It was a "perfect storm" of issues. Builder dependencies were being executed twice because for some dumb reason I wasn't reusing the builder execution macro, which meant the check to see if the builder was already built was not happening. This turned out to be a blessing though, because in the customTargets.xml I was bundling XULRunner in the folder that releng copies zips to, instead of the "working" directory. The error was happening because on the first richfaces execution, it couldn't find the zips, since the builder was still running, and the zips weren't in the releng dir yet. But on the 2nd try of the richfaces build, the zips were there so it executed correctly (and gave the illusion of a false positive). At first I tried disabling the builder dependencies because I thought it was the source of the problem. This gave us a richfaces zip w/o XULRunner since it was never executed a 2nd time, which helped me find the actual cause of the issue: The XULrunner bundling code needed to use the "intermediary" zips instead of the "final" zips in customTargets.xml.

Max Rydahl Andersen wrote:

cool - what was the cause?

-max

Alright -- the error has been fixed.  We shouldn't be seeing false
positives anymore (and XULRunner plugins are being bundled correctly
again). Forcing a build now..

Marshall Culpepper wrote:

This is my fault -- I'm trying to isolate the "false positive" problem
and I've finally found it. Should be fixed really soon

-- 
Marshall Culpepper
Red Hat Developer Studio, JBoss Tools
marshall@jboss.org