I was going to wait to comment later in
the day ... to see if servers recover fairly well by the end of today ...
but ... given Laurent's note, and others questions I'm getting via
IM ... thought I should clarify some now.
First, we should strive to never have
unsigned bundles as part of any of our repositories at any time ... that
is, at least, any repository that's "declared" for anyone to
install, even during development. If you do, then a) there is still some
security risk (the reason we sign in the first place), and b) then the
next time you provide an update, you need to make sure all your qualifiers
increase ... or, your newly signed bundles will not be picked up by some
users or adopters that rely on the version changing to signify a change
in content (which, is the reason for all the versioning rules and 4 part
version number requirement in the first place) ... and "signed or
not signed" is a change in content. So, usually best to "go with
old stuff" until you can sign again.
In addition ... what everyone is waiting
to hear about ... at this point, I still think our RC3 end dates are achievable.
Obviously many will be "late" (past their +N day) and I might
have to stay up later than usual on Wednesday night :) but I do think it
is possible to "make the date" and following our "heck or
high water" process, that is our current plan. Of course, some projects
may decide just to go with their RC2 results as their contribution towards
Indigo RC3, and roll-forward their RC3 contribution into their RC4 contribution
next week, but that's up to each project. At this point, I'd like to follow
the usual rules: if you contribute past your +N day, just let everyone
know here on this list after you have contributed; and, by Wednesday afternoon,
if your contribution is/will be past "EOD" (5 PM Eastern) to
let us know, here on this list, that you'd like a respin to pick up your
"late" contribution. We will all be plenty sympathetic. If you
just stay with your RC2 content for RC3 I don't think you need to notify
everyone here, but that's up to you ... in some cases (such as if you are
low in the food chain) it might be better to do so.
Naturally, we'll adapt and adjust plans
as we observe more about how the servers, etc., recover, but my guess is
we will recover enough to have a useful RC3. This will (or, has) pushed
out more risk to RC4 than we'd like ... more than usual ... but, that's
how we do it. And ... rest assured ... we've done it before ... that is,
we've had catastrophic failures in previous years during "the end
game" and still shipped on time. In fact, I think it was just two
years ago, we had "complete disk failure" and even had to wait
for a replacement to arrive!?
But, I can't help close on a whimsical
observation ... we committers first sucked up all the bandwidth ... that
problem was solved ... then we sucked up all the disk space ... that problem
was solved ... and now we've moved on to consuming power supplies! We sure
keep our webmasters busy! :) Much thanks for them to working tirelessly
to help us recover and keep on track.
And, thanks to all you committers and
"release engineers" for working through these difficult, last
minute problems. I know its stressful.
As always, keep the questions coming,
let us know how its going ... and I'll do the same.
Laurent Goubet <laurent.goubet@xxxxxxx>
Cross project issues
05/31/2011 09:55 AM
server status and the RC3 build
With the current issues on the server, what is expected from the
different teams as regards to the RC3 builds? Should we build our bits
"as we can" (disabling signing is part of it) and promote an
"RC3" build to the Indigo aggregator? Should we skip the RC3
hope that we'll be able to build in time for next week's RC4?
When I try to build EMF Compare, I simply get a cryptic
Caused by: java.lang.LinkageError: loader (instance of hudson/remoting/RemoteClassLoader):
attempted duplicate class definition for name: "hudson/model/Result