On Tue, Apr 04, 2017 at 09:32:19PM +0100, James Hogan wrote: > I think the correct fix is to prevent the read rather than change the > size of the array. buf2[] is intentionally 3 so that out of 5 sorted > samples the last element is the median, whereas buf1 is 4 elements so as > to work out the 75th percentile. > > When inserting the 5th sample into buf1 (i.e. j = 4), there are already > 4 entries, so the highest element it needs to compare against is the 4th > one (buf1[k=3]), so thats fine. > > For buf2 however its still trying to insert 5 elements, so by the 5th > one (i.e. i = 4) it may try to compare against the 4th element to know > whether to insert before it, at which point we simply don't care about > the ordering as its past the median. > > So I think something like this would be more correct. Does that fix your > problem? Yes, this fixes it too. Please feel free to submit it yourself with your explanation. You can add my Tested-by if you like. Thanks.
