[LEAPSECS] earth speeding up

[Warner Losh]

On Apr 15, 2014, at 7:50 AM, Steve Allen <sla at ucolick.org> wrote:


> On Tue 2014-04-15T09:42:20 -0400, Joseph Gwinn hath writ:

>> This first negative leap second may end civilization - essentially no

>> leap-second handling code is really ready for a step backwards.

> 

> I think not.  I think many of the computing systems which fail for

> positive leaps are ready for negative leaps.

> 

> A negative leap is saying "I missed experiencing some time".  Any time

> sharing system should find it routine that a process was asleep while

> some time elapsed, and then it wakes up later.


Given the Linux exemplar of what can happen when there’s positive
leap seconds that software has dealt with many times before, I
suspect the capacity to foul up the implementation is not bounded
by what’s reasonable in a naive implementation. This is effectively
untested code, and untested code tends to be broken code.

That said, most naive implementations that don’t try anything too
much fancier than ts->second++ should fair fairly well.

Warner