On Fri, 9 May 2014, Henrik /KaarPoSoft wrote:

> I presume that the problem is, that all three cache entries have the same
> timestamp down to microsecond level.
>
> In real life, this would probably not be a problem; who cares which cache
> entry is evicted if they are less than a microsecond apart.
>
> However, it is not nice if the test case fails (on fast systems under light
> load).
>
> Introducing a 1 millisecond sleep fixes the testcase.

Thanks!

(btw nyou added 100 microseconds which is 0.1 milliseconds)

The only little nit with the fix is that usleep isn't working on all
platforms. I think we should be able to use the first.c:select_wrapper()
function to accomplish the same thing that can work on Windows too...

-- 
  / daniel.haxx.se
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html