PATCH: Mauve powerpc test failures

[Mark Wielaard]

Hi,

On Fri, 2002-04-19 at 19:16, Andrew Haley wrote:
> The logic that handles longs in Java .class files is broken -- it
> assumes that a HOST_WIDE_INT is 32 bits and doesn't handle sign
> extension properly.
> 
> This patch fixes 57 test failures, no regressions.

Great work. I installed the patch on my powerpc auto-build machine and
the mauve test results are now comparable with i686.
<http://gcc.gnu.org/ml/gcc-testresults/2002-04/msg00764.html>

The only extra failures (compared with i686) are:

FAIL: gnu.testlet.java.lang.Integer.IntegerTest: Error: test_Basics
failed - 1 (number 1)

On powerpc the following statement produces false, not true.
System.out.println(Integer.MIN_VALUE == 0x80000000);
(Filed as PR java/6388)

And the following tests from java.lang.ref fail:
FAIL: gnu.testlet.java.lang.ref.PhantomReference.phantom: object
finalized (number 1)
FAIL: gnu.testlet.java.lang.ref.WeakReference.weakref: unreachable
(number 1)
FAIL: gnu.testlet.java.lang.ref.WeakReference.weakref: contents of weak
reference (number 1)

But I think we can ignore those for now since those tests are not very
reliable anyway (since they depend on when the garbage collector gets
run).

Cheers,

Mark