On Wed, 2009-12-02 at 14:37 -0500, Peter Jones wrote:
On 12/01/2009 11:16 AM, John Poelstra wrote:
> Jesse Keating said the following on 11/29/2009 10:52 PM Pacific Time:
>> On Sun, 2009-11-29 at 15:30 -0800, Chuck Forsberg WA7KGX N2469R wrote:
>>> Is it time for the images directory to reappear?
>>> In the past many problems with updated rawhides
>>> have been solved by doing a fresh install from
>>> boot.iso or pxeboot.
>> From now on, Rawhide will not have install images. Rawhide is a never
>> stopping never freezing repository of packages. To get to rawhide
>> you'll need to start with say F12 and either point to the rawhide repo
>> during install, or yum update to it post install.
> The part about not making install images all the time is listed as a
> "discussion point", part of the official proposal:
> 'Do we always make install images for rawhide, or only make images for
> pending release tree?'
> Big picture it seems like we've suddenly taken away access to a huge
> amount of potential nightly and periodic installer testing. Is this
> addressed somewhere in the proposal that I missed?
I'm not sure how you figure we've taken away access - we've made it so
an individual tester, working alone, has to do a compose themself, which
is admittedly a higher barrier to entry. But the side effect is that
people aren't testing trees we don't expect to work, and we've done that
without adding a serious communication overhead ("hey guys, will the
tree work today?")
It's an interesting difference between the two styles here.
Having testing tightly coupled with development. Versus, more
formalized boundaries between devel start/stop ... qa start/stop.
Different projects/teams decide on different approaches, so there isn't
a right answer here. However, at the pace at which we move in
Fedora ... a tight coupling between development and testing seems to fit
best. The alternative is QA bombards development with defects well
after they have context switched out of that code.
> Are there particular reason for not creating the images any more
> might help everyone understand more why this is a good idea?
Usually it's a waste of time and energy, especially early in the rawhide
cycle where we (the the anaconda team) are doing heavy development.
During this time ... should new packages be landing that are expected
not to work? There's been talk for a while that installer development
and testing would happen off to the side ... and then be tagged
appropriately so that it lands in rawhide after it passes some level of
verification. I'm certainly open to helping that succeed.
Is that the problem we are attempting to address by removing rawhide
> Can you add something to the proposal explaining when, how
> where install images will be created?
Maybe we ought to have sortof a fire-drill test day sometime soon where
we spin an anaconda we think will work and do some smoke-test composes and
installs, and then do a compose for the test day just to get in the habit
of the new process.
Using the current schedule , we have the following milestones where
QA will need install images for testing.
= Alpha =
2010-02-04 - Pre-compose install testing 
2010-02-11 - Test Alpha 'Test Compose' (boot media testing)
2010-02-18 - Test Alpha Candidate
= Beta =
2010-03-11 - Pre-compose install testing 
2010-03-18 - Test Beta 'Test Compose' (boot media testing)
2010-03-25 - Test Beta Candidate
= Final =
2010-04-15 - Pre-compose install testing 
2010-04-22 - Test RC 'Test Compose' (boot media testing)
2010-04-29 - Test RC Candidate
This model assumes that rawhide install images were available for anyone
to test at anytime. As Adam points out elsewhere in the thread, where
this helps us is so people can verify bug fixes. We also benefit from
community testing against rawhide leading up to these milestones.
If I'm summarizing the pain correctly ... the push back seems to come
from providing testing before development is ready to accept bug
Does this only happen during early development? Meaning, after
Alpha/Beta is having nightly built rawhide install images still a
problem? When in the schedule  would be an acceptable time to start
providing test feedback?
Another point, I think we can't limit the discussion to just
anaconda-devel. There are a *lot* of other critical components pulled
into the install images that are developed outside that group.
 Not yet present in the scheduled, but mirrors F-12