On Thu, Oct 9, 2008 at 2:15 PM, Paul King <pau...@asert.com.au <mailto:pau...@asert.com.au>> wrote:

Paul Duffy wrote:

Jason Dillon wrote:

On Oct 7, 2008, at 6:04 PM, Paul King wrote:

But just on the "eat your own dog food" analogy, I think it will remain important to support ant groovy and groovyc tasks for numerous releases, so we should presumably still be eating that food too?

IMO this analogy only really applies to projects which produce build tools, so like the gradle project should should gradle, and gant should use gant to follow in the "eat your own...". But groovy isn't a build tool so IMO the analogy doesn't apply.

Completely agree. I don't get the analogy either.

The Groovy distribution bundles an Ant task. The current build uses that task, so we know that it at least works for cases we cover. If we move to a different build system that no longer uses that task, we have more chance of shipping a broken Ant task (without even knowing it - the current test coverage covers the basics but is by no means complete). Just explaining the analogy not trying to stir the debate further.

The current tests provide more coverage on the ant task then the build itself exercises. While there could be more coverage than exists, moving to gradle won't create any testing gaps IMHO.