[Bug classpath/22783] wrong parsing behaviour of NaN, Infinity, -Infinity

[gcc-bugzilla at gcc dot gnu dot org]

Classpath (CVS HEAD 09.25.04) throws a NumberFormatException although
[+/-]NaN|Infinity are possible float/double values according to Java
documentation.


------- Comment #1 from from-classpath at savannah dot gnu dot org  2004-10-07 
14:29 -------
Simple test cases:

try {
  if(Double.POSITIVE_INFINITY == Double.parseDouble("Infinity")) {
    // OK
  } else {
    // WRONG
  }
} catch(NumberFormatException nfe){
  // should not happen (but is what its doing now)
}

aside from the given comparison it can be replaced with:
Double.POSITIVE_INFINITY == Double.parseDouble("+Infinity")
Double.NEGATIVE_INFINITY == Double.parseDouble("-Infinity")
Double.NaN == Double.parseDouble("NaN");

Float.POSITIVE_INFINITY == Float.parseFloat("+Infinity")
Float.NEGATIVE_INFINITY == Float.parseFloat("-Infinity")
Float.NaN == Float.parseFloat("NaN");

for all examples the string may contain whitespace before and after the 'word'.


------- Comment #2 from from-classpath at savannah dot gnu dot org  2004-10-07 
16:05 -------
I've written some Mauve tests that cover these failing cases (plus some other
non-failing cases).  I'll post them to the mauve-patches mailing list now.


------- Comment #3 from from-classpath at savannah dot gnu dot org  2004-10-07 
22:42 -------
I've submitted a quick-fix for this to the classpath-patches list. 


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22783