Proposal to revise forest graph and integration practices for JDK 9

[Henry Jen]

On 11/25/2013 05:40 AM, Dalibor Topic wrote:
> On 11/23/13 5:34 PM, Joe Darcy wrote:
>
> If long term features require a throw-away integration forest,
> create one as you need it for as long as you need it, integrate and
> close that forest.
>

I think this nails it, that's what the ideal case in my simple
mind. Whatever forest we have now is just a recursive version of that.

In a sense, hotspot is such a repo only that it lasts longer as
continious evolving of the codebase. The deciding factors makes us think 
we need a hotspot repo is probably why we end up with a thicket.

 From Joe's proposal, I assume the time-delayed integration from dev to 
master is not just hg pull, but a staging work like what we did for 
lambda to TL; otherwise, that seems sufficient to have just use tag.

As clone is cheap in mercurial/git, it's really that if we can have an 
infrastructure easily to support clone, we should really have many small 
feature-oriented repos and close it once done as Dalibor suggested.

Cheers,
Henry