable.

Maybe insserv should be changed to emit a warning even when -f is
passed. Currently it succeeds silently when -f is passed.


-- 

Saludos,
Felipe Sateler