--explain-call to report multiple best candidates

[vasslitvinov]

This PR makes --explain-call option show all "best" candidates being considered in the case of return intent overloads. This may be useful when the user is expecting one overload to be invoked and does not realize that another overload is also being considered because of return intent overload.

Testing: standard and gasnet paratests.

[mppf]

Looks reasonable to me. One request: could have adjust it to print "best candidate is" (as it did before) if return intent overloading is not relevant (i.e., when only one of bestRef, bestValue, bestConstRef is not nullptr)?

[vasslitvinov]

@mppf -

when only one of bestRef, bestValue, bestConstRef is not nullptr

My understanding is that in this case the function where my change is located is not invoked, instead the 1-candidate function is:

  if (numMatches == 0) {
    ....
  } else if (numMatches == 1) {
    ResolutionCandidate* best = NULL;

    if        (bestRef  != NULL) {
      best = bestRef;

    } else if (bestVal  != NULL) {
      best = bestVal;

    } else if (bestCref != NULL) {
      best = bestCref;
    }

    retval = resolveNormalCall(info, checkState, best);

  } else {
    // >>> my change is here <<<
    retval = resolveNormalCall(info, checkState, bestRef, bestCref, bestVal);
  }

What do you think?

[mppf]

That's fine; please just verify that the behavior is the way we expect when return intent overloading is used (it prints out the best candidate, not calling it ref or anything). We might have a test that already checks this (I did not check).