Bug ID: JDK-4649030 sparc v9 fails CTW of rt.jar, at 6783

Type: Bug
Component: hotspot
Sub-Component: compiler
Affected Version: 1.4.1

Priority: P2
Status: Closed
Resolution: Duplicate
OS: generic
CPU: sparc

Submitted: 2002-03-07
Updated: 2002-03-15
Resolved: 2002-03-15

producer:cliffc/tests> J C2 fg v9 -Xbatch -Xcomp -XX:+CompileTheWorld -XX:CompileTheWorldStartAt=6783 -XX:+PrintOpto -XX
:CompileOnly=.createNativeCursor
VM option '+CompileTheWorld'
VM option 'CompileTheWorldStartAt=6783'
VM option '+PrintOpto'
VM option 'CompileOnly=.createNativeCursor'
CompileTheWorld : Compiling all classes in /net/jaberwocky/export/cliffc/j2sdk1.4.0/jre/lib/rt.jar

default to Motif 2.1, os is: 5.8
CompileTheWorld (6783) : sun/awt/motif/X11CustomCursor
  1  sun.awt.motif.X11CustomCursor::createNativeCursor (860 bytes)
#
# HotSpot Virtual Machine Error, assertion failure
# Please report this error at
# http://java.sun.com/cgi-bin/bugreport.cgi
#
# Java VM: Java HotSpot(TM) 64-Bit Server VM (1.4-internal-debug compiled mode)
#
# assert(_base == AryPtr, "Not an array pointer")
#
# Error ID: /net/jaberwocky/export/cliffc/C2/src/share/vm/opto/type.hpp, 923 [ Patched ]
#
# Problematic Thread: prio=5 tid=0x100236418 nid=0xb runnable
#
Dumping core....
Abort

###@###.### 2002-03-07

EVALUATION I chased it for awhile. In bytecode parsing, at block #30, bytecodes 564-571 there's a backedge which adds a safepoint. We trim to live locals only and the methodLiveness info reports local #2 is dead. Later we parse bytecode 677 which uses local #2. Yes there is a path from bytecode 571 to 677. There's some looping going on, but it's all well structured, no exception handlers, no re-use of local #2 (its an incoming argument). In short, it looks like a bug in methodLiveness that I've stumbled over. Since I'm full up with other bugs and this one doesn't look like one I've introduced I'm not going to fix it with this current putback. ###@###.### 2002-03-07

07-03-2002