RE: [eclipse-pmc] Helios end game plan

[John Arthorne <John_Arthorne@xxxxxxxxxx>]

> Do
you have data at hand what had caused the delays?

From what I can see in the mail archives:

RC1: two teams (p2 and releng) asked
for a rebuild on Friday. The build and tests completed Friday evening,
and was promoted Saturday morning.
RC2: Friday morning rebuild for a Workspace
bug, and another on late Friday afternoon for a p2/releng problem.
RC3: no rebuilds, promoted on time
RC4: There was a rebuild for a PDE doc
issue, and to accept a new version of ECF. The build finished at 11pm Friday
and all teams didn't give a "go" until Sunday morning.

I don't know if there is any general
lesson to learn from this. One issue is that if the build isn't completely
done (along with tests) by middle of Friday afternoon, then you end up
losing a day because all teams can't give a "go" promptly once
you get into Friday evening and the weekend (even Friday 5pm EDT is 11pm
in Europe). 

Another observation: RC3 had no rebuilds,
and it still took until middle of Friday afternoon for all teams to sign
off on the builds. This means there was no slack in that schedule at all
to have a rebuild and still ship on time. We have to account for the fact
that final verification and sign-off is an important step and we need to
schedule time for it. From the above information, my conclusion is that
this was more of a scheduling error than a failure by the component teams
- we just didn't put enough time into the schedule for any rebuilds, which
is a possibility I think we should always be prepared for (should leave
enough room in the schedule to ship on time with a single rebuild after
the scheduled build).

John

"Oberhuber, Martin" <Martin.Oberhuber@xxxxxxxxxxxxx>
Sent by: eclipse-pmc-bounces@xxxxxxxxxxx

03/15/2010 11:17 AM

Please respond to
eclipse-pmc@xxxxxxxxxxx

To	<eclipse-pmc@xxxxxxxxxxx>, "Thomas Watson" <tjwatson@xxxxxxxxxx>
cc
Subject	RE: [eclipse-pmc] Helios end game plan

> I
think this change is necessary because last year we proved incapable of
producing a good build in less than 24 hours.
 
This concerns me. 
 
Do you have data at hand what had caused
the delays?
 
Martin

---

From: eclipse-pmc-bounces@xxxxxxxxxxx
[mailto:eclipse-pmc-bounces@xxxxxxxxxxx] On Behalf Of John Arthorne
Sent: Monday, March 15, 2010 3:57 PM
To: eclipse-pmc@xxxxxxxxxxx; Thomas Watson
Subject: [eclipse-pmc] Helios end game plan


Fellow PMC members, 

According to the Helios simultaneous release requirements, we must make
our end game plan available by Helios M6, which is this coming Friday the
19th.  I have written an initial draft of this end game plan here:

http://www.eclipse.org/eclipse/development/plans/freeze_plan_3.6.php

Apart from the usual one day shift in dates, I have made two changes from
last year: 

1) Removed one day from the RC2 test pass, to make it a single day test
pass. The feedback during the Galileo retrospective was that the extra
testing at this point was too late in the cycle. Instead teams would like
some extra testing earlier in the end game. I have *not* added this extra
testing, since I thought we should discuss with the whole team during this
week's planning call. The two choices are to have a test pass immediately
after M7 for two days, or to start the M7 shutdown a couple of days earlier
to allow for an extended M7 test pass. I can update the the freeze plan
once we decide on that. 

2) Moved the RC builds from Thursday 8pm to Thursday 8am (EDT). This is
unfortunate because it cuts down on the time available for the fix pass.
Hopefully teams that *do* make a good contribution the first time can spend
all of Thursday and Friday doing either testing or fixing for the next
week's RC. I think this change is necessary because last year we proved
incapable of producing a good build in less than 24 hours. From the archives,
this is when we shipped each RC last year: 

3.5 RC1 Saturday 9am 
3.5 RC2 Sunday 7pm 
3.5 RC3 Friday 5pm 
3.5 RC4 Sunday 9pm 

Hopefully making the build 12 hours earlier will help us hit closer to
the target this time around. When we slip our milestone/RC dates it causes
great disruption to the +1 projects that need to have their builds completed
on the following Monday. 

Comments and criticism welcome on any aspect of the freeze plan, either
my two proposed changes or any other aspect of the plan that has not changed
from last year. We can discuss on this list, and/or during Wednesday's
call this week. Also let me know if there are any national holidays in
countries where we have committers that I missed. It's useful to have these
marked in the freeze plan for reference. 

John_______________________________________________
eclipse-pmc mailing list
eclipse-pmc@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipse-pmc