I agree. Given that I think its better to stick with tasks that produce
properties and references to Paths and filesets.

On 3/27/07, Xavier Hanin <[EMAIL PROTECTED]> wrote:

On 3/27/07, Gilles Scokart <[EMAIL PROTECTED]> wrote:
>
>
> I started to try to scoping for the ivy:settings with the same idea than
> Maarten (<ivy:settings id="my-settings" settings="settings.xml" />
> <ivy:resolve resolveId="my-resolve" settingsRef="my-settings" />
> <ivy:cachepath resolveId="my-resolve" settingsRef="my-settings" />)
>
> And during the test I notice a side effect that I didn't expected.  The
> trace of the old configure task are now present in the first task to use
> the
> settings (of curse, you would say).  For the settings, it is not really
an
> issue.  However, if we go up to the ivy:cachepath datatype, it might be
> more
> problematic.  We could end with all the resolve traces under a [java]
> task,
> which might be disappointing for the user.
>
> We should probably see it in action before really evaluate the impact,
but
> that's a potential issue.
>
> What is your feeling about that?


I agree that users will certainly be disapointed by traces like that, but
fortunately we can control the output. It will require some more work on
the
Messages API, but I think we can log using another task logger, and thus
seeing the configure trace in the settings or configure task, even if it's
triggered by a cachepath in java.

BTW, according to your discussion about datatype lifecycle on ant mailing
list, I think the major concern for now is that datatype have no real
lifecycle, so I don't think converting cachepath and cachefileset to
datatypes is a good idea with current Ant version.

- Xavier

Gilles
>
>
>




--

- Eric

Reply via email to