define ids or paths and use Ant's location magic for filename
resolutions in the XML file.
PR: 11321, 12045
Submitted by: Paul Christmann <paul at priorartisans.com>
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273484 13f79535-47bb-0310-9956-ffa450edef68
ant-. It's just cosmetic - if you feel 'need' is better I'm fine.
Also added a property to customize haltonfailure - sometimes it's
nice to run all tests and then look at the logs and see all failures.
Plus a quick way to run a test without a full rebuild.
Again - if you see any problem, I can revert.
PR:
Obtained from:
Submitted by:
Reviewed by:
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273464 13f79535-47bb-0310-9956-ffa450edef68
Submitted by: Robert Clevenger <Robert.Clevenger at oracle.com>
H comes before I not after N, at least in German alphabets ;-)
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273456 13f79535-47bb-0310-9956-ffa450edef68
Remove the attempted merge with UnknownElement - while it
would be nice, it's not possible without affecting backward compat
( since tasks could use both of them ).
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273449 13f79535-47bb-0310-9956-ffa450edef68
Added an experimental TaskDiscovery that will use commons-discovery
to find tasks using the ProjectComponentFactory hook.
This is just to play with different antlib mechanisms.
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273439 13f79535-47bb-0310-9956-ffa450edef68
I commented out for now the special case that does imediate
creation of task, and changed the behavior to all-lazy.
This seems to pass the gump test, and doesn't seem to
create any major problem. Given that lazy creation was allways
a possibility, I would think it is a task bug if a task that can't deal with
that.
After this settles down I'll eliminate the latest special case for
nested elements ( now it'll all be dealt only by UnknownElement ).
Also, added more code to deal with namespaces. The logic should
be in UnknownElement/RuntimeConfigrable - we should just pass
the info ( in the case that an XML source is used )
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273438 13f79535-47bb-0310-9956-ffa450edef68
Again - this is a proposal and an experiment, all features will
eventually get to a majority vote and removed or moved out
of ant if it doesn't pass.
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273437 13f79535-47bb-0310-9956-ffa450edef68
stores to implement their own policy.
This feature will not work in ant15 ( for now ).
Also changed Project to Object - and use it as a context. This makes
the callback more generic and the hooks less dependent of ant.
The only major issue on dynamic properties for me is the JNDI stuff.
I think the Context interface could be a better abstraction, but
on the other side it is much more complex. I also want to finalize
the JNDI property source and abstract the properties via JNDI
so tasks can be created without any dep on ant, but still accessing
properties and refs.
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273434 13f79535-47bb-0310-9956-ffa450edef68
I would like to extend it a bit more in the future - to support
properties like "${property}" without a string conversion.
In other workds, if a dynamic property is used ( which return an
object ), the object should be used as such in the introspection.
That would greatly simplify tasks and allow more flexibility.
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273433 13f79535-47bb-0310-9956-ffa450edef68
behavior and control the creation.
This will work for all code that checks instanceof, but
will obviously fail if any code explicitely looks at the class
name ( which is very unlikely - so far no code did that in the
gump run or otherwise ).
The UnknownElement2/RuntimeConfigurable2 will be further refactored
to make them more independent of the XML representation and
have a cleaner API. All behavior will be moved out of the XML
processor and into this pair. Of course, full backward compat will
be provided.
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273432 13f79535-47bb-0310-9956-ffa450edef68
Full backward compatibility is provided - the project
is used to create tasks. The mechanism allows full
flexibility in implementing any kind of antlib policy,
and allows applications embeding ant to integrate in the
component creation process.
Probably a good improvement would be to move the 'default'
behavior in a separate plugin ( DefaultComponentFactory - to
use the project component storage ), and deprecate the use
of Project for project component management.
After more feedback is received I'll propose this for the
main branch.
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273430 13f79535-47bb-0310-9956-ffa450edef68
(and may thus ignore some attributes or nested selectors).
On a side-note. I found it amusing to read
* <p>Currently Delete extends MatchingTask. This is intend <i>only</i>
* to provide backwards compatibility for a release. The future position
* is to use nested filesets exclusively.</p>
in Delete.java. This is sticking in there since almost exactly two
years ago. Revision 1.8 2000-10-04. Revision 1.9 went into Ant 1.2. 8-)
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273428 13f79535-47bb-0310-9956-ffa450edef68
dependencies.
* <patternset> --> <selector> for most stuff.
I've kept optional.jar for now as I'm not sure whether I've really
covered all classes (and no class appears twice).
Feel free to change names or optimize my "mutiple nested selectors"
constructs. ;-)
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273426 13f79535-47bb-0310-9956-ffa450edef68
Stefan Bodewig
Stephane Bailliez
Steve Loughran
Peter Donald
Costin Manolache