This test was intended only to test a quirk on Solaris and be run only there.
If run on macOS 10.13 it fails there.

The test looks quirky but it just tries to confirm that text renders the same whether it is the first text ever rendered or not.

On 10.13 it to me as if the first time through we get AA glyphs when expecting
B&W.
These then end up very blocky because we render them as B&W
In a 2nd VM, we render one string, throw it away (erase it)
and render the same as we did the other VM. This time we get B&W glyphs.

This is on 10.13 so its not exactly the same as https://bugs.openjdk.java.net/browse/JDK-8220150
but the result is the similar. Its not the same because subsequently we *do*
 get B&W glyphs

So it looks like on first rendering we are getting (unexpectedly) AA glyphs until
something runs and sets the pipeline up properly.

This is likely an old problem since rendering to the screen or Swing backbuffer
we always use OGL where we treat all glyphs as AA.

Also since 10.14 and above always return AA glyphs we need to treat them
as such in software rendering loops, as are used here.

So the root cause is probably irrelevant except on 10.13 and below.