sorry for the delayed answer but I'm on vacation.
In summary I have the same concerns as Jeff and
Derek and I would like to discuss this further before
making a decision. Because, SR1 is very close, I
suggest we don't change anything just yet and release
8.5 from master with SR1, I don't think there is a big
drawback to that in itself. And we continue this
discussion for 8.6 (which we can still decide to plan
for December, although that is not my vote). I try to
express my hesitations in detail below.
First, as Jeff pointed out, doing an 8.4.1 release
off the 8.4 branch, instead of 8.5 from master implies
more work this time because we were not expecting
that. I was under the impression that the 8.4 branch
was used by integrators that needed specific fixes for
an older release; with that in mind, I didn't focus on
duplicating bug fixes from the master branch to the
8.4 branch. So, like Jeff, and possibly others, I
would need to go back and sift through the commits and
cherry-pick the right ones. To do things right, I
will need to test each bug fix. That's 14 bug fixes
in the Debug component. I'm not very fond of
retesting each one.
Second, some consumers of CDT repackage and release
to their customers. This is our case. We include
other projects in those releases, not just CDT, and
those projects follow the Eclipse release train. This means that we probably
cannot get away from having our internal releases
match with SR1 and SR2. So a release in December
will not provide
much benefit for us, but will create more work (see
The internal releases
I mention have to be tested. We run thorough tests
on each one (by 'we' I mean the Ericsson
committers). Any issue found is fixed in
open-source, so the community benefits directly from
our efforts. As our packages include other projects
that interact with CDT, we cannot assume everything
will just work because CDT has a small SR release.
So we will need to keep on testing each release.
Now, we are discussing adding a forth release each
year. That's another test cycle. That's more work.
I have other concerns (e.g., will users know to go fetch an
un-aligned December release, compared to getting
the release automatically with the SRs?), but let's focus on the ones
We should focus on the
benefits to the community, and my points might seem
focused on Ericsson. In truth that is not the case.
I still don't see the extra benefit to the
community in having a December release compared to
one at SR1 and one at SR2. Both allow the community
to get features faster. If the SR1 cycle is too
short, then the SR2 one comes shortly after; how is
that different than a December release?
I do see a
disadvantage to the community in the fact that if
the committers are tied-up doing release efforts, we
spend less time fixing bugs and working on features.
So I believe it is a direct benefit to the
community to minimize the overhead of the
The only question mark
for me is the release review. I am not familiar
with it and don't understand how much work that
entails. But besides that, I believe that not being
aligned with the release train will slow us down
with a benefit I have yet to understand.
But I'm enjoying the
I'd iike to hear from the Ericsson gang before
going much further.
To answer some of the questions below,
IMHO, this is actually less releases than what
we were planning. The SR's are now just
maintenance releases which are a lot less work
(no release review required, for example). We
go from 3 minor releases a year to 2 which
allows us to make each one more feature rich
and gives us enough time to test them.
And, yes, sorry for the late call with
8.4.1. I was always worried about the 4 month
cycle (which is actually only 3 with SR-1) and
should have held my ground there. It will mean
more work for folks like Jeff who want to get
things in the SR.
If we decide this is the way to go, we'll
need to firm up the 8.5 schedule. I don't mind
doing it early in December, which is really
only 2 months after SR-1, so if there are
things that can wait that long, you wouldn't
need to post it to the SR.
The more I think about having a holiday
season release, the more I like it. Especially
with the personal work I'm doing with Wascana
supporting hobbyists on Windows and Arduino
(and maybe even Raspberry Pi). It would be
nice to get this all into their hands by then.
I'm OK with either proposed
of the current release schedule. We are
releasing approximately twice a month
from nightly builds, so it doesn't make
a difference for us.