Yup, we are certainly aware of these efforts and it would be great to
have another workshop. My understanding from the previous workshop
intent (I was not able to attend), the current efforts are focused on
simplifying the instantiation of a build system to build "add-on" bits
(i.e., features and bundles that are added to existing Eclipse
installs). Basically making it easier for project teams to get a build
system up and running. This covers quite a few projects at Eclipse so
is quite valuable. The challenge I mentioned in the original post is
somewhat different.
In essence "build" is an overloaded term (my bad for using it in the
original message). The system I was talking about is the combination of
PDE Build and the Releng BaseBuilder where the latter is the set of
scripts and other stuff that drives the former to create the Eclipse
project builds. For all its foibles, PDE Build does quite a reasonable
job of actually building (aka, compiling) bundles, fragments, features,
... The current setup handles innumerable special cases and quirky
things that come with building the base part of Eclipse so we have to
be careful to avoid biting off more than we can chew.
Historically we have not had a reasonable provisioning story so we
overloaded PDE build and the BaseBuilder to do a lot of grovelling in
the bytes, fetching, zipping, chmoding, transporting etc. With the
advent of p2 we have the opportunity to improve the situation with
modest effort *around* the core parts of PDE build by letting it do its
thing and let p2 do the byte pushing/pulling/management. That
opportunity was what I was trying to highlight in my original post.
Anyway, I just wanted to provide that clarification and state my
support for having a build workshop. FWIW, the first week of Oct is
not great for me. After that should be good.
Jeff
Bjorn Freeman-Benson wrote:
Eclipse PMC members,
As you probably know from blog posts [1],
bugs [2],
and the wiki [3],
there's a general community effort afoot, supported and abetted by the
Foundation, to create a common build infrastructure across all Eclipse
projects. It would seem appropriate (and the best use of limited
resource) to have a common effort in this area, improving not only the
Eclipse PMC's build systems but also the common build infrastructure
(for all projects who want to use it)*.
I'd like to propose a face-to-face kick-off of a unified effort. Early
October? Ottawa? Build workshop number three with the specific goal of
putting together a technology plan and a project plan. How about it?
- Bjorn
In the 3.4 round build was a big problem. PDE build and the releng
builder have evolved over time to handle many subtle and complicated
usecases. While it is great that it handles these, the infrastructure
has evolved into an unmanageable, brittle and monolithic code mass.
Fundamentally it is inhibiting teams from making progress quickly and
it
is stressing the build and releng teams. Further, as our consuming
community grows and diversifies and things like e4 and Equinox blossom,
the load on these teams and the technology will only grow.
* note that "common" means available to all the projects if they want
as opposed to "standard" as in "standard ip log" which means the
projects must do it that way. In my vision, projects can still do their
own builds in their own custom way if they want to.
--
New Page 1
[end of message]
_______________________________________________
eclipse-pmc mailing list
eclipse-pmc@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipse-pmc
|