Alpha (Re: [kaffe] Testing for 1.1.0 release)

[mcmahill at mtl.mit.edu]

I've tried out the latest (well, as of a few days ago now) CVS kaffe on
NetBSD/alpha with --with-engine=intrp and it is passing all but 3 of the
regression tests.  I'm posting them here to see if anyone has some good
ideas on how to fix.

On Wed, 28 May 2003, Dalibor Topic wrote:

> Hi Dan,
> 
> --- mcmahill at mtl.mit.edu wrote:
> > 
> > Hi Dalibor,
> > 
> > This is what I now get for the InetSocketAddressTest.
> > 
> > diff -u2 test/regression/InetSocketAddressTest.out
> > test/regression/InetSocketAddressTest.fail
> > --- test/regression/InetSocketAddressTest.out   Mon May 26 13:32:22 2003
> > +++ test/regression/InetSocketAddressTest.fail  Mon May 26 13:32:23 2003
> > @@ -4,6 +4,6 @@
> >  java.lang.IllegalArgumentException: Bad port number: -1
> >  java.lang.IllegalArgumentException: Bad port number: -128
> > -Unresolved bad.bad.bad: true
> > -Unresolved toString(): bad.bad.bad:0
> > +Unresolved bad.bad.bad: false
> > +Unresolved toString(): bad.bad.bad/216.40.33.117:0
> >  Unresolved localhost: false
> >  Resolved toString(): localhost/127.0.0.1:0

maybe this is caused by my machine not really being setup properly for
DNS? 


>  
> > My other failures are:
> > --- test/regression/DoubleComp.out      Mon May 26 13:09:58 2003
> > +++ test/regression/DoubleComp.fail     Mon May 26 13:09:59 2003
> > @@ -1,2 +1,3 @@
> >  Double tests:
> >  Float tests:
> > +assertion "eptr != NULL" failed: file "Float.c", line 84, function
> > "toCharArrayWithPrecision"
> > 
> > 
> > --- test/regression/DoublePrint.out     Mon May 26 13:09:42 2003
> > +++ test/regression/DoublePrint.fail    Mon May 26 13:09:43 2003
> > @@ -24,11 +24,3 @@
> >    1.0
> >    NaN
> > -  Infinity
> > -  -Infinity
> > -  1.4E-45
> > -  3.4028235E38
> > -  666666.56
> > -  123.75
> > -  1.0E7
> > -  -0.001
> > -  0.0/0.0 = NaN
> > +assertion "eptr != NULL" failed: file "Float.c", line 84, function
> > "toCharArrayWithPrecision"
> > 
> > this seems to be because s = "INF" and buf = "-INF" at the preceeding line
> > so the search for 'E' fails.
> 

> Infinities are supposed to have been dealt with in java code, as in Float.java:
> 
>     88    public static String toString(float value) {
>     89      // Handle exceptional values
>     90      if (isNaN(value))
>     91        return "NaN";
>     92      if (value == POSITIVE_INFINITY)
>     93        return "Infinity";
>     94      if (value == NEGATIVE_INFINITY)
>     95        return "-Infinity";
>     96
>     97      return toStringWithPrecision(value, MAX_DIGITS);
>     98    }
> 
> Since the test is run with Float.POSITIVE_INFINITY, I'd assume that line 92
> would be true, and that would be it. Something fishy seems to happen with
> floating points here. You could try looking at those float comparison routines
> again, and see if you can find what's going wrong.



> > also I'm getting (randomly) some of those 'undefined PLT symbol' errors.
> > I'm bugging some who are more in the know about netbsd's ld.elf_so about
> > that.
> 
> thanks for investigating, please report on the mailing list how it went.

there is a bug in ld.elf_so on NetBSD/alpha which has been fixed a few
months back in -current CVS but the fixes have not made it into a patch
release yet.  After patching my ld.elf_so sources, all of the 'undefined
PLT symbol' errors went away.

-Dan