[kaffe] Bug Report: accessing array doesn't throw NullPointerException under MIPS

Kevin D. Kissell kevink at mips.com
Wed Mar 12 01:40:01 PST 2003

Dalibor writes:
> --- "Kevin D. Kissell" <kevink at mips.com> wrote:
> > And as an FYI to developers, I don't know what's
> > been changed elsewhere
> > in the tree that would account for it, but while I
> > can build and debug a JIT3
> > version of kaffe-1.0.7+patch, I cannot get "make" to
> > complete for a JIT3
> > kaffe using the current CVS sources - it blows out
> > during the compilation
> > of the Java class libraries, complaining about
> > floating point values on the
> > order of 1-point-something being "too small".
> I've synced up with the JanosVM 0.8 jit3
> implementation on Tue Feb 11 16:54:43 2003 UTC (4
> weeks ago), I though that might have caused the
> problems. I'm seeing the same problems on irix, by the
> way, and as I've reported them back on the 3rd
> February, it can't be that patch to blame.
> http://www.kaffe.org/pipermail/kaffe/2003-February/028787.html
> Looks like it's binary search build time again ;)

Note that the problems you reported on Feb 3 were not
the same as those that I'm seeing with the MIPS/Linux
platform.  In your case, you had 3 regression failures.
I can't even get the build to complete, due to failures
building the class libraries.  I supposed I could build
the libraries using an interpreted kaffe and go back to
JIT for the regression tests, but for the moment, my
hypothesis is that, if I finally figure out what's wrong
with MIPS/jit3 in 1.0.7, that same fix will also improve
the behavior with CVS sources.

It's been hard slogging trying to identify the current
problem.  Since some of the failing regression tests were
among those which failed under intrp before I fixed the
argument-passing code in o32-sysdepCallMethod.h,
I had really expected to find the same conceptual error
(the calling convention *is* baroque and poorly documented)
in the JIT code, but so far, it looks to have been done 
correctly.  I'll keep pounding at it, and will send a patch
to the mailing list if I come up with something, but for
the moment, all I've managed to do is fix some erroneous
debug trace output and get rid of some preprocessor 
whinings about trailing text in #undefs.

            Kevin K.

More information about the kaffe mailing list