libjava test failures on some targets

[Kaz Kojima <kkojima at rr dot iij4u dot or dot jp>]

Hi,

There are failures on the libjava testsuites for sh4-unknown-linux-gnu

FAIL: Class_1 execution - source compiled test
FAIL: Class_1 execution - bytecode->native test
FAIL: Class_1 -O3 execution - source compiled test
FAIL: Class_1 -O3 execution - bytecode->native test
FAIL: InvokeInterface execution - bytecode->native test
FAIL: InvokeInterface -O3 execution - bytecode->native test
FAIL: PR12416 execution - bytecode->native test
FAIL: PR12416 -O3 execution - bytecode->native test
FAIL: PR260 execution - bytecode->native test
FAIL: PR260 -O3 execution - bytecode->native test
FAIL: SyncTest execution - bytecode->native test
FAIL: SyncTest -O3 execution - bytecode->native test
FAIL: invokethrow execution - bytecode->native test
FAIL: invokethrow -O3 execution - bytecode->native test
FAIL: pr179 execution - bytecode->native test
FAIL: pr179 -O3 execution - bytecode->native test
FAIL: pr8415 execution - bytecode->native test
FAIL: pr8415 -O3 execution - bytecode->native test
FAIL: stringconst2 output - source compiled test
FAIL: stringconst2 output - bytecode->native test
FAIL: stringconst2 -O3 output - source compiled test
FAIL: stringconst2 -O3 output - bytecode->native test

and it looks that the same failures have been seen for {alpha, hppa,
ia64, powerpc, s390}*-*-linux-gnu in gcc-testresults, though there is
no such failure in the testresults on {i686, x86_64}-*-linux-gnu and
powerpc-apple-darwin*.  A typical error is:

Exception in thread "main" java.lang.NoClassDefFoundError: two
   at InvokeInterface.class$ (Throwable.java:37)
   at InvokeInterface.main (InvokeInterface.java:42)

on sh4-unknown-linux-gnu.  I did a binary search and it seems that
they've started to fail after the patch in

http://gcc.gnu.org/ml/java-patches/2005-q2/msg00368.html

and if it's reverted with adding #if 0 to natClassLoader.cc:
_Jv_RegisterClassHookDefault:

void
_Jv_RegisterClassHookDefault (jclass klass)
{
  // This is bogus, but there doesn't seem to be a better place to do
  // it.
  if (! klass->engine)
    klass->engine = &_Jv_soleCompiledEngine;

#if 0
  if (system_class_list != SYSTEM_LOADER_INITIALIZED)
    {
      unsigned long abi = (unsigned long) klass->next_or_version;
      if (! _Jv_ClassForBootstrapLoader (abi))
	{
	  klass->next_or_version = system_class_list;
	  system_class_list = klass;
	  return;
	}
    }
#endif

then the above testcases pass.  I don't understand what is going on
here, though.
BTW, I couldn't find any PR for it in bugzilla.  Should I file it?

Regards,
	kaz