No offense taken. :)
The "unhacking" you refer to is at this point mostly related to hard-coded path assumptions that were valid on emf.torolab and later on
emft.eclipse.org. Once that is made more generic (or, really, once
build.eclipse.org's paths are supported), the rest will flow more easily.
The proof of concept of this will be getting GEF to build -- and not just build, but have all the web UI work too (which again, is busted due to path inconsistencies between servers), then publish a build (including update site, signing, packing, digesting, etc.).
Once that's proven, we can migrate the PDT build I have (running on
emft.eclipse.org using the "modeling build" version) to
build.eclipse.org, and even look at creating a VE build. The reason I suggest all these as test-candidates is that none of them are in Modeling -- they're all in Tools. In creating the PDT build, I've fixed a number of issues in the Modeling build, but have not yet ported them into Dash due to time constraints (and my now-past wedding / vacation).
In terms of what others can do to help... well, I'll need to ponder the best steps for inserting others into the flow. Being available online when I'm working on stuff (in the event I can't figure something out (foundation database queries, eg., into the portal) or don't have write perms (because I'm not root)) is perhaps, for now, the best support option available... especially since my current priority is wrapping up GSoC and Ganymede SR1.
Eventually, we'll be able to split off into workgroups -- for example, taking a given PHP or shell script and improving/replacing it with something more optimal for use on build.eclipse. This includes accessing signed zips locally instead of via ssh, getting dependencies locally instead of over http, etc. There are also things to improve in terms of security -- using the dev.eclipse/portal single signon instead of using .htaccess to control access to the build UI.
I'd also like to look at having a third in-person workshop session, perhaps in late October or early November. Could be Ottawa or Toronto, depending on who wants to attend.
Nick
On Tue, Aug 26, 2008 at 1:56 PM, Denis Roy
<denis.roy@xxxxxxxxxxx> wrote:
Dashers,
I have a question. During our short but interesting workshop, I spent a fair amount of time watching Nick Boldt 'unhack' modeling-specific hacks from his current Build Infra code. I mean no disrespect to Nick's build system -- it appears to be complete, and it seems to work very well for all the building he does. But a generic, simple builder it is not.
My question is: Should we keep pushing forward with the current goal of taking Nick's code and adapting it to work (which is currently stalled as most of us are useless without Nick doing all the work), or should we (Bjorn, myself, Nick, Wayne, anyone?) consider creating a small, simple system from scratch?
Thoughts?
--
Denis Roy
Manager, IT Infrastructure
Eclipse Foundation, Inc. -- http://www.eclipse.org/
denis.roy@xxxxxxxxxxx
_______________________________________________
dash-dev mailing list
dash-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/dash-dev
