> On Jun 24, 2018, at 1:29 PM, Doug Lea <d...@cs.oswego.edu> wrote:
>
> On 06/22/2018 01:28 PM, Paul Sandoz wrote:
>
>>> 8203864: Execution error in Java's Timsort
>>> http://cr.openjdk.java.net/~martin/webrevs/jdk/jsr166-integration/TimSort/index.html
>>> https://bugs.openjdk.java.net/browse/JDK-8203864
>>>
>>
>> 406 if (n > 0 && runLen[n-1] <= runLen[n] + runLen[n+1] ||
>> 407 n > 1 && runLen[n-2] <= runLen[n] + runLen[n-1]) {
>> 408 if (runLen[n - 1] < runLen[n + 1])
>> 409 n--;
>>
>> Minor comment, would be good if consistent style is used for the index
>> expressions. I have a marginal preference to retaining ordering from a
>> “picture this in my head" kind of thing, specifically:
>>
>> runLen[n - 2] <= runLen[n - 1] + runLen[n]
>
> I marginally agree, but I'm keeping it this way just for boring
> consistency with published versions.
>
Ok, consistency overrules!
Paul.
