At 03:38 16/11/00 -0800, you wrote:
>Now wait a minute, do you mean "discouraged" or "not supported"? The
>problem with initializing properties outside of targets is that this makes
>initialization "a priori" with respect to executed targets. It follows that
>I can't do something nifty like have one of two different property files
>sucked in depending on my present OS.
nope. Not at the moment.
>Regardless of that specific complaint, I don't see why the semantics of the
>two examples should be different, and I protest: by the time I use my
>defined path, "myprop" has been resolved, so what's the problem with
>interpolating ${myprop} when I actually use the path reference?
because it would brake current semantics. It will change in future for
Ant2.0 but the next couple of iterations will not have it changed.
>It sounds like Ant is trying to eliminate the conditional setting of
>properties entirely, and is trying to have its cake and restrict it too:
>either don't support property initialization in targets at all, I argue, or
>support it properly with runtime interpolation of path and other referenced
>elements. As it is now, the semantics feel very "buggy" and I'm going to
>have to throw out referenced paths entirely to proceed.
Right - but many other people complained for each differnt change. This
change will make Ant2.0 (I believe) but not now because we want to keep
compatability for short term.
Cheers,
Pete
*------------------------------------------------------*
| Despite your efforts to be a romantic hero, you will |
| gradually evolve into a postmodern plot device. |
*------------------------------------------------------*
