These are chat archives for HdrHistogram/HdrHistogram

17th
Nov 2015
Gil Tene
@giltene
Nov 17 2015 14:39 UTC
@mikeb01 to the "is this big specific to IntHistogram question: assuming you are asking about the linear and logarithmic iterator edge bug: no. It's generic to iteration, and appears when the range of values covered by a single iteration step includes the highest recorded value, as well as a (higher) value whose index would exceed the current counts array length. For this to happen, a value would need to be recorded close enough to the edge of the array to be included in such a step (otherwise the iteration would terminate before reaching the edge and won't trigger the bug). The bug has always been around, but is much more likely to be hit by auto sizing histograms, since pre-sized histograms tend to be oversized and don't (tend to) have any values recorded close to the top of their actual array-covered range.
Michael Barker
@mikeb01
Nov 17 2015 20:09 UTC
@giltene Cheers, I'll modify the unit test to see if I can recreate the issue.
Michael Barker
@mikeb01
Nov 17 2015 20:35 UTC
Looks like there is an issue on the C version, doesn't crash, but omits the last block of the linear iterator.