Analysis of Mauve failures - Part 3

Analysis of Mauve failures - Part 3

[Mark Wielaard]

Hi,

Here is Part 3 in the Mauve failures soap.
We are almost done with the lists and then it is time for action!
But first, lets go on with the analysis.

I fixed a couple of link problems by turning the testlets into self
contained tests. But the following remains (since it uses inner
classes):
> FAIL: Link for gnu.testlet.java.util.AbstractMap.AcuniaAbstractMapTest
Have to look into the actual dejagnu script to fix this for real.

> FAIL: gnu.testlet.java.util.AbstractSequentialList.AcuniaAbstractSequentialListTest: should throw a ConcurrentModificationException -- 7 (number 1)
Same as with some other tests set(int, Object) does not make the
Iterator throw a ConcurrentModificationException.

> FAIL: gnu.testlet.java.util.Properties.AcuniaPropertiesTest: should throw NullPointerException -- 1 (number 1)
Properties.get(null) (comes from Hashtable). behavior seems not to be
defined by the spec.

> FAIL: gnu.testlet.java.util.Properties.AcuniaPropertiesTest: v does not contain:\$longvalue=I'mtryingtogiveavaluelongerthen40characters\$ (number 1)
Classpath seems to have a better/different character escape
implementation. Should be tested.

> FAIL: gnu.testlet.java.util.Properties.AcuniaPropertiesTest: should throw NullPointerException -- 1 (number 1)
Same as above get(null).

> FAIL: gnu.testlet.java.util.Properties.AcuniaPropertiesTest: all elements are the same (number 1)
> FAIL: gnu.testlet.java.util.Properties.AcuniaPropertiesTest: checkvalue -- got:\$ents=\$ (number 1)
> FAIL: gnu.testlet.java.util.Properties.AcuniaPropertiesTest: checkvalue -- got:\$dog=nocat   \$ (number 1)
> FAIL: gnu.testlet.java.util.Properties.AcuniaPropertiesTest: check if all elements were found -- got: [dog=no\\cat   ] (number 1)
Should all be tested with Classpath version. But Classpath version
throws some NullPointerException in the middle of the test...

> FAIL: gnu.testlet.java.util.Vector.AcuniaVectorTest: should throw NullPointerException (number 1)
removeAll(null) seems not to be specified by the spec. Note that it does
throw a NullPointerException when the Vector contains at least one
element.

> FAIL: gnu.testlet.java.util.Vector.AcuniaVectorTest: make sure all elements are removed (number 1)
Fallout from previous failure?

> FAIL: gnu.testlet.java.util.Vector.AcuniaVectorTest: should throw an IndexOutOfBoundsException, not IllegalArumentException -- 3 (number 1)
removeRange(3,2) Cannot find this in the spec.

> FAIL: gnu.testlet.java.util.Vector.AcuniaVectorTest: ArrayList must not be changed -- 3 (number 1)
Fallout from previous failure? Or does it actually remove things from
the Vector and then throw the IllegalArgumentException?

> FAIL: gnu.testlet.java.util.Vector.AcuniaVectorTest: should throw an IndexOutOfBoundsException, not IllegalArumentException -- 5 (number 1)
> FAIL: gnu.testlet.java.util.Vector.AcuniaVectorTest: ArrayList must not be changed -- 5 (number 1)
Same as above for removeRange(15,13)

> FAIL: gnu.testlet.java.util.Vector.AcuniaVectorTest: should throw NullPointerException (number 1)
retainAll(null) Cannot find it in the spec.

> FAIL: gnu.testlet.java.util.Vector.AcuniaVectorTest: checking new size() -- 2 (number 1)
Fallout from last failure?

> FAIL: gnu.testlet.java.util.Vector.AcuniaVectorTest: should throw a ConcurrentModificationException -- set (number 1)
The famous set(int,Object) method.

> FAIL: gnu.testlet.java.util.Vector.AcuniaVectorTest: should throw a ConcurrentModificationException -- setElementAt (number 1)
Same for setElementAt(Object,int) old Vector method.

The result of Part 1,2,3 and some of the recent fixes to libgcj and
Mauve gives us the following summary:

> 		=== libjava Summary ===
> 
> # of expected passes		12727
> # of unexpected failures	201
> # of unexpected successes	3
> # of expected failures	37
> # of untested testcases	14

There is just one more chapter to write "Analysis of ignored Mauve
tests", which will list the tests that we are currently ignoring or
marking XFAIL. But that will have to wait till tomorrow.

Good night,

Mark