At the same time, our blocker criteria are generally interpreted to
apply to
common bugs. If user switching is broken for one guy with an obscure setup,
that shouldn't be a blocker.
Yes, that's how release criteria usually work. It needs to be a general problem
affecting everybody or a large part of the audience (best guesses applied here). We
usually don't block on obscure setups, unless it has e.g. catastrophic consequences
(severe data loss, etc).
Is there a particularly strong use case for user switching when
booted
from live media? If not, can't this be fixed with an update?
Release criteria affect more things than just those which are important for LiveCD usage
and/or can't be updated. For example, keyring functionality is not too much useful on
LiveCD, it can be updated post-install, yet it has to work on the release day. The same
goes for lot of apps which you're not likely to use from LiveCD, e.g. gnome-documents,
gnome-contacts, etc.
But it's true that software which can be updated post-release gets lower severity when
discussing the impact of the bugs (especially when it's not a general issue affecting
everybody, but just a part of Fedora user base). The same would apply here, the point is
not to block on obscure bugs but to block on widespread bugs.
I agree with Matthias, user switching isn't worth considering a
blocker.
Besides, even if broken, it can easily be fixed by subsequent updates.
I proposed this criterion because history shows that the approach you mention (let's
fix it with updates) didn't really work, at least for GNOME. So I'm trying
something else, to make sure it gets the attention I believe it deserves. (We reached the
same conclusion during our blocker bug review). It seems that people opinions on
importance of this feature vary wildly, which is unfortunate. If we don't reach
consensus, it can't be a criterion.
A thing to consider - if we know user switching is completely broken (not just broken in
some use cases or for someone, but completely broken), are we OK with releasing Fedora
like that? Should that feature deserve a prominent position in the system/user menu, if
we're not willing to guarantee at least basic correctness?
In that case we would probably need to specify an explicit exemption for user switching in
https://fedoraproject.org/wiki/Fedora_22_Final_Release_Criteria#Default_p...
. Please note the description box below the criterion where it says:
# The key question is "would this bug cause significant inconvenience or just a
really bad first impression to a typical user or reviewer of the release?"