On Fri, 2010-10-01 at 12:36 +0200, Kevin Kofler wrote:
Adam Williamson wrote:
> Again, you're extrapolating way too far from a single problem case. The
> problem is simply that we have the xorg-x11-drivers metapackage which
> requires every single X driver and is in the critpath. There's various
> ways we could adjust this so it's no longer the case. It's hardly
> something that renders an entire policy invalid.
Another example for how the critical path policy breaks things:
https://admin.fedoraproject.org/updates/firstboot-1.113-4.fc14
This update adds support for xfwm4 and openbox to the firstboot code.
Updates for those 2 window managers:
https://admin.fedoraproject.org/updates/xfwm4-4.6.2-2.fc14
https://admin.fedoraproject.org/updates/openbox-3.4.11.2-4.fc14
which add the virtual firstboot(windowmanager) Provides have already been
pushed to stable! So now we have 2 WMs satisfying firstboot's dependencies,
but not actually supported by its code. The result: the Xfce and LXDE spins
will be outright BROKEN. (And it's not my fault, I only did the firstboot
build and update requests, the other 2 packages were pushed by cwickert.)
I CANNOT push the firstboot update which UNBREAKS those 2 spins because of
the update policy. So instead of preventing breakage, the policy CAUSES
breakage! How can it fail more spectacularly for you to finally realize it's
a failure?
To all proventesters: please +1 that update, EVEN IF YOU HAVEN'T TESTED IT,
we need to get out of this impasse!
This is a bad idea. I don't advocate supplying positive karma feedback
without following basic test procedures to verify the update is sane. I
understand that sometimes things are out of ones' control, but lowering
quality standards to resolve this issue isn't a precedent I support.
In retrospect, if the three updates you list were in fact
interdependent, should they have been submitted and tested as a group to
avoid the current situation?
Thanks,
James