canonical-ci-engineering team mailing list archive

[Andy Doan <andy.doan@xxxxxxxxxxxxx>]

On 12/11/2013 05:54 PM, Chris Johnston wrote:
> I seem to remember that when we were in the beginning stages of
> discussing/planning the new CI Engine that we had discussed having each
> 'micro service' as its own branch or potentially its own project. In my
> opinion, each service should be in a separate branch.  Keeping
> everything in the same branch is going to make the bzr log absolutely
> horrible. If I'm looking for the last commit to a certain part of the
> engine, if the commit message doesn't explicitly state that it did
> something to that part of the engine, I may miss it entirely when
> searching. With each service being it's own branch, every commit will be
> related to a single service, and things will be much easier to follow.

There are reasons for disliking this, but this is not one. If you want 
to see all the changes for the ticket system just run:

 bzr log ticket_system

> I actually think that each service should be in its own project as well,
> but I'll settle for atleast separating things into their own branches.

That thinking leads to:

 https://launchpad.net/lava

which was a giant pain to work on every time you need to make a core 
change. We also did a similar thing with touch test cases during Saucy 
and after significant pain, you wound up being the person who 
consolidated our 3 touch testing branches into one.

> By separating each service into it's own project, it would be much
> easier for each of us to follow bugs and MPs related specifically to the
> code we are working on.

I think that's a bit of silo thinking. eg, We all need to care about 
what's going into the image-builder and at least be cognizant of the 
fact that there are changes going into it.