ant-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Steve Loughran <>
Subject Re: Exception in Style
Date Tue, 25 Apr 2006 09:53:51 GMT
Ron Jeffries wrote:
> On Monday, April 24, 2006, at 5:51:54 AM, Steve Loughran wrote:
>>> A better man would take the hour or two it would require to do all
>>> that. My learning, teaching, and proselytizing go in other
>>> directions, and I think that's where my energy needs to be put.
>>> Right now, I think I'll stop where I am, with running code.
>> what, you dont test your build process? What about regression tests?
> In fact I don't, which is no doubt terribly damning given what I
> supposedly stand for. But I've never seen anything on how one tests
> an Ant script other than by running it. Musing ...
> I suppose I could build up a small fake web site containing just a
> few example files to be converted ... then clean all the output to
> make it a worst-case build ... though it always mostly is except for
> article generation ... all the indexes are always rebuilt ... then I
> could have the last thing in the test target compare all the output
> files to a canonical list ... could be done. My experience with the
> process is that while that would be the religiously correct thing to
> do, it wouldn't pay off, since the process now either explodes, or
> builds something that doesn't look right, and does the latter only
> when I'm changing things, which would mean that the regression tests
> wouldn't work in that case anyway ...
> What's known "out there" about how to test and regression test an
> Ant script?

the purpose of a sequence of tasks is to move a build from one state 
"clean" to another "ready-to-ship" or "deployed".  All you need is 
assertions in the build to verify that the final state meets the 
criteria you desire.

If you check out Ant from SVN, there is an antunit antlib that provides 
a set of assertions for testing things. I'll cover it the forthcoming 
2nd edition of Java Development with Ant, albeit in the context of 
testing tasks you write yourself.


>> If you can tease out the problem we could always add it as part of Ant's
>> unit tests for <xslt>; that way we know which runtimes are more trouble...
> I'll add it to my list, but think it'll be a low priority story
> unless I decide to write an article about how it's harder to submit
> a problem report on Xalan than it is to submit one to Microsoft.

File a bugrep against ant and we can cross charge it to them. One thing 
the Ant team cannot do is reassign a bugzilla bug to MS.

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

View raw message