I can only ask this in vague terms since the code is IP. But basically what happens is this, I have a Hudson autobuild setup hooked with perforce that has 2 builds steps.
1. A setup script that deletes old source/jars in our test framerwork. Then copies the new (synced from perforce) source/jars back
2. build and test through ant (framework compile and copied in source compile, run tests)
Problem is, once in a while, the build will break with compilation problems from the code that is copied over (cannot find symbol,etc). No matter how many rebuilds, it doesn't fix it. So, I would think it would be the setup script.
But when I run it with the setup script but commenting out the part that copies over the source(it still cleans up, and copies the jars) then runs the ant call (compiles the framework, tries to run the test, but no tests to run) it succeeds.
If I then, uncomment the part I commented out earlier and build again, it succeeds. It keeps succeeding until the next failure happens (so far about every 8). What I don't get, is why it compiling without the copied over code makes it work, but then compiling with keeps it working whereas before it wouldn't.
I was thinking maybe it was the perforce plugin, since usually this happens the next day I come in. But there are also Perforce triggered builds that succeed. hmmm
Sorry, this is a bit open-ended and vague since I can't post any specifics, but if anyone at least has a compilation conflict from what I've seen, it is totally worth it.