[m-rev.] for review: simplify profiler feedback code

[Zoltan Somogyi]

For review by Paul.

Paul, I would like you to check the following in particular:

- Should the calls to add_feedback_candidate_parallel_conjunctions be
  calls to replace_feedback_candidate_parallel_conjunctions instead?
  Under what circumstances did you plan to have existing feedback
  information simply replaced by new information of the same kind?

- The old code had two sources for the name of the program that created
  the information in a feedback_info: the one recorded in the feedback_info
  itself, and the caller of whatever predicate needed that program name.
  Why was the second source added, and could the two sources ever disagree?
  If they can disagree, why? If they cannot, I will delete the extra
  ProgramName args. In both cases, I will delete the assertions on them.

  I note that those assertions did not fail during a bootcheck, but the
  bootcheck may not have exercised the full functionality of the feedback
  mechanism.

- Do you have any plans for the use of the now-commented-out feedback type?

I hope you had a good long weekend, and that the upcoming procedure on
your eye is successful.

Zoltan.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Log.feedback
Type: application/octet-stream
Size: 1296 bytes
Desc: not available
URL: <http://lists.mercurylang.org/archives/reviews/attachments/20141104/b2fa4a1d/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: DIFF.feedback
Type: application/octet-stream
Size: 40846 bytes
Desc: not available
URL: <http://lists.mercurylang.org/archives/reviews/attachments/20141104/b2fa4a1d/attachment-0001.obj>