I create this new bug to separate the C2 problem from 4428927.

The following JVMPI tests fail with Merlin server VM in -Xcomp mode
when compiling make_jvmpi_method_exit:

nsk/jvmpi/GetCallTrace/calltr001
nsk/jvmpi/DisableEvent/disablev001
nsk/jvmpi/EnableEvent/enablev006
nsk/jvmpi/GetMethodClass/getmethc001
nsk/jvmpi/GetThreadStatus/thrstat002
nsk/regression/b4328316

Using JDK1.4 build 57 + build 59 server VM, these tests fail with an assertion in -Xmixed mode on both Solaris Sparc and WinNT:
#
# HotSpot Virtual Machine Error, assertion failure
# Please report this error at
# http://java.sun.com/cgi-bin/bugreport.cgi
#
# assert(dead_locals_are_killed(), "garbage in debug info before safepoint")
#
# Error ID: /usr/re/hotspot/hotspot1.4/ws/solsparc/hotspot1.4beta/src/share/vm/opto/graphKit.cpp, 507 [ Patched ]
#
# Problematic Thread: prio=5 tid=0x10b248 nid=0xc runnable
#
Dumping core....
Abort (core dumped)

Using JDK1.4 build 59, some tests (such as disablev001 and b4328316) pass with -server -Xmixed mode on Solaris Sparc!! I haven't tried them on Win32. They fail with -server -Xcomp mode but got a SIGSEQV:

 >[1] Node::in(0x0, 0x0, 0x0, 0x1, 0xff33c7a0, 0x0), at 0xfded89ac
  [2] SafePointNode::control(0x0, 0x0, 0xd9584, 0xd9340, 0xd9550, 0xd7528), at 0xfded8318
  [3] GraphKit::use_exception_state(0xf0380cf0, 0x0, 0xd7528, 0x0, 0x0, 0x0), at 0xfdec865c
  [4] Parse::do_exits(0xf0380cf0, 0xd7870, 0xf03811e8, 0xff22a6c4, 0x0, 0x0), at 0xfdfdeff0
  [5] Parse::Parse(0xf0380cf0, 0x1156d4, 0xae594, 0x3f800000, 0x0, 0x0), at 0xfdfdd550
  [6] ParseGenerator::generate(0xfa52c, 0x1156d4, 0xce240, 0x0, 0x0, 0x0), at 0xfdd93304
  [7] Compile::Compile(0xf03811e8, 0xf03819b0, 0x0, 0xae594, 0xffffffff, 0x1), at 0xfde01b58
  [8] C2Compiler::compile_method(0x0, 0xf03819b0, 0x0, 0xae594, 0xffffffff, 0x0), at 0xfdd8efbc
  [9] CompileBroker::invoke_compiler_on_method(0x114c90, 0xf4410a10, 0x6, 0x0, 0x0, 0x0), at 0xfdbe8f58
  [10] CompileBroker::compiler_thread_loop(0x0, 0x5, 0x0, 0xff38e000, 0xf0381e10, 0xf0381e00), at 0xfdbe8988
  [11] compiler_thread_entry(0x114030, 0x114030, 0x1, 0x0, 0x0, 0x0), at 0xfdace58c
  [12] JavaThread::thread_main_inner(0x114030, 0x1072e0, 0x6, 0x0, 0x0, 0x0), at 0xfdac9868
  [13] JavaThread::run(0x114030, 0x0, 0x0, 0x0, 0x0, 0x0), at 0xfdac96f8
  [14] _start(0x114030, 0xff255d18, 0x0, 0x5, 0x1, 0xfe401000), at 0xfd9b96b4



Name: elR10090 Date: 05/14/2001


This bug also affects the following testbase_nsk test:
    /nsk/jvmpi/events/mentry2001

Pavel N. Visotsky, ###@###.###
======================================================================