ant-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ray Tayek <>
Subject Re: some design questions about "ant in anger"
Date Fri, 19 Apr 2002 07:48:06 GMT
At 09:45 PM 4/18/02 -0700, you wrote:

>----- Original Message -----
>From: "Ray Tayek" <>
>To: "Ant Users List" <>
>Sent: Thursday, April 18, 2002 17:59
>Subject: some design questions about "ant in anger"
> > ... recommends having standard top-level targets like: build, test, ...
> > ... it would seem that ant is smart enough to look at the build.xml 
> file in each
> > directory and do the right thing. is this more or less how it works?
>well, you use <ant> to hand off whatever targets you want to projects in
>subdirs; it is not automatic.

ok, he is not that smart. but iirc, he will compile everthing in the tree 
(including subdirectories), so perhaps we need to exclude those files or 
only include the ones we are interested in so we don't compile something in 
a subdirectory that has it's own build.xml file, correct?

> >
> > ... buildtool that contained all of the JAVA_HOME, ANT_HOME properties (or
> > perhaps the build tools themselves?). perhaps that should be one of the
> > top-level targets?
>not a toplevel target, but you should keep as much as you can in SCM; ant
>and all other java libs can go in there.


>  I draw the line at java versions,
>oh, wait a minute, I draw the line at JDKs. We do have a JRE installer under
>SCM so that when we burn CDs it pulls everything (even 50MB app server
>installs) from one place.

oh wow! murphys moores law :)

> >
> > the author suggests having "internal" targets at each level like: init,
> > compile, link/jar, staging. not sure what this is all about, buy it sounds
> > sorta like the separation between interface (the standard top level
> > targets) and implementation (the internal "targets").
>more that you should have well known targets, be they public ones or
>internal ones.

got it.

> > otoh, if i am using ant calls, what happens to all of the properties that
> > have already been defined by descending the tree up to the point of this
> > particular ant call? do these properties get lost when the new ant call
> > gets made?
>your choice

yes, eric pointed out the ant and antcall (inheritAll).

> > when ant sees new build.xml and files do these
> > "go away" when that ant call terminates?
>yes. View it as a heirarchical context of properties; as you return up the
>chain the context gets rolled back.


> > another issue that seems problematic is that of testing. we will be doing
> > servlets, so there are at least four (4) levels of testing: junit, cactus,
> > httpunit, and someOtherTesting. ...
>cactus needs ant; httpunit is simply another junit target you run *after* 
>so it goes: build, unit tests. deploy. httpunit. cactus
>-maybe swap http unit and cactus if you want

yes, cactus seems to deploy in place somehow (at least with tomcat).

thanks for your assistance.

ray tayek
actively seeking telecommuting work
vice chair orange county java users group
hate spam?

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

View raw message