Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [modeling-pmc] Builds: The never ending nightmare


On Feb 24, 2010, at 1:45 PM, Thomas Hallgren wrote:

Hi Miles,

On 01/-10/-28163 08:59 PM, Miles Parker wrote:
Hi Ed et. al.,

I had a "chance" to play around a bit with Buckminister, because I wanted to do some product builds. I thought I had build all figured out until I tried to build an RCP product version, then found that all of the mechanisms for packaging and building are completely different! :( There was no simple path to easily turn a Hudson Athena build into an RCP build that I could discover. Unless one's idea of "easy" involves reading pages of Wiki and then hand-editing XML files for Buckminster. see for example: http://www.ralfebert.de/blog/eclipsercp/rcp_builds/. Their was a bunch of really fiddly steps needed to try to get a local instance of Buckminster up and running, including actually bootstrapping the headless environment from command-line, trying to figure out the inevitable plugin conflicts, etc... I finally gave up and just went with the PDE manual build. And in the end, its still down to a veneer over PDE, so you still really need to (try) to understand all of that.

  
The way I recall it, you were using our bleeding edge 3.6M4 and reported an error which we then fixed in our next milestone. I realize that trying to be an early milestone adopter while at the same time having to learn about Buckminster can be somewhat overwhelming :-), but as I'm sure others can testify, we do our utmost to try and help our community and it makes me sad to hear that you needed to give up.


Yes, that's true -- I think I had gotten to using the milestone via a circuitous route of not being able to figure out something about the setup for OS X; and I realized that I was missing some features that were in the tutorials. My giving up was really down to my own timing -- I have a bad habit of keeping at something and this time I was trying to be smart and just go with the simplest thing that could possibly work. So I really was tackling this at the worst possible place in terms of patience and time. I don't mean to come off as negative about Buckminster itself -- its just that I'm the only person on my project so far which means that I've had to replicate all of the same work that everyone else is doing on a team basis and prob. spent 30% of my development time on Eclipse build and packaging issues and so I think I'm starting to lose perspective. :) As we're all aware, something that seems reasonable and straightforward once you know it can seem really opaque when you don't. And then you get to the point where you're not even sure what questions to ask.

This makes me gratified that people are working on common solutions -- and I think having a common one for Modeling based on Buckminster is a great one. W/O the common Athena build that I could copy from the other Modeling projects using it I would have really been up a creek.

For the record, there is no longer need for hand editing our XML files in our latest milestone. We now have the EMF editors in place.

You know, I think the worst part for me was actually the properties files over the xml. I'm sure you've made progress on that as well. I've just found that the long chain of paths and other properties that need to be followed through from cmd-line / shell scripts -> buck / xml / cquery -> athena -> ant -> PDE and various permutations thereof became more than my mind could handle. For example, seeing an every growing number of build / provisioning paths, etc..  Of course, this is the very issue that Buckminster is designed to solve, but its inevitable to have some growing pains to get there.


Aside from product creation, now that Athena / PDE can now spit out PFS, that takes care of poor man's development time provisioning. So I guess I'm asking -- non-rhetoriacally -- what the advantage might be for those projects already on Athena?
There's always the saying - if it ain't broke, don't fix it - and that applies here too. I think you should use whatever makes your team and your consumers comfortable. We would of course love to have you on board with Buckminster and will do what we can to help if asked.

I must say I do like that idea of having more build support!

  
And we would love to provide it.

I certainly cannot fault the level of support and help that you offered! It had just gotten to the point for me where I couldn't figure out which end was up. Plus this wasn't really for an Eclipse hosted build. As it turns out, I then ran into a bunch of issues with getting a target platform setup on my Eclipse install.. It's still a bit wonked. So anyway, don't get me wrong -- I would love and really need to have an automated product build and it would be excellent to have that integrated with the Athena setup so that I can replicate locally; and I'll be ready to jump back in at some point.  

cheers,

Miles

Back to the top