kernel 2.6.16-1.2107_FC5 breaks graphical boot (and cups and httpd)
wwoods at redhat.com
Fri May 5 14:12:43 UTC 2006
On May 5, 2006, at 4:28 AM, Rahul Sundaram wrote:
> On Fri, 2006-05-05 at 09:06 +0200, Arjan van de Ven wrote:
>> I'm sure Red Hat has some system for their own testing of RHEL, maybe
>> for such security fixes it can be used for Fedora?
> That is one of the things the new Fedora QA lead has to figure out.
Hey, that's me!
As Arjan says, Red Hat indeed has systems/procedures (automated and
otherwise) for testing packages before they go into RHEL. I'm hoping
that I can bring some of these things to Fedora to keep this sort of
thing from happening. But it takes a while to build this kind of
infrastructure. We'll start small.
>>> The classic solution would be a small farm of automated test
>>> but I imagine that's a bit too resource intensive for the project
>>> right now.
Actually, no! It's not a gigantic test lab, but we have some hardware
dedicated soley to Fedora QA and we're working on timesharing
agreements so we might be able to use the RHEL test systems when
they're not busy. This should allow us to do automated testing on
releases / packages before they go fully public.
>>> How about a core team of volunteers that yum updates off a
>>> private repo
>>> several hours before it goes public?
Volunteers? YES! Oh please yes! But not just for security updates
(although we will definitely work on that).
As for the simple case - testing packages before they go public -
Isn't that what updates-testing is for? The brave (and/or foolhardy)
can enable that repo and give us some advance warning if a package
has gone bad.
> There is a push towards that. Not sure whether automated testing would
> have detected this particular problem though.
Depends on what the automated tests do, really - on my machine, any
test that used TCP connections would have fail. We definitely have
automated tests here that do that sort of thing - the question is
whether the RHEL team is allowed to give us their tests. We'll see
what happens. We can always write our own tests if the freely-
available ones don't do enough.
More information about the test