Next week, I'll start implementation of UI tests for all screens.
Emmanuel
Jason van Zyl a écrit :
On 17 Oct 06, at 2:34 PM 17 Oct 06, Brett Porter wrote:
I agree with Emmanuel. IIRC, the profiles are already in the model,
and basic choice of which JDK and maven/ant installation to use should
be straightforward and extremely useful. I agree that making it more
pervasive and using the toolchain support would be even better, but I
don't believe it needs to wait for that.
I would be against any more radical changes until the testing setup is
rock solid. We're slipping in this area. We don't really know what
machines this stuff runs on and I don't think anything is automated
anymore. We need to stop paying lip service to what we are preaching.
Jason.
- Brett
On 18/10/2006, at 12:54 AM, Emmanuel Venisse wrote:
The introduction of continuum profiles isn't impacted by the
DefaultContinuum refactoring.
If we don't provide a full continuum profiles implementation in 1.1,
I think we can do a minimal one with only the possibility to choose
the maven1/maven2/ant/java home directories and use them instead of
using maven/ant/mvn/java from the PATH. This feature isn't big to do.
In 1.1, I'd like to see the possibility to choose in build
definitions if a project is built with an update (like it's done
actually) or with a clean checkout.
The last point, I'd like to see in 1.1 is the dependency/parent
change build-trigger.
All these features are awaited by users since a long time. I don't
think the implementation will take lot of time, so they can be add in
1.1.
Of course, we need a database migration tool for the release too.
Emmanuel
Jesse McConnell a écrit :
I was going to try and wrap my head about what needed to get wrapped
up for a 1.1 release of continuum this week when I got to talking to
emmanuel this morning.
I had been under the impression that we were getting near a point that
we might want to polish things up and cut a 1.1 release but emm was
thinking that we ought to have another big push for new features
before we start polishing things up. So I told him I would mention
our talk and see what kinda interest we got from people on new
features and who might want to tackle what in the short term, or if we
wanted to put some things out into the longer term bin.
One of the major things that I had been thinking we would push off to
the 1.2 release was the profiles. Its a slightly overridden term as
it has little to do with maven profiles, but in my mind at least the
profiles were going to be 1/3 of a trinity by which builds could be
setup to run.
The trinity being: profile (maven instance, env vars, maven profiles,
jdk instance, etc), temporal ( scheduled cron, when dependency
changes, scm activity, etc) and the project group (bundle of
projects). I was figuring that those three things taken together
ought meet the requirements for building what, where and when. It
would be a matter of setting up the permutations of those three
components, for example: 2 profiles, 1 schedule, 1 project group would
make 2 instances of schedulable FOO.
Anyway, I digress...profiles would be one large feature that would be
wonderful to have in continuum, sooner the better. But it would be
pretty massive effects on the codebase. So massive that I would think
we ought to consider splitting up the DefaultContinuum object into the
workflows that have been kicked around, making things like 'Add
Project To Project Group' extensible by users so they can trigger any
other processes they might want running on those events. Trygve has
some definite ideas in this area, perhaps using the plexus-spe code.
The actions in continuum have been a first pass attempt at starting to
break things out of the DC object which is pretty big atm.
If we were going to rip the top off of the DefaultContinuum object and
add/modify in the profile concepts into the store then we really ought
to clean up the whole store api which is more painful to work with
then it really should be. joakim and I had a lot of success with
structuring things nicely in the plexus-security jdo stores and we
could probably apply a ton of the concepts there in terms of api to
the continuum-store and make it scads easier to work with.
and on and on.
I agree with Emmanuel that since 1.1 as it currently stands is not
backwards compatible (I think) with the old database we ought to just
add in what we need now...But doing this will definitely move out a
1.1 release into the new year...and is that something we want to do?
I dunno really, personally I would be cool with adding in profiles and
refactoring the core chunks of continuum up now and get it over with,
but does anyone else have anything to say on the matter? I know we
have had a lot more interest recently by folks like rahul and
christian on participating, would you guys be interested in taking on
some of these challenges with us? Theres nothing like ripping through
the guts of code to really get involved :)
thoughts? should we open this out to the users list maybe?
jesse
