On Mon, 9 Feb 2004, Adam R. B. Jack <[EMAIL PROTECTED]> wrote:

> For directories ... I see that <ant has the concept of 'basedir'
> (relative to the <module, I believe) and this is for kicking off a
> build.

Yes.  I assume it is named after Ant's own basedir attribute for the
project tag in Ant build files.

> I see that <project can override this -- but it doesn't mean "this
> is the area for a project under a module", it is not where "stuff
> is" (per se).

True.  As for overrides, see below.

Sometimes projects put build files into strange locations like a build
subdirectory of where "stuff is" and expect the build to be executed
from there.

> I see we have 'home' dir on a project, but that is for results --
> not where "stuff is".

because some projects tend to place the outputs in a very different
place than where "stuff is".  Take JUnit for example, it even moves it
outside of the current source tree.

> [I also see 'source directory' (a misnomer?) for a module. Not sure
> if this get played w/ much.]

Used by <property reference="srcdir"/>, I agree, it's not too
frequently used.

> Basically --- I feel a project (and it's contents) are not
> clearly/easily specified.

Because projects are neither clear nor easy 8-)

> Can we have a concept of 'this is the root of a project under this
> module' -- and perhaps allow values to be set when this doesn't
> work, or do projects (in Gump) just not work that simply?]

Hmm, take bootstrap-ant, ant and dist-ant.  Where does each project
live?  test-ant?

> Does it mean ' override', or does it mean 'override default'?

It means override.  This is not for a straight <ant> nested into
<project> case where the override mechanism doesn't make any sense.

It becomes important when you have multiple definitions for a
<project> and they get merged.  This gives you a chance to augment the
project definition (coming from Apache's Gump for example) with some
extra pieces inside your workspace definition.

See the rubypad.xml workspace where Sam has overridden the build
targets for Cactus and Xalan's smoketest.

> I'm starting to wonder if we need to take a step back and see if the
> Gump metadata is exactly a model we'd like to continue with,

Probably not 8-)

Stefan

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to