Re: Math.min

[John Leuner]

> This patch looks fine to me, but as I don't have Classpath running right
> now (and won't for a bit until I get done hacking Mozilla) I can't test
> them.  Therefore it is on the patches list now, at
> https://savannah.gnu.org/patch/?group_id=85.
> 
> If someone could run this against the Muave tests right quick (John
> Leuner perhaps?) that would be wonderful :)

The patch didn't cause any improvements when used with my JVM.

skami> ~/kissme/cvs/kissme -bootclasspath 
.:/home/jewel/java/cvsclasspath/classpath:/home/jewel/kissme/cvs/classes 
gnu.testlet.SimpleTestHarness
gnu.testlet.java.lang.Math.min
FAIL: gnu.testlet.java.lang.Math.min: Double NaN comparisons (number 2)
FAIL: gnu.testlet.java.lang.Math.min: Double infinities (number 1)
FAIL: gnu.testlet.java.lang.Math.min: Double infinities (number 4)
FAIL: gnu.testlet.java.lang.Math.min: Small floats (number 7)
FAIL: gnu.testlet.java.lang.Math.min: Small floats (number 8)
FAIL: gnu.testlet.java.lang.Math.min: Float NaNs (number 1)
FAIL: gnu.testlet.java.lang.Math.min: Float NaNs (number 3)
FAIL: gnu.testlet.java.lang.Math.min: Float infinities (number 1)
FAIL: gnu.testlet.java.lang.Math.min: Float infinities (number 4)


This passes:
 harness.check (Double.toString (Math.min (2.0, Double.NaN)), "NaN");
This fails:
 harness.check (Double.toString (Math.min (Double.NaN, 2.0)), "NaN");

This prints out 2.0:
 System.err.println(Double.toString (Math.min (Double.NaN, 2.0)));

This passes:
 harness.check (Math.min (Double.NaN, 2.0), Double.NaN);
      

To verify Eric's patch we also have to ensure that my JVM is implementing the 
comparison operators correctly (especially for NaN). (It doesn't check NaN for 
floats yet).

I will check the patch in.

John Leuner