do-release-upgrade should block release upgrades in some circumstances
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
ubuntu-release-upgrader (Ubuntu) |
Fix Released
|
Undecided
|
Brian Murray | ||
Xenial |
Fix Released
|
Undecided
|
Brian Murray | ||
Bionic |
Fix Released
|
Undecided
|
Unassigned | ||
Cosmic |
Fix Released
|
Undecided
|
Brian Murray |
Bug Description
Impact
------
do-release-upgrade will allow people to upgrade to a new release of Ubuntu even if they have outstanding updates to install or need to reboot from a libc6 upgrade. This behavior is different than that of update-manager which forces you to install all updates or reboot before upgrading. Additionally, this specifically allows users to upgrade without receiving fixes that would improve the upgrade process e.g. bug 1780996.
Test Case for missing updates
-------
1) Install Ubuntu 18.04 and don't install updates.
2) Modify /etc/update-
3) Run 'do-release-upgrade -d'
With the current version of ubuntu-
Test Case for libc6 update
-------
1) Install Ubuntu 18.04
2) Modify /etc/update-
3) Run 'sudo apt-get install --reinstall libc6'
4) Run 'do-release-upgrade -d'
With the current version of ubuntu-
Regression Potential
-------
The strings are not translated but that seems better than people having a bad outcome when upgrading. It's also worth noting that
/usr/lib/
Related branches
Changed in ubuntu-release-upgrader (Ubuntu Cosmic): | |
status: | New → In Progress |
assignee: | nobody → Brian Murray (brian-murray) |
description: | updated |
tags: | added: id-5b8465b2c124d91ccc4a6766 |
Changed in ubuntu-release-upgrader (Ubuntu Xenial): | |
status: | New → In Progress |
assignee: | nobody → Brian Murray (brian-murray) |
I saw this in the unapproved queue and have the following questions:
What's the reason for singling out the libc6 reboot-required reason rather than asking for *any* reboot-required?
If you do strongly want to keep it limited to libc6 only, wouldn't opening the file in the python code instead of opening a grep subprocess be faster?
I'm semi-convinced by your reasoning on untranslated strings, but if we could make some effort towards getting as many as possible I'd be happier, such as mailing the translator list so it gets in the next langpack refresh in bionic. (And it seems to me like if the string could avoid mentioning "libc6" that would be more friendly to any future tweaks in this area.)
Is it right to say that this could equally be a 0-day SRU and so this upload isn't important for the release itself - there's no need to rush it in for final freeze?
I'll reject from the queue for now, but someone else could find it in the rejected queue and accept if that is the decision.