On 9/25/06, Stefan Bodewig <[EMAIL PROTECTED]> wrote:
On Mon, 25 Sep 2006, Peter Reilly <[EMAIL PROTECTED]> wrote: > On 9/25/06, Stefan Bodewig <[EMAIL PROTECTED]> wrote: >> >> On Fri, 22 Sep 2006, Peter Reilly <[EMAIL PROTECTED]> >> wrote: >> >> > I propose that we defer handing id's until the processing >> > stage. >> >> Will probably break something like [...] >> which is copied straight from the examples for <script> in the Ant >> manual for version 1.6.5. > > Yep, this is correct ;-) , I have removed the example from the > documentation. I've seen that 8-) >> I.e. it breaks something that was documented to work before. One >> could even assume it was a best practice, given it was used in the >> manual. > > Yep, one may assume it was best practice, but it is not! Never claimed it was. >> This example predates Ant 1.2 IIRC. > > I know, that is why I have raises the reference deferral > as a issue. > > I really cannot see the use-case described by the example. > I would not think that there are any examples in the "wild". You've identified two cases in Ant's tests yourself.
I meant the use-case of script tasks getting tasks in other targets, modifying them and then executing them. The two cases in ant's tests do give me pause for thought, but I feel on balance that they work on the anti-pattern of "it works by accident" and thus are very brittle. One checked the contents of a reference to a fileset defined in another target as a nested element. It took me a long time to figure out why it worked at all. I feel uncomfortable about changing this, even though you are probably
right that it it should better be changed.
I am uncomfortable as well, but I think that this change will stop a lot of strange bug reports and will allow some interesting uses of references. Peter Stefan
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
