harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Stepan Mishura" <stepan.mish...@gmail.com>
Subject Re: [general] M3 schedule
Date Fri, 14 Sep 2007 06:15:45 GMT
On 9/13/07, Tim Ellison wrote:
> Hi Stepan, thanks for bringing this up again.
>
> Stepan Mishura wrote:
> > OK, at least nobody suggested that we have longer then 3-month
> > schedule. That means that we should have next milestone no later then
> > September, 30.
>
> Agreed, let's shoot for the end of the month for M3.
>
> > So I'd like to remind that we are close to the mid of September and
> > propose the following M3 schedule:
> > - Septemer 15: feature freeze - svn trunk is frozen
> > I think this time it makes sense to freeze the trunk instead of
> > branching to avoid doing merges for bugfixes
>
> Yep, so no new areas of significant functionality should be introduced
> after Sept 15th; but we can continue to do bug/perf/tidy-up type changes
> during this period.
>

Sure.

> > - Septemer 22: code freeze - M3 branch(tag?) is created and the trunk
> > is unfrozen. All fixes for M3 go to the branch.
>
> So what are you proposing gets branched? everything (and only those
> things) that contributes to the federated build? e.g. drlvm, classlib,
> and jdktools.  I assume that the BTI code is not branched too?
>

Everything that is required for building snapshot (i.e drlvm,
classlib, and jdktools).
BTI is not branched.

> Can BTI and other scripts cope ok with switching to the code on a branch
> rather than dealing with HEAD?
>

Well, it seems that I hurried up a bit with the proposing to branch
the trunk for M3. I think now that testing ifra is not ready to
provide testing for the branch. I'd like to mention that we use many
machines (up to 30) to run a set of scenarios on 4 platforms. All
these CC instances will have to be reconfigured to test the branch.
Reconfiguration is not a problem itself - CC is stopped, svn urls are
updated to point to the branch and CC is started again ... but taking
into account the number of machines it won't be 2 minutes switch (and
after M3 we have to switch them back).
Also CC reconfiguration will mean that the trunk testing will be
stopped (IMO this is not good).

Having said that I'd propose to follow the same procedure as we did
for M2 - the trunk is frozen and no more commits without agreement
from two committers on the dev list during code freeze.

Sorry, for making a disturbance with the branch proposal.

Thanks,
Stepan.

> Once the M3 branch is created then I assume that the HEAD is opened once
> again for code and feature changes, but changes to the stable M3 branch
> require our established `review then commit` style.
>
> > - Septemer 29 :release date - M3-branch is merged with the trunk.
>
> Though there is no reason why we would not merge the M3 branch into HEAD
> during the M3-shut-down period either, right?  At the end of Sept we do
> the final merge, then delete the M3 branch.
>
> > Objections?
>
> No, I'm keen that we get another stable build published.
>
> Regards,
> Tim

Mime
View raw message