On Jan 10, 2008 5:13 PM, Dominique Devienne <[EMAIL PROTECTED]> wrote:
> On Jan 10, 2008 8:25 AM, Xavier Hanin <[EMAIL PROTECTED]> wrote: > > So, what do you think? Do you think this would be useful? Would you be > > interested in contributing? Do you think a new Ant sub project would be > a > > good fit? > > I too like Ant, and in the past worked full time on doing ad-hoc > builds for many different projects which required *a lot* of > customization. Quite a few projects were mixed Java and C++, and a lot > of the different teams for which I was doing builds had their own way > of doing things, so having a standard and uniform project structure a > la Maven was not possible (and Maven is mostly Java-only, and > difficult to extend IMHO). > > Despite this diversity, I was still regularly refactoring builds to > reuse as much as possible, and even achieved a high level of reuse, > where a project's build was down to a custom descriptor for the > project (I was doing my own dependency stuff before Ivy came along), > an import, and zero or more override targets. > > But the cost for this common build that was supposed to be generic > enough to work with all these diverse projects was increased > complexity, to the point that almost no one could customize a > project's build because of the need to have a deep understanding of > the common build and Ant's intricacies. For example, every target from > a normal (concrete) build had to be split into 3 in the common > (abstract) build, to add a pre and post "hook" target to allow > customization in the concrete build. Dealing with paths and filesets > was complex. I don't remember all the details (it's been 2+ years), > but I clearly remember I was "fighting" Ant, and I have a lot of > experience with Ant. I too sometimes feel like I'm "fighting" Ant, but I always find a solution. And the more I learn the less I fight. But I understand your comment, and I agree that Ant has some drawbacks that we will have to fight with while building a resuable build system based on it. But it also has many advantages, one of the most important IMO being its very large user base. This means that a lot of people out there are able to understand how a well written build script work. So if we base a build system on Ant, a very large number of people will be able to review it and tweak it, especially in the java community. But this won't be a silver bullet either, and I don't foresee this system replacing Ant itself. For those having very specific needs, I guess writing ad hoc build files will still be the solution. What I'm targetting is the users starting projects from scratch, or people who want to move away from Maven, or considering to move to a structured and uniform build system. > > So I guess I'm saying that I agree with you on the goal, but having > tried to achieve it myself to serve the need of the builds I was > managing proved quite difficult, and Ant has to come a long way still > to support what you describe IMHO. Sure, I should have tried to modify > Ant itself to make it easier, but that's no easy task, Agreed, but that's also why I'm discussing this on this list: here we have all the Ant gurus (including you :)), so if we can get some of them at least interested in the project we could even maybe tweak Ant when necessary. and life made > it so that my involvement with Ant took a back step, and I'm mostly > watching from afar what's going on in Ant. I've never used Ivy for > example. I'd probably have replaced my dependency stuff with Ivy by > now, although my stuff was C++ aware and was pulling not just jars > with also C++ headers and libraries, packaged sources of dependent > modules for debugging, etc... BTW Ivy is language agnostic, so you can use it for non java dependency management. But this is out of topic. > > All the power to everyone that wants to achieve the goal of reusable > scripts, but my own experience was quite mixed, despite the fact that > it was successful enough to still be in use today more than 2 years > after I'm gone, mostly untouched. I understand, thanks for sharing your experience, I guess if the job was *so* easy it would have already been done :-) > > > I hope these words don't discourage you though! Don't worry :-) > But from the energy I > can read in your posts, I don't think I could ;-) Cheers, and good > luck, --DD Thanks, Xavier > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > -- Xavier Hanin - Independent Java Consultant http://xhab.blogspot.com/ http://ant.apache.org/ivy/ http://www.xoocode.org/
