Re: Proposed release goal for lenny: Switch to dependency based boot system
> I have to say that I disagree with that. I have tried insserv on my
> laptop last month and was not all that happy with the way it worked.
> My objections are summarized in:
I am aware of this, but believe you do not have to worry as much as
you do. As far as I can see all of your problems originated from the
lvm2 script missing headers, and my bad decision to drop override
files for packages with headers in unstable, while forgetting to make
sure those headers were also making it to testing before the new
version of insserv.
When scripts have correct headers, the boot sequence stays correct,
and if there are scripts with incorrect headers, installing them will
either be rejected (because of loops), or that packages script will
get an incorrect location in the boot and shutdown sequence
(obviously, as the dependency information is used to order the
In your case, the lvm2 script got the default headers (depend on
$remote_fs and $syslog), which is wrong for lvm2. It got it because
insserv was upgraded and dropped the correct override file for lvm2.
It will not happen again (because I will not remove override files
before scripts are fixed in testing), and did not break the boot
sequence as it refused to mess with the working sequence as soon as
the resulting loop was detected.
Since your report, I have checked the packages you noticed with large
jumps in the shutdown sequence, and reported those with bugs in their
header, or missing header, to BTS. Your feedback was very valuable,
and I hope for more test reports to weed out the remaining bugs in the
dependency information provided by init.d scripts.