Marc, doing everything at once at init sounds a bit heavyweight and monolithic. Imagine you're resolving and retrieving 15 different task libraries when all you need for a given build is two or three.
Instead, I would obtain tasks on an as-needed basis. So for example, the moment you need a particular task is the moment you do: 1. ivy:resolve 2. ivy:retrieve 3. taskdef But this approach also presumes that you've created more fine-grained Ivy confs that correspond to your different optional tasks. BTW, when you're using two different Ivy module descriptors in a build, make sure to specify the file attribute on the ivy:resolve each time so Ivy doesn't use the wrong one. Good habit anyway, as far as I'm concerned. On Wed, Mar 10, 2010 at 12:46 PM, marcdb <mdeb...@gmail.com> wrote: > > Hi Mitch, > > Thanks for your suggestion. I'll try to do the following: > In my central build file (used for building all projects), I will create an > init-target which will: > * execute an ivy:init > * do an ivy:retrieve using that common ivy.xml file for the dependencies of > my build-system itself. The appropriate versions of the jars that I need to > create my ant-tasks will then be retrieved in an appropriate directory for > my build-system. > * call the <taskdef> task in ant, to create/define the additional ant-tasks > that I need. > > I hope this will work. It actually seems simplier than I first thought > :thinking: > > Regards, > Marc > > > -- > View this message in context: > http://old.nabble.com/Use-ivy-dependencies-to-create-ant-tasks-tp27850372p27855533.html > Sent from the ivy-user mailing list archive at Nabble.com. > >