On Tue, 15 Aug 2023, 13:15 Santiago Vila, <[email protected]> wrote:
> > On a Debian 11 system where logcheck is installed, removing > /etc/logcheck/header.txt and then upgrading to Debian 12 > makes such file to reappear again. > > > thanks - i agree this is a bug and a patch to fix it would be great :) In the meantime anyone who doesnt want the header can either delete it again (which should be preserved by dpkg in future) or edit logcheck.conf to disable the inclusion if the header. Previously, header.txt was "a configuration file which is not a conffile > in the dpkg sense", so it was created at the first install to never touch > it again. This old behaviour was in fact the result of this suggestion > I made more than 20 years ago: > > https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=191891 > > and it worked well enough for a lot of time. > sorry to argue, but i disagree that it worked well enough. The contents of the header.txt have been improved several times and yet users who had installed were not given the benefit of those improvements. That is not the experience i expect from debian unless there is some compensating benefit. i know when i first installed logcheck many years ago i found the wording a bit clunky - and was surprised that when i moved to a new installation it magically changed. i always assumed this was a bug in dpkg conffile handling, so this had, for at least one user, a cost to debian's reputation for quality as well. keeping the file as a configuration file which is not a conffile > (i.e. handled by postinst) would have avoided such bug as well. > Only at the cost of maintaining technical debt and complexity in the packaging that - as this bug shows - no-one active understands. (i still dont understand why a header lives in debian/ and wasnt installed like a normal file - if it had been done as a normal conffile everyone would have been better off. Had the 20 year old bug included some comments as to why it was done like this, we could have done something better for bookworm) (Note: I'm not using any particular severity, but I believe this to be > a regression of the type that deserves a fix in stable). > > Thanks. > (i agree with this last bit - although i am not personally able to make that happen. ) >
