[phobos] periodic test failures + auto-testing

[Jonathan M Davis]

On Monday, December 19, 2011 10:00:21 Brad Roberts wrote:
> I still believe you're testing more than is useful.  What kinds of bugs in
> this D code can these tests find?  So far, they're far more pain than
> benefit.  You're not going to find issues with time not moving forward, and
> even if you do, what would you do about it?  You might find bugs in the
> underlying OS's dealing with file time metadata.  Assuming you do, what
> would you do about it?  Add to that my loathing for tests that require
> sleeps.  Just get rid of them.  At best, disable them but keep the code on
> the extremely remote chance someone reports a bug in the area and you can
> point them at the unittest to help debug.

Okay. For the moment, I've increased the leeway and versioned off some of the 
tests, so hopefully that's enough to fix the autotester for now.

I think what I need to do in the long run is refactor and combine the tests so 
that they're more condensed and don't require as many sleeps and don't need to 
do as much writing to and reading from disk (and general reduce code 
duplication in the tests). Also, it looks like it might be possible to make 
the file times more precise, which would allow the tests to be more precise and 
allow the sleeps to be further reduced. So, obviously I have something more 
for my todo list.

- Jonathan M Davis