Proposal to revise forest graph and integration practices for JDK 9

[Joe Darcy]

Hi Alan,

On 11/24/2013 4:17 AM, Alan Bateman wrote:
> On 23/11/2013 16:34, Joe Darcy wrote:
>> Hello,
>>
>> The current arrangements of sets of integration forests for a JDK 
>> platform release, like JDK 8, impose high overheads on development. 
>> I'm proposing we use an alternate forest arrangement for JDK 9 that 
>> will dramatically reduce the propagation time of fixes across the set 
>> of forests. More details below.
>>
> Good to see this, the lag getting fixes and changes involving hostpot 
> + another repository into specific integration forests has been a 
> bugbear in the system for a long time.
>
> It would really nice to get the new structure into a diagram along 
> with a summary of the testing that happens at each node or directed 
> edge. At least for the short term, then does the weekend 
> pre-integration testing on hotspot still happen before the changes 
> push to dev?

All else being the same, I think the level of testing that should occur 
before a HotSpot integration into dev should be at least equivalent to 
what now occurs for a HotSpot integration into master.

>
> I haven't see any comments yet from folks that currently push to 
> jdk8/awt and jdk8/2d but one thing that is an unknown (to me at least) 
> is whether there is any non-automatic testing that needs to happen 
> before the changes go into master. If there is then I just wonder what 
> this means for a weekly or bi-weekly integration from dev to master. 
> Maybe it's no different to today, at least in the short term.

A detail I omitted from the earlier description, the client libraries 
(awt, 2d, swing, etc.) and core libraries are integrated into master 
simultaneous even though they come from separate forests now. (The 
integrator creates a side forest with the combined set of client and 
core fixes. This goes through build + test before getting pushed to master.)

One side-effect of the proposal is that the client libs and core libs 
changes will get mingled a bit sooner than currently.

>
> Probably a separate discussion but one thing that is not clear to many 
> of us is the relationship between the hsx and jdk8 projects (some 
> people have different roles in one vs. the other). Are hsx roles 
> applicable in the JDK 9 project and the proposed structure? I'm just 
> thinking of someone pushing to hotspot + jdk at the same time and 
> whether they need to wear more than one shirt.
>

That is a relevant point to raise. I think it would be a fine 
simplification if those who have a certain status in the hsx project 
were initialized to have the same status in the jdk9 project, similar to 
what is done for the jdk8 -> jdk9 transition.

Thanks,

-Joe