Re: Testing monotonic clock

Matthew Dillon wrote:
> I'm sure you will come up with a solution, even if it kills you. 
> My one consolation is that I am putting it on your shoulders to solve
> :-)

I hate when it happens ;).

OK, I ended up bringing in some changes from FreeBSD - I verified these 
and these make perfect sense. I haven't verified yet what it breaks in 
our tree :).

The source of the problem seems to be the stuff we have in sys/_posix.h. 
Besides being buggy, using these defines doesn't make sense IMHO. 
Volunteers to review (and kill?) it?

* There is more POSIX standards than 1003.1b-1993 which should be used in 
  ifdefs.
* Using '#ifndef _ANSI_SOURCE' doesn't make sense if __POSIX_VISIBLE is 
  already used in ifdefs - __POSIX_VISIBLE is 0 if _ANSI_SOURCE is defined 
  (sys/cdefs.h).

The patch is attached.


PS. In progress I realized that the configure test which started this is 
buggy as well ... ARGH!

-- 
Hasso Tepper