Should a working fedup in Fedora N's stable repository be a release criterion for N+1?

[Adam Williamson]

On Wed, 2013-12-18 at 12:31 -0800, Andrew Lutomirski wrote:

> I wonder if fedup (the program) should bail, or at least warn, if the
> kernel/initramfs version doesn't match.  If I had gotten a message
> that said something like "You are trying to use fedup 0.7.0 to upgrade
> to a release that was composed with fedup-dracut 0.8.0.  This may be
> unreliable.", then I might have grumbled, but I wouldn't have spent a
> bunch of time wondering why the upgrade silently failed.
> 
> I certainly understand that this stuff is *hard*, especially when you
> factor in the combinatorial blowup of having multiple version to
> upgrade from and to at the same time.

I don't think that's a practical approach, at least at this point in
time. We really haven't nailed down the expectations for what should
work with what. fedup and fedup-dracut versions are not necessarily in
lockstep for the duration, I don't think.

Do bear in mind all this stuff  is _new_ and we made up the procedures
for how everything gets built and tested on the fly, over just the last
three release cycles. During f18, I remember, there was a large amount
of confusion between wwoods and releng as to how the initramfs was
supposed to get built and delivered to users at all, and it was quite a
struggle for QA to follow and figure out what and how we ought to be
testing.

Now the process has settled down at least a bit and we know where the
kinks are, we can clean up the test procedure and documentation
somewhat, and it should be possible to avoid this happening in future.
I've already been working on this in the last dozen hours or so, in
between firing off 'use 0.8!' notices everywhere I could think of.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net