ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Conor MacNeill" <co...@cortexebusiness.com.au>
Subject Re: [PROPOSAL] new propertyprecedence attribute for <ant> task
Date Tue, 12 Jun 2001 14:43:54 GMT
From: "Glenn McAllister" <glenn@somanetworks.com>
To: <ant-dev@jakarta.apache.org>
Sent: Tuesday, June 12, 2001 11:04 PM
Subject: Re: [PROPOSAL] new propertyprecedence attribute for <ant> task


> Jose Alberto Fernandez wrote:
>
> > I think that rather than adding another attribute here and another
there, we
> > need to rethink the way properties are passed and then fix the problem
at
> > the root. I think that is part of the agenda for ANT2.
>
> Fair enough, but Ant2 is way down the road - we won't be seeing this for
a
> couple of months at least and we will probably see at least one more Ant
1.x
> release before then.  In point of fact, we may see a number of 1.x
releases as
> people who need the JDK 1.1.x capability keep maintaining it.
>

OK, that is a reasonable position to take. The implication, however, for me
is that the "distance" between Ant1 and Ant2 will continue to grow with
that philosophy. That will lead to a bigger break that we may have hoped
for. It is probably true that one of our goals for minimal Ant1 -> Ant2
change is already in danger :-)

Nevertheless, I see this as the wrong way to solve this problem. The
problem for me has always been the propagation of the complete
"environment" of the parent to the child build. I believe an explicit
property passing approach would be cleaner. My approach to solving this
problem, therefore, would be to allow a mode for <ant> where only
explicitly named properties are passed to the child

<ant explicit="true">
    <property ....
</ant>

This approach would require minimal changes to the current code. It is
subtly different to the propertyprecedence approach, which I feel may
require more changes - either tagging properties with their precedence or
linking parent and child property sets in precedence chains. Both of these
may be workable, I'm not sure (I have considered property precedence
tagging in the past). Anyway, I haven't gone into the code to confirm what
would be involved but it may not be trivial.

Conor



Mime
View raw message