Is Rawhide supposed to be useful?
Jeff Spaleta
jspaleta at gmail.com
Tue May 10 16:26:51 UTC 2011
On Tue, May 10, 2011 at 7:34 AM, Jonathan Corbet <corbet-ft at lwn.net> wrote:
> Could it be that Fedora lacks the resources to maintain both Rawhide and
> the next-release branch? In retrospect, was No Frozen Rawhide as good an
> idea as it seemed?
I need to redo my tongue-n-cheek seasons of rawhide in the new No
Frozen Rawhide era to give you a clearer picture as to what to expect
from Rawhide. Rawhide still has seasons of activity, and while the
current rawhide winter of your discontent might not be _frozen_ its
not necessarily as active as other times. Nor do I think anyone
expected it to be equally active at all times.
The point of No Frozen Rawhide was two fold. To make it easier for
pre-release testers to transition into the gold release without
getting pinched by the confusion in the transition of rawhide to to
N+1 near release day. And secondly, it makes it _possible_ to
introduce breakage into rawhide earlier than release day so there
isn't a flood of pent up changes which get introduced in a pulse as
soon as rawhide unfreezes. And in that respect the policy is working
as expected. Pre-release day breakage achievement unlocked!
There's nothing in the policy which speaks to curb breakage which
lingers for more than a few days once introduced. Whether that
breakage is introduced pre-release or post-release. As before
maintainership is a best effort affair. There's been no best effort to
define a minimum best effort.
Yes its dumb that a _simple_ maintainer mistake sat there for weeks.
This sort of stuff happens. It's happened to my packages, where I
thought I applied some simple brown bag packaging fix on a branch(a
release branch even!) and I plum forgot to do it till someone poked me
in the eye. It's a _human_ mistake. Until we have skynet doing the
package maintainence for us, we have to have a maintainership system
which can recover when mistakes happen to mitigate the effect no
matter if its rawhide or not.
-jef
More information about the devel
mailing list