java/1204: Definite assignment problem

[green@cygnus.com]

>Number:         1204
>Category:       java
>Synopsis:       Definite assignment problem
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    apbianco
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Dec 20 12:17:32 PST 2000
>Closed-Date:    
>Last-Modified:  Thu Apr 13 17:56:57 PDT 2000
>Originator:     Anthony Green
>Release:        Latest cvs
>Organization:
>Environment:
Red Hat Linux 6.1
>Description:
The compiler is not recognizing a situation
where a variable is guaranteed to be assigned to,
and is exiting with a definite assignment error.

This bug occurs with for local variables assigned within 
either a try/finally or try/catch block.
>How-To-Repeat:
I've checked in a test case: libjava.compile/assignment.java
Also see libjava.compile/assignment_2.java
>Fix:

>Release-Note:

>Audit-Trail:

Formerly PR gcj/148


From: Bryce McKinlay <bryce@albatross.co.nz>
To: green@cygnus.com, java-gnats@sourceware.cygnus.com
Cc:  
Subject: Re: gcj/148: Definite assignment problem
Date: Wed, 16 Feb 2000 22:13:31 +1300

 Somewhat related - I'm getting the "Variable may not have been
 initialized" message from the latest compiler from cvs in totally bogus
 situations that used to work fine. This bug seems to have appeared
 recently (certainly since 19991213 which is the second-most-recent build
 I have lying around)
 
 In my case I have
 
 String var_value = line.substring(...);
 Variable v = new Variable(first_token, var_value.trim(),  ...);
 
 and the compiler is telling me that var_value may not have been
 initialized. I'm going to try and come up with a test case...
 
   [ bryce ]
 
 
 green@cygnus.com wrote:
 
 > >Number:         148
 > >Category:       gcj
 > >Synopsis:       Definite assignment problem
 > >Confidential:   no
 > >Severity:       serious
 > >Priority:       medium
 > >Responsible:    apbianco
 > >State:          open
 > >Class:          sw-bug
 > >Submitter-Id:   net
 > >Arrival-Date:   Sun Feb 13 09:40:02 PST 2000
 > >Closed-Date:
 > >Last-Modified:
 > >Originator:     Anthony Green
 > >Release:        Latest cvs
 > >Organization:
 > >Environment:
 > Red Hat Linux 6.1
 > >Description:
 > The compiler is not recognizing a situation
 > where a variable is guaranteed to be assigned to,
 > and is exiting with a definite assignment error.
 > >How-To-Repeat:
 > I've checked in a test case: libjava.compile/assignment.java
 > >Fix:
 >
 > >Release-Note:
 > >Audit-Trail:
 > >Unformatted:
 

From: Bryce McKinlay <bryce@albatross.co.nz>
To: java-gnats@sourceware.cygnus.com
Cc:  
Subject: Re: gcj/148: Definite assignment problem
Date: Wed, 16 Feb 2000 23:27:24 +1300

 Here's my test case for this strange bug:
 
 interface I
 {
   int huk = 5;
 }
 
 public class DA implements I
 {
   DA(String s) {}
 
   private void tmp(String line)
   {
     String broken = line.substring(huk);
     DA da = new DA(broken);
   }
 }
 
 produces:
 
 DA.java: In class `DA':
 DA.java: In method `tmp(java.lang.String)':
 DA.java:13: Variable `broken' may not have been initialized.
     DA da = new DA(broken);
 
 Both the static variable being declared in an interface and calling a
 constructor with the result appear to be significant. Works fine in 2.95.2
 & 19991213. Broken on 20000215. Works fine from .class file.
 
   [ bryce ]
 
 
 
 

From: Bryce McKinlay <bryce@albatross.co.nz>
To: apbianco@cygnus.com, java-gnats@sourceware.cygnus.com
Cc:  
Subject: Re: gcj/148: Definite assignment problem
Date: Thu, 17 Feb 2000 00:37:52 +1300

 I narrowed the cause of this down to Anthony's Class Initialization
 Optimization patch. Removing the patch makes the problem go away.
 
   [ bryce ]
 
 

From: Anthony Green <green@cygnus.com>
To: bryce@albatross.co.nz
Cc: java-gnats@sourceware.cygnus.com, apbianco@cygnus.com
Subject: gcj/148: Definite assignment problem
Date: Wed, 16 Feb 2000 09:48:02 -0800

 Hi Bryce -
 
 > I narrowed the cause of this down to Anthony's Class Initialization
 > Optimization patch. Removing the patch makes the problem go away.
 
 Are you saying the toolchain you found the problem in was 20000215+my
 patch?  I don't think I committed it (although I was so tired at the
 end, maybe I did by mistake?)
 
 Thanks,
 
 AG
 
 -- 
 Anthony Green                                                        Red Hat
                                                        Sunnyvale, California

From: bryce@sourceware.cygnus.com
To: java-gnats@sourceware.cygnus.com
Cc:  
Subject: gcj/148
Date: 14 Apr 2000 00:47:35 -0000

 CVSROOT:	/cvs/java
 Module name:	libgcj
 Changes by:	bryce@sourceware.cygnus.com	00/04/13 17:47:35
 
 Modified files:
 	libjava/testsuite: ChangeLog 
 Added files:
 	libjava/testsuite/libjava.compile: assignment_2.java 
 
 Log message:
 	2000-04-14  Bryce McKinlay  <bryce@albatross.co.nz>
 	
 	Additional test for PR gcj/148:
 	* libjava.compile/assignment_2.java: New file.
 
 Patches:
 http://gcc.gnu.org/cgi-bin/gcc/libjava/testsuite/ChangeLog.diff?cvsroot=java&r1=1.80&r2=1.81
 http://gcc.gnu.org/cgi-bin/gcc/libjava/testsuite/libjava.compile/assignment_2.java.diff?cvsroot=java&r1=NONE&r2=1.1
 
>Unformatted: