java crash on amd64 when switching to java perspective in eclipse

[Sean McNeil]

On Mon, 2006-02-06 at 17:24 +0200, Panagiotis Astithas wrote:
> Sean McNeil wrote:
> > Got the following:
> > 
> > #
> > # An unexpected error has been detected by HotSpot Virtual Machine:
> > #
> > #  SIGSEGV (0xb) at pc=0x00000008055aa9df, pid=56131, tid=0x133d000
> > #
> > # Java VM: Java HotSpot(TM) 64-Bit Server VM
> > (1.5.0-p2-root_02_feb_2006_13_05 mixed mode)
> > # Problematic frame:
> > # J
> > org.eclipse.jdt.internal.compiler.parser.Scanner.jumpOverMethodBody()V
> > #
> > # An error report file with more information is saved as
> > hs_err_pid56131.log
> > #
> > # If you would like to submit a bug report, please write
> > # a letter to freebsd-java at FreeBSD.org mailing list
> 
> I did some research. There is a comment with a workaround at the bottom 
> of this page:
> 
> http://wbeaton.blogspot.com/2005/09/eclipse-in-64-bits.html
> 
> A more thorough discussion is here:
> 
> https://bugs.eclipse.org/bugs/show_bug.cgi?id=71987
> 
> The relevant JDK bug entry is here:
> 
> http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=5060628
> 
> The Sun engineer's fix is described here:
> 
> http://blogs.sun.com/roller/page/javawithjiva/20051031
> 
> It is reported to be fixed in jdk 1.5 update 6 and jdk 1.6 build 59.

Great!  I've read this information and guess-timated where the issue
with REX and NEGL happens to be.  With a simple little patch I've
managed to eliminate my repeatable crashes.  The system seems much much
more stable now.

--- hotspot/src/cpu/amd64/vm/amd64.ad.orig      Wed Feb  8 15:18:24 2006
+++ hotspot/src/cpu/amd64/vm/amd64.ad   Wed Feb  8 15:17:35 2006
@@ -7246,7 +7246,7 @@

   format %{ "negl    $dst\t# int" %}
   opcode(0xF7, 0x03);  // Opcode F7 /3
-  ins_encode(REX_reg(dst), OpcP, RM_opc_mem(secondary, dst));
+  ins_encode(REX_mem(dst), OpcP, RM_opc_mem(secondary, dst));
   ins_pipe(ialu_reg);
 %}