From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <java-prs-return-7223-listarch-java-prs=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 14265 invoked by alias); 18 Nov 2005 23:18:40 -0000
Received: (qmail 14247 invoked by uid 48); 18 Nov 2005 23:18:39 -0000
Date: Fri, 18 Nov 2005 23:18:00 -0000
Subject: [Bug java/24938]  New: gcj -C miscompiles gnu/CORBA/DynAn/gnuDynValue.java
X-Bugzilla-Reason: CC
Message-ID: <bug-24938-3639@http.gcc.gnu.org/bugzilla/>
Reply-To: gcc-bugzilla@gcc.gnu.org
To: java-prs@gcc.gnu.org
From: "mark at gcc dot gnu dot org" <gcc-bugzilla@gcc.gnu.org>
Mailing-List: contact java-prs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Subscribe: <mailto:java-prs-subscribe@gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/java-prs/>
List-Post: <mailto:java-prs@gcc.gnu.org>
List-Help: <mailto:java-prs-help@gcc.gnu.org>, <http://gcc.gnu.org/ml/#faqs>
Sender: java-prs-owner@gcc.gnu.org
X-SW-Source: 2005-q4/txt/msg00365.txt.bz2
List-Id: <java-prs.sourceware.org>

There are two pieces of code in gnuDynValue that generate unverifiable (and
wrong!) code.

The following method in this file:

  public int component_count()
  {
    return isNull ? 0 : super.component_count();
  }

Generates the following byte code:

Method name:"component_count" public Signature: ()int
Attribute "Code", length:38, max_stack:2, max_locals:1, code_length:14
  0: aload_0
  1: getfield <Field gnu.CORBA.DynAn.gnuDynValue.isNull boolean>
  4: ifeq 9
  7: iconst_0
  8: ireturn
  9: aload_0
 10: invokespecial <InterfaceMethod org.omg.DynamicAny.DynAny.component_count
()int>
 13: ireturn
Attribute "LineNumberTable", length:6, count: 1
  line: 369 at pc: 0

PC 10 is wrong. And the verifier actually catches that:

gnu/CORBA/DynAn/gnuDynValue.java: In class 'gnu.CORBA.DynAn.gnuDynValue':
gnu/CORBA/DynAn/gnuDynValue.java: In method
'gnu.CORBA.DynAn.gnuDynValue.component_count()':
gnu/CORBA/DynAn/gnuDynValue.java:369: error: verification failed at PC=10:
didn't see expected constant

Another method in this file that exhibits this is:

  public boolean equal(DynAny other)
  {
    if (other instanceof DynValueOperations)
      {
        DynValueCommon o = (DynValueCommon) other;
        if (isNull)
          return o.is_null() && o.type().equal(official_type);
        else
          return !o.is_null() && super.equal(other);
      }
    else
      return false;
  }

That last super.equal() call generates the same wrong code.

This might be an existing bug, but I couldn't quickly find it.
I have a workaround for this issue that I want to get in before reducing this.


-- 
           Summary: gcj -C miscompiles gnu/CORBA/DynAn/gnuDynValue.java
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Keywords: wrong-code
          Severity: normal
          Priority: P3
         Component: java
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: mark at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24938