hrtime regression

[BridgeAR]

I have the feeling 89f056b introduced a regression. The nanoseconds might return negative values if you compare the actual value with the one before. This does not occur in 5.3 or before.

> process.hrtime(t)
[ 45, 373756947 ]
> process.hrtime(t)
[ 46, -1831860 ]

Easy to reproduce with:

var t = process.hrtime(); while(t[1] > 0) t = process.hrtime(t);

[jasnell]

@Fishrock123 @trevnorris

[ChALkeR]

I can reproduce this.

[ChALkeR]

It does not actually mean negative values, it looks to me that the values are just not normalized.
I.e. it returns [1, -692093386] when it should return [0, 307906614].

[ChALkeR]

89f056b#diff-6ff379484cbabad48301d485db111c08R198does:
ret[1] = hrValues[2] - ar[1]; does not take into an account that hrValues[2] could be less than ar[1].

[ChALkeR]

The simplified fast test would be: process.hrtime([0, 999999999])[1] >= 0.

[mscdex]

I also noticed this recently and was really confused.

[trevnorris]

My bad. I'll write up a patch and additional test case for this ASAP.