On Tue, 17 Apr 2012 10:57:03 +0200, Zygmunt Krynicki zygmunt.krynicki@linaro.org wrote:
W dniu 17.04.2012 03:05, Michael Hudson-Doyle pisze:
On Sat, 14 Apr 2012 20:14:44 +0200, Zygmunt Krynicki zygmunt.krynicki@canonical.com wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi
I'd like to announce a new project in the LAVA family.
(README file follows) http://xkcd.com/927/
Seriously, lava has way to many tiny pieces. It all started when we renamed launch-control to lava-dashboard and created lava-server in the process. Now it's getting in the way.
Hi. I approve of the intent, but am a bit confused as to the scope. What existing components do you see as ending up in lava-core?
I guess at least lava-server, lava-dashboard, lava-scheduler.
lava-dispatcher? Not so sure about that one.
Actually I only see the need of two packages that we work on:
- lava-core, it will have everything that we work on that is installed
on "server" machines: lava-{dispatcher,server,dashboard,scheduler,android-test} are going to be a part of that. The aggregate code size is tiny and the value of splitting the packages is arguably small in comparison to the fuss that is needed to keep them up to date separately.
As I said on IRC, I'd be wary of overly tight coupling here -- we don't want to have to have rabbit running to be able to install or run just the dispatcher. That sort of thing isn't an inevitable consequence of putting everything in the same branch, but it makes it easier to fall into that trap.
- lava-device-agent, it will have everything that we need to put on a
test device.
Uh, what about client side things like lava-dashboard-tool?
It _will_ share code with lava-core but the exact method is not set in stone yet. I need some time to experiment and see.
Well OK, but I hope we can still get useful work done while you're experimenting. I guess I'm worried that I don't have an idea of the scope of your plans in this area or the timeframe you are thinking of. I'm wary of temporary hacks that turn out not to be temporary.
Ideally we could keep it on one repository and just release separately. I really want one tree that it easy to manage, test, release and upgrade.
One repository, multiple releases would make some kinds of development easier for sure, but OTOH would not make releasing or upgrading easier AFAICS.
I see a branch proposing lava-tool for inclusion -- but lava-tool is used by the client side tools, and so seems to me to be a candidate for keeping somewhere else...
The copy of lava-tool in lava-core will be independent. Before the transition is complete this will simplify the process.
The reason I asked about lava-tool is that lava-dashboard-tool etc depend on it, so where do you see lava-*-tool going? Into lava-core as well? If not, then I really don't think we want to have two copies of the lava-tool code floating around.
Cheers, mwh