Bug#219102: libc6: IBM JDK 1.3.1 core dumps

[Peter Nagy <peter.1.nagy@nokia.com>]

On Tue, 2003-11-04 at 15:20, Daniel Jacobowitz wrote:
> > Or do you mean that this is a bug that can't be fixed because it's
> > present in the JDK via static linking or simliar, and not in glibc?
> 
> That's correct.  It's a problem in the JDK, and one of the changes in
> later versions was to fix it.  Sorry I wasn't clear.

I've downloaded the latest 1.3 JRE from IBM, and it still segfaults at
JVM startup. It works fine with libc6 2.3.2-9, it crashes with
2.3.2.ds1-9.

VM Version:
Classic VM (build 1.3.1, J2RE 1.3.1 IBM build cxia32131-20030618 (JIT
enabled: jitc))

And stack trace:
#0  0x4034ce39 in allocGuardPage () at
/userlvl/cxia32131/obj/x86_linux_2/jvm/sov/xe/mmi/mmi_invokers.s:2279
#1  0x4034d1fd in xeThreadInit () at
/userlvl/cxia32131/obj/x86_linux_2/jvm/sov/xe/mmi/mmi_invokers.s:2279
#2  0x4034edde in eeInitNewThis () at
/userlvl/cxia32131/obj/x86_linux_2/jvm/sov/xe/mmi/mmi_invokers.s:2279
#3  0x40354146 in xmInitializeJVM () at
/userlvl/cxia32131/obj/x86_linux_2/jvm/sov/xe/mmi/mmi_invokers.s:2279
#4  0x40265d69 in ciCreateJVM () from
/home/pnagy/newroot/opt/IBMJava2-131/jre/bin/classic/libjvm.so
#5  0x40270136 in JNI_CreateJavaVM () from
/home/pnagy/newroot/opt/IBMJava2-131/jre/bin/classic/libjvm.so
#6  0x080499ae in InitializeJVM ()
#7  0x08049038 in main ()

I've found some reports that this JVM also crashes on RedHat 9 with the
same stacktrace. It does tell a workaround with LD_ASSUME_KERNEL=2.2.5,
which forces glibc to use the old pthreads instead of NPTL.
http://www.mainframeforum.com/t584476.html

This workaround doesn't work on my debian, though :-)

Actually, the IBM java startup scripts _do_ contain code that sets
LD_ASSUME_KERNEL. It just doesn't work for us.