[D-runtime] [D-Programming-Language/druntime] d7ab02: Add an OS variable to posix.mak and disable core/t...

[Brad Roberts]

On 2/20/2011 2:49 AM, Brad Roberts wrote:
> On 2/20/2011 2:29 AM, Jonathan M Davis wrote:
>> On Sunday 20 February 2011 01:39:50 noreply at github.com wrote:
>>> Branch: refs/heads/master
>>> Home:   https://github.com/D-Programming-Language/druntime
>>>
>>> Commit: d7ab02b7be3e172352855daa05edd5ce4ddf9e43
>>>    
>>> https://github.com/D-Programming-Language/druntime/commit/d7ab02b7be3e1723
>>> 52855daa05edd5ce4ddf9e43 Author: Brad Roberts <braddr at puremagic.com>
>>> Date:   2011-02-19 (Sat, 19 Feb 2011)
>>>
>>> Changed paths:
>>>   M posix.mak
>>>
>>> Log Message:
>>> -----------
>>> Add an OS variable to posix.mak and disable core/time on freebsd until it
>>> can be fixed
>>
>> What's broken about core.time on FreeBSD?
>>
>> - Jonathan M Davis
> 
> One of the unit tests is calculating a slightly wrong value.  I only spent a few minutes digging into it and opted to
> disable and move on for the moment.  I'll dig out the relevant info once I get the auto-tester up and going.

See my most recent comment to bug 5629:

in TickDuration's static this.  freebsd has clock_gettime, so it uses the
clock_getres api.  The results:

    ts.tv_nsec = 280, ticksPerSec = 3571428

Later, during the unit test:

            auto t = TickDuration.from!"nsecs"(1_000_000_000);
            assert(t.nsecs == 1_000_000_000);

t.nsecs is slightly less: t.nsecs = 999999840

280 * 3571428 == 999999840
280 * 3571429 == 1000000120

So, it's a matter of precision.