ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Steve Loughran <>
Subject Re: The Cactus/JUnit-Task problem
Date Fri, 15 Feb 2008 14:35:24 GMT
Stefan Bodewig wrote:
> Hi all,
> [Petar, it would be good if you subscribed to dev@ant, it is not that
> high traffic anyway]
> last night I chatted with Petar about the backwards incompatible
> change to the JUnit task we introduced in Ant 1.7.0 that broke Cactus.
> Cactus' Ant task extends the JUnit task (and if memory serves me right
> is one of the reasons that a bunch of methods in JUnitTask have
> protected access in the first place).  It used to override execute()
> completely and invoke the execute variants that acceps tests or lists
> of tests (I don't recall which).
> This doesn't work any longer since execute() performs setup work on
> the delegate that decouples Ant from the junit library and the execute
> variants rely on this setup.
> On my way home I thought that maybe the easiest solution would be to
> have the execute variants check whether the setup has been performed
> and if not - simply do it themselves.  The appended patch does just
> that and I'd like to get some feedback.
> The patch would make deleteClassloader protected so that subclasses
> can cleanup themselves, this may not strictly be necessary.
> With this patch in place, Cactus should be able to use Ant without any
> modifications, but could benefit from more control over resource
> cleanup if it wants to.
> BTW, I'd love to merge whatever solution we agree on to the 1.7 branch
> and have it go into 1.7.1.  Right now Cactus users are forced to stick
> to 1.6.5 and we should change that.


> Of course Petar should make sure that Gump can build Cactus so that he
> can hit us if we break it again. 8-)


Steve Loughran        
Author: Ant in Action 

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message