The attached file contains the test for the lib.equal.range algorithm.
I have one small question about the complexity of this algorithm:
The standard talks that it should be at most 2 * log (last  first) + 1.
But for some small sequences I got 2 * log (last  first) + 2 (this is
the sum of lower.bound and upper.bound "complexities").
I am not sure that this may be considered as a bug, so the test consider
2 * log (last  first) + 2 as a correct value.
I'll check this on other stl implementations (at first glance there will
be the same result).
With best wishes,
Anton Pevtsov
