[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [ecf-dev] ECF build v2 -- Build easier

Hi Thomas,

Sounds like a discussion which needs to go off-line, over a good beer. I appreciate your comments and replies, and I hope I can convince you I do not mind hearing people speak their mind about my rantings, no matter how strong the verbiage ;) I am the first to admit that some of my statements could benefit from a stronger argumentation, and some, as you point out, are based on misunderstandings. What's left should guarantee a lively discussion at the bar, at the next EclipseCon :)

Even though I prefer open, direct discussions, I also hope you understand that I only refrain from rebutting too much, because I see a very long interleaved reply train coming up here, possible without end. Especially on the EMF thing, which now affects Buckminster, too. I *do* have strong opinions on the matter, and I'm waiting for a better moment to voice my concerns on some e4 design decisions, as Doug Schaeffer already did http://cdtdoug.blogspot.com/2009/07/my-biggest-fear-for-e4.html

My concern is genuine, and based on actual experience of trying *everything* the Eclipse ecosystem has to offer in production scenarios, only to see productivity in the team plummet in some cases, because of scale and continuity issues. But, I'm not going to open up that can of worms, now. But it is similar to what developers have been reporting about the WSAD+Rational ball-and-chain experience.

So far, my experience with Buckminster, that while ago, was quite the contrary: a real productivity boost for several teams over different sites. Hence my cringe moment when I saw some dreaded acronyms pop up, all over again.

I do prefer to sit this one out on the bench, though, as far as post-EMF Bucky is concerned!

Best regards,
 Dann





Thomas Hallgren wrote:
Hi Dann,
I feel that some of your statements are based on misunderstandings and others on the status of the project as it were two years ago (still in incubation). Some answers inline in an attempt to clarify our position.


Dann Martens wrote:
Hi Scott,

Just to give you some feedback on your specification of missing elements in a Buckminster-based build system, I'd like to share some of my tribulations.

As I have been an early adopter of Buckminster, I have a solid two years of experience running a complete build system with said tool. I have been negotiating recently with the project lead to return some of the proprietary build actors (e.g. a very necessary packaging actor), but surprisingly met resistance in terms of those actors considered a reinvention of the wheel (sic!), as I should have written Ant scripts to do the same work, and call those Ant scripts using Buckminster.

As I recall it, the mention of "reinvention of the wheel" concerned a Javadoc actor, nothing else. Writing it was on your TODO list and my "resistance" was more intended as a hint about priorities then anything else. I'm still interested in your "packaging" actor unless there are issues with third party libraries (JBoss). There's an unanswered question in the thread about what that actor does exactly.

Needless to say, I disagree with this position. The original appeal of Buckminster was, from my point of view, to move beyond the Ant script bonanza. I can only testify that our pursuit of what we believed the vision to be, worked incredibly well, and we managed to roll this thing out in the company over different teams without the headaches of getting started with Buckminster as it still stands today. Unfortunately, that is no longer the road the Buckminster wagon seems interested in traveling on.

Your text about the Javadoc actor and my answer, in verbatim:

>> And high on my TODO list: the Javadoc actor; I'm not a fan of falling back to Ant for standard artefacts in release engineering.
>>
> Neither am I. But I'm not a big fan in reinventing the wheel either ;-) and Ant has a fairly good Javadoc task readily available, so unless you
> have endless time to spare, why bother?


Do you see the "Neither am I" in there? What I meant was that there's a balance between purism and effort. We probably would like a native Javadoc actor eventually but given the effort it would take to write one, perhaps other things are more interesting and should have a higher priority. The Ant actor will be retained in any case (hard to remove once people has started using it).

In addition, I was surprised to see the project itself had made no substantial progress over a very long time, at least in terms of what you might expect from a product maturing and evolving as they normally do.

I think that we've made substantial progress. Both in Buckminster project itself and in other projects (P2 especially but we've helped ECF and PDE build as well) in order to improve the Buckminster offering.

Ironically, we managed to get Buckminster to do everything a releng nerd desires with our proprietary extensions, already years ago, yet a lot of that stuff is still not available today.
You managed Thomas Spiessens back then and he contributed the Subversive support for Buckminster. That was an example of very good cooperation within the community and how users can contribute to the project. But besides Thomas contribution, none of your proprietary extensions has been contributed. We don't know what "that stuff" is since you haven't told us. Please enter bugzillas with enhancement requests. That's the way to get things done.

As much as it pains me to say this, I have become very skeptical about the project. I will spare you a description of the look on my face when I discovered that, apart from an overhaul to replace overlapping parts with p2 (which came out of left field to duplicate a lot of functionality), now it seems that this particular build tool seems lost without EMF, thus prompting the introduction of EMF into B.
You shared some words with Ed Merks and myself about the subject and I must agree with his final assessment. The start of that discussion can be found here: http://dev.eclipse.org/mhonarc/newsLists/news.eclipse.tools.buckminster-dev/msg00580.html



But don't expect it to make your packages for you, or signing of those for that matter. You'll need an Ant script to that.

The ant-script needed for signing is already bundled. You'll find actions like "site.signed" and "site.p2" in cspecs generated for the "eclipse.feature" component type. Page 135:

http://www.eclipse.org/downloads/download.php?file=/tools/buckminster/doc/BuckyBook.pdf


And again, I'm not particularly fond of delegating to Ant and Buckminsters use of it o may change over time. But getting rid of Ant is not a top priority.


The problem I see is that there is no reason to consider moving to Buckminster from an existing system, as it stands today. You would need to contribute additional components yourself, first.
I doubt that ECF will need to do that. AFAIK, Buckminster can build ECF out of the box.

As a project manager, I would like to see a compelling reason to warrant such an effort under the ECF project umbrella. Speaking from experience, hooking yourself onto the Buckminster train means a lot of work in terms of keeping up to date with updates and bugfixes. We got tired of the heavy maintainance of our plug-ins and simply forked and never looked back.
Perhaps contributing them had been a better option? Chances are we have had invited you (or the author) as a committer to the project in order to maintain them.

Not an option anymore given Buckminster's status in the ecosystem. Even though I'm still considering forking myself (on the last EMF-free code base),
There won't be any major changes in the 3.5 codebase so you can safely remain there and benefit from the bugfixes that we provide for 3.5.x releases. We are keeping the API stable nowadays. It's not like it was 2 years ago when Buckminster had incubation status.

and have a complete snapshot at my disposal to that end, right now I'm checking out if it simply doesn't make more sense to see if building something on top of p2 isn't the way to go. It would appear p2 has almost made Buckminster redundant; I'm still investigating the validity of that claim.

So are we. And we are constantly improving both P2 and Buckminster as we move forward.

Any which way, sounds like hell of a lot of work still to do before you can actually build and release. Can you afford that?
Markus, don't you have a build system in place for Buckminster already?

Kind Regards,
Thomas Hallgren


_______________________________________________ ecf-dev mailing list ecf-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/ecf-dev