Revert the move of jvm dependency analyzer to its own task. [REVERT]

Review Request #2728 — Created Aug. 30, 2015 and discarded

jsirois, stuhood, zundel
JVM compilation in a given chunk must know whether fatal
missing dependencies are present, otherwise it will write
artifacts to the cache.  Even if the compile goal later
fails because of the missing dep detection task, the artifacts
have been written, and subsequent runs will have cache hits
which are not checked by missing dep detection and will
end in a successful run even though fatal missing deps are

RB2721 was also reverted because it was a direct followup
that prevents RB2668 from being reverted cleanly.

This reverts RB 2721
Revert "Removed scan workunit; mapping workunits now debug"
This reverts commit ded93c1af9bce736550d1b0637148ab9125279a3.

This reverts RB 2668
Revert "Refactor JvmDependencyAnalyzer into a task"
This reverts commit a3c316a7d4ed9a9bdf1606b822a948be58db5eb6.

CI is green:

  1. Since this reverts so much, and presumably its only important to be reverted by the coming Friday release and before too much more happens in this space, I'd feel more comfortable waiting for Stu et. al to chime in Monday and be given the chance to propose fix or ACK revert.
    1. Knowing that this revert is going to happen blocks a couple pieces of work I have lined up for improving missing dep analysis. I don't want to work on it until I know whether I'm going to work on reverted or non-reverted code, though I can't honestly imagine this not being reverted. Which is to say, it's important for continuing work that it be reverted soon, not just before the release.

      But yes, I wouldn't expect anyone to be reading reviews on a Sunday and I intend to wait to hear what Stu and Cody think (hopefully tomorrow).

  1. Opened to expand testing and confirm that we fail stably for missing deps.

Review request changed

Status: Discarded

Change Summary:

Revert wasn't necessary