Where an UnknownElement is involved, need to look a little harder
Commented out for now the code which associated a project level message
with the currently active element - better to have it go to the project
level
PR: 7926
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273817 13f79535-47bb-0310-9956-ffa450edef68
- implement the BuildLogger - so it can be used with -logger
- few hacks/changes to support logging messages to stdout/stderr without
going to infinite loop. The trick is to create the logger with stdout
beeing the "real" one. Just to be sure it also logs with stdout set to
the real one. This could be done more cleanely probably - with some
changes in core.
- less verbosity ( task started/stoped doesn't need info, etc ).
- the log category is based on the real task name, combine with the
target. That means you can arbitrarily enable debug on tasks and
to tasks that are in particular target.
PR:
Obtained from:
Submitted by:
Reviewed by:
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273772 13f79535-47bb-0310-9956-ffa450edef68
There are normal cases ( antcall) where the same build file is parsed many times
( this could be avoided now - as an optimization ).
It is a problem only in import, because of the target conflict resolution.
( targets are qualified with the project name )
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273769 13f79535-47bb-0310-9956-ffa450edef68
This is _not_ an antlib, just a mechansim to make the task/type
management pluggable and simplify Project ( which will just delegate
for backward compat. ).
Please review - and -1 ( or just remove/change ) if you think this
is bad.
The actual details of the plugin mechanisms are not perfect -
we need to settle on them before 1.6 is released ( same for PropertyHelper).
PR:
Obtained from:
Submitted by:
Reviewed by:
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273745 13f79535-47bb-0310-9956-ffa450edef68
Display a warning.
I don't understand very well the code - but we take the name from
"references" ( which is specified by the user ), and assume that
all of them exist. For some reason that's not allways true.
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273734 13f79535-47bb-0310-9956-ffa450edef68
Input is calling setNewProperty - but if defaultValue and value are null
we'll get NPE
PR:
Obtained from:
Submitted by:
Reviewed by:
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273732 13f79535-47bb-0310-9956-ffa450edef68
There is no standard for how jsp generates names - and jasper
changes from time to time :-)
PR:
Obtained from:
Submitted by:
Reviewed by:
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273728 13f79535-47bb-0310-9956-ffa450edef68
This is just the first round, it can be improved - not sure what's
the best way to deal with classpath specific issues in path.
The problem is that CLASSPATH ( as given in the system property,
and added in concatSystemPath ) may contain relative paths,
and they are relative to user.dir, not the project basedir.
Since gump is using relative paths, the launched java didn't find
the classes it needed, returning the strange -1 error.
Of course, the test suite could be more informative too :-)
PR:
Obtained from:
Submitted by:
Reviewed by:
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273727 13f79535-47bb-0310-9956-ffa450edef68
The test relies on the order of attributes ( cr set before eol ). Probably
other tasks are in the same situation.
The original change tried to reduce the dependencies on SAX - we may
use DOM or direct API calls in future, and the code will become very
complex. I just used the same thing that SAX is using ( 2 Vectors to
preserve the order ). We could also pick one ( SAX2 attributes ) and use
it in all cases.
PR:
Obtained from:
Submitted by:
Reviewed by:
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273725 13f79535-47bb-0310-9956-ffa450edef68
is present, use it.
I think this can be further cleaned up to use the right parent loader
etc. But probably the best would be to just use the loaderRef.
Or just deprecate Definer completely - when we have a good antlib solution
Also added getters ( it's nice to have access to info - especially with
dynamic properties :-)
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273720 13f79535-47bb-0310-9956-ffa450edef68
ProjectHelper delegates for backward compat.
I deprecated some of the methods, easy to revert if you think they
are needed.
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273712 13f79535-47bb-0310-9956-ffa450edef68
Project delegates to PropertyHelper ( for backward compatibility ).
As soon as the API is stable ( i.e. do we pass ns ? Object ? )
we can deprecate the methods in Project.
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273711 13f79535-47bb-0310-9956-ffa450edef68
This is "slightly" different from embed - if dynamic properties will be
accepted in 1.6, it is better to do it right. Embed uses few hacks to
trick the ProjectHelper.
PropertyHelper includes all the code related with property manipulation
from Project (cut&paste). I added a very simple hook mechanism ( Filter/Valve
like ) for the most common operations.
The API is obviously far from final - someone who really understand "user"
and "inherited" properties should review it and make few changes.
In Project, all properties fields are private - so all can be removed.
The methods related with properties will just delegate to PropertyHelper.
From a user point of view - no visible change ( hopefully :-). Even grant
will continue to work ( but won't be able to use the new features ).
Benefits:
- cleanup of Project. Less code and better organised.
- Property handling will hopefully be cleaner too
- we get to add APIs for namespace support, and maybe support non-string
properties ( JSP-EL like - that needs to be disussed, but IMO it would
be very helpfull ). While refs are Objects, the main distinction is imutability.
Also:
- apps embeding or extending ant can fully customize _all_ aspects of
property processing, including ${} replacement and even the syntax.
- it should be very easy to hook a different storage mechanism ( i.e.
integrated with the embeding app, without requiring copy of properties ).
- it should be possible to avoid copy when creating execution frames
( by using a chain that keeps changes and delegates getters ).
- dynamic properties support ( my original itch :-)
Please wait few days before firing, it seems all tests are passing
and gump works - but I'm sure there are few problems :-)
( I'll commit the change in Project later today )
PR:
Obtained from:
Submitted by:
Reviewed by:
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273709 13f79535-47bb-0310-9956-ffa450edef68
support imports. The original helper won't work ( because of eval order ),
and embed will fail too ( since it can't use 1.6 features - one of the
goals is to work in ant1.5 ).
I'm ready to move the dynamic properties and the component hooks - I'll
porbably do it this weekend ( most people are in vacation, so fewer
will feel any breakage ). In any case [embed] should now used only
if you want to use (some) new features in ant1.5.
git-svn-id: https://svn.apache.org/repos/asf/ant/core/trunk@273707 13f79535-47bb-0310-9956-ffa450edef68
Stefan Bodewig
Conor MacNeill
Magesh Umasankar
Costin Manolache
Erik Hatcher