[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
Re: [cbi-dev] Bundles that may be improperly/incorrectly build with Tycho
|
I was asked to help with #2 at end of last summer and I believe we made
major progress already. Most of Eclipse Platform projects are migrated
to Tycho already and we need to deal with the few remaining ones.
Instead of trying to fundamentally redesign CBI after this much effort
has been invested into it already, I think it makes more sense to
understand problems with each of the remaining projects and deal with
them individually.
As for the projects outside of Eclipse Platform, this is really up to
them to migrate to CBI or stay on whatever build tools they are using.
The goal of CBI is to provide *common* build approach and supporting
infrastructure, not to wrap multitude of existing build solutions.
--
Regards,
Igor
PS: the above is my personal unauthoratative opinion, I am neither an
Eclipse Platform committer nor authorized to make any project management
decisions on behalf of Eclipse CBI effort.
On 12-07-17 2:41 AM, Krzysztof Daniel wrote:
I think the real question here (especially when considering -1 on
maven-ant-runner plugin [1]) is what is really the CBI is supposed to
do:
(1) Is it supposed to be a thin layer on the top of existing projects
for the sake of entire Eclipse build easiness?
(2) Is it supposed to migrate all projects to maven (kind of back-door)
and obsolete the PDEBuild?
While I really would like to see 2 to happen, I don't believe it is
possible in the short or even mid-term:
* Some projects were created long before maven and they have a structure
which does not fit into the maven lifecycle. The common maven
anti-pattern is storing ant pre-compiled code (native and jars) in the
repo. (This is also a big problem for Fedora). On the other hand, this
anti-pattern is a really good patern for some other goals (saving
bandwidth). Heavy project refactoring (and bugs like [2]) seem
unavoidable.
* I cannot commit to that refactoring. Time is important, but what
really limits me is the lack of domain knowledge, visible in [2]. I can
read the build files, but if they seem to be not processed, or manually
invoked, how can I automate them? And that brings the third and the last
point:
* Many commiter's do not have sufficient command of maven to introduce
it to their projects before SR1 (note that in many cases it will be a
very dangerous change, and in some places it will require rewriting of
the customBuildCallbacks).
Given that quality concerns appeared again, I really would like to get
expectations that you all have about CBI.
Note that in case we go for (2) we may need much more coordinated
effort, which should include all non-tycho project leaders.
[1] https://bugs.eclipse.org/bugs/show_bug.cgi?id=385154
[2] https://bugs.eclipse.org/bugs/show_bug.cgi?id=384874
