We are currently using a modified version of ant to do this. It basically provides the functionality to override target in sub builds (with the ant task). We use it so that projects can be built independently or as a sub project of another project using the same build.xml file. This allows you to redefine or set a sub project target to a no-op.
I posted a long description of why we did this and why we felt it was necessary a couple months ago on this list if you want to look at the archives or I can dig it up and repost if you want. Tim Tye -----Original Message----- From: Steve Cohen [mailto:[EMAIL PROTECTED] Sent: Friday, June 06, 2003 9:54 AM To: [EMAIL PROTECTED] Subject: Overridable targets? Here is a radical idea that I'd like to kick around a little. The prohibition against more than one target of the same name in the same build script seems logical, but it impedes a form of reuse that I think would be very handy to use. If you include one build file within another (using ENTITY inclusion mechanism) and you want to use all the targets in the included file except one, which you would like to replace with a slightly different functionality, and that target is called by another in the INCLUDED file, you have to go through a lot of annoying stuff using if, unless, etc. and perform redesigns on your script that render it far less clear and less readable. In many ways it feels like a straitjacket. SO... what if... targets followed the same rule as properties???? That is, if two targets with the same name are found in one build script (after inclusions performed), the first is used and the second discarded (instead of as now, when this throws an error). Then you could include your generic bag of generally useful targets, but redefine one "inner" target differently, to customize a particular build outside of the generic pattern in some way. The rule against two targets with the same name is an ant rule, not an xml rule, so this is in theory doable. It doesn't seem like it would break that much because this has never been allowed before, there is no previous set of semantics working against it against which backward compatibility must be preserved. There is probably some reason why this cannot work but I can't think of it, so I will throw it out to the group for discussion. Have at it! Please, though, not too hard. :-) ---------------------------------------------- Steve Cohen Sr. Software Engineer Sportvision Inc. [EMAIL PROTECTED] http://www.sportvision.com --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
