ubuntu: ucf tracking of valid known md5sums should be limited to only those md5sums that affect a given distro release

[Chad Smith]

Currently the project tracks all valid md5sums of permutations of 50unattended-upgrades.conf in a single md5sum file that contains every md5sum of every historic version of all unique distros:

 50unattended-upgrades.Debian
 50unattended-upgrades.Devuan
 50unattended-upgrades.Raspbian
 50unattended-upgrades.Ubuntu

Ultimately ucf for a given packaging release should only track the applicable md5sums which are expected to be seen on that particular distribution and release.

For example:
   On Ubuntu Bionic: valid md5sums should be limited to the md5sum of the most recent Ubuntu Xenial 50unattended-upgrades.conf and the md5sums of previous Ubuntu Bionic releases to allow Xenial->Bionic and Bionic->Bionic upgrades without prompt.

[Balint Reczey (rbalint)]

It is highly unlikely that the configuration file on one distro is replaced with one that was shipped on a different one. It may be a bit more likely that a config file is overwritten by a variant from a previous release, but I think this is still unlikely and I believe trimming the md5sum list is not a general practice for UCF managed configuration files.

As an example openssh-server ships the historical list, too:
$ cat /usr/share/openssh/sshd_config.md5sum
# Historical md5sums of the default /etc/ssh/sshd_config up to and including
# 1:7.3p1-5.
0d06fc337cee10609d4833dc88df740f
10dc68360f6658910a98a051273de22c
11f9e107b4d13bbcabe7f8e8da734371
16c827adcff44efaca05ec5eea6383d7
2eeff28468576c3f2e538314e177687b
386c8b9079625b78f6d624ae506958ae
38fc7b31b3e3078848f0eec457d3e050
395c5e13801f9b4f17c2cb54aa634fbd
...

[Robie Basak (racb)]

See also: bug 1915547

> It is highly unlikely that the configuration file on one distro is replaced with one that was shipped on a different one.

I think it's more likely than you say in cases that a configuration shipped is primarily a set of boolean values and enumerations of a limited set of strings, and the difference in distributions is mostly in choices of those values. The extreme example would be a configuration file that ships just one boolean setting. I think this unattended-upgrades case is closer to that extreme example than it is to openssh's sshd_config.

[Brian Murray (brian-murray)]

The Hirsute Hippo has reached End of Life, so this bug will not be fixed for that release.