On Sat, 26 Feb 2005, Robert Virkus <[EMAIL PROTECTED]> wrote: > the classpath is fine, but it isn't set for the two target devices > anyhow - so in both rounds it is empty.
That's good news. > The bootclasspath, however, is appended (the first device is a > MIDP/1.0 device, the second one a MIDP/2.0 one), so in the second > round I have something like this: > > BootClasspath=/home/enough/workspace/enough-polish-build/import/midp1.jar:/home/enough/workspace/enough-polish-build/import/midp2.jar I'll look into this. > This is not causing the current trouble (since all the "duplicate > class"-errors are in neither of both JAR files), so where are they? Ant should log the command line it passes to javac at the verbose level, does it look suspicious. > So re-using the Javac task seems not to be possible. Maybe we can fix it, > So what could I do? Right now, I wouldn't give up on fixing the task. > I now would create a "wrapper"-task that accepts all attributes as > well as nested elements just like the Javac task, so that I could > instantiate a fresh Javac for each round. This basically is what Dominique talked about when he said it would be best if you could get hold of the UnknownElement (which is just that wrapper, but written and populated by Ant). Even if there might now be a way to get hold of an UnknownElement directly (I'm not sure, it may be possible), you can use it as your wrapper and populate it yourself. Stefan --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
