When putting together an Ant script, it’s common to maintain some sort of hierarchy of “targets” which get invoked in a certain order (by setting dependencies for each). A typical script goes something like this (pseudo-ant!):

target name = "clean"
target name = "build"
target name = "test"
target name = "deploy"

Now, for a clean target, I used to do something like this:

<target name="clean">
  <delete dir="src" />
</target>

Simple and to the point eh: get Ant to delete your source directory, ready for the “build” target to, well, re-build it. Super, smashing, great. But if you use source control (and of course, you do) you may run into a whole world of trouble with this approach. For example, one of my clients uses Subversion, and I started to see these sorts of errors when attempting to commit the changes resulting from the build:

Working copy not locked; this is probably a bug, please report
svn: Directory '/Users/SeymourButtz/Workspaces/Eclipse/Naffproject/src/.svn'
containing working copy admin area is missing

Lawks! “Fear not” I said to myself, and popped off to the Ant manual “Delete” page. Once there, I discovered that I simply needed to get down with the kids, and update my Ant skillz. Don’t use the square old delete dir=… directive! Use this badger instead:

<target name="clean">
  <delete includeemptydirs="true">
    <fileset dir="src" defaultexcludes="true" />
  </delete>
</target>

Slightly more verbose, but also a lot cleaner (pardon the pun): it wipes out your src directory as before, but having that defaultexcludes flag means the operation preserves any “required” sub-folders, which typically comprise such hidden gems as .svn and .cvs

Hip hip hooray!

(By the way, more on the Subversion error can be found in the FAQ).

Filed underant, tips