Re: java/4230: gcj -C segfaults on static string continuation '+' in gcj 3.1 expermental.

Re: java/4230: gcj -C segfaults on static string continuation '+' in gcj 3.1 expermental.

[Alexandre Petit-Bianco]

The following reply was made to PR java/4230; it has been noted by GNATS.

From: Alexandre Petit-Bianco <apbianco@cygnus.com>
To: Tony Knaus <awk@panic.spinnakernet.com>
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: java/4230: gcj -C segfaults on static string continuation '+' in gcj 3.1 expermental.
Date: Wed, 5 Sep 2001 20:59:07 -0700 (PDT)

 Tony Knaus writes:
 > gcj 3.1 generates a segmentation fault while building the xerces
 > parser and generating byte code. The compiler does not seg fault
 > when compiling to object code or just generating dependencies.
 
 Hmm, sorry about that. It appeared recently.
 
 > public class StringCrash {
 >     static final String viramaString = 
 >     "f"
 >     +"o"
 >     +"o"
 >     +"b"
 >     +"a"
 >     +"r";
 > 
 >     public static void main(String[] args) {
 > 	System.out.println("viramaString= " + viramaString);
 >     }
 > };
 
 
 Thank you for the test case. After a several hours in build hell, I'm
 testing this patch.
 
 ./A
 
 2001-09-05 Alexandre Petit-Bianco  <apbianco@redhat.com>
 
 	* jcf-write.c (generate_classfile): Issue an error in case of
 	field/initial value mismatch.
 	* parse.y (analyze_clinit_body): Keep <clinit> if an array is
 	being initialized and we're generating bytecode.
 	(java_complete_lhs): In CASE_EXPR section: added comments,
 	set DECL_INITIAL properly when appropriate.
 
 Index: jcf-write.c
 ===================================================================
 RCS file: /cvs/gcc/gcc/gcc/java/jcf-write.c,v
 retrieving revision 1.87
 diff -u -p -r1.87 jcf-write.c
 --- jcf-write.c	2001/08/31 04:14:43	1.87
 +++ jcf-write.c	2001/09/06 03:51:55
 @@ -2899,8 +2899,8 @@ generate_classfile (clas, state)
  	{
  	  tree init = DECL_INITIAL (part);
  	  static tree ConstantValue_node = NULL_TREE;
 -	  // This conversion is a work-around for front-end bug.
 -	  init = convert (TREE_TYPE (part), init);
 +	  if (TREE_TYPE (part) != TREE_TYPE (init))
 +	    fatal_error ("field initializer type mismatch.");
  	  ptr = append_chunk (NULL, 8, state);
  	  if (ConstantValue_node == NULL_TREE)
  	    ConstantValue_node = get_identifier ("ConstantValue");
 Index: parse.y
 ===================================================================
 RCS file: /cvs/gcc/gcc/gcc/java/parse.y,v
 retrieving revision 1.307
 diff -u -p -r1.307 parse.y
 --- parse.y	2001/09/04 21:50:31	1.307
 +++ parse.y	2001/09/06 03:52:04
 @@ -7813,6 +7813,11 @@ analyze_clinit_body (bbody)
  	break;
  	
        case MODIFY_EXPR:
 +	/* If we're generating to class file and we're dealing with an
 +	   array initialization, we return 1 to keep <clinit> */
 +	if (TREE_CODE (TREE_OPERAND (bbody, 1)) == NEW_ARRAY_INIT
 +	    && flag_emit_class_files)
 +	  return 1;
  	/* Return 0 if the operand is constant, 1 otherwise.  */
  	return ! TREE_CONSTANT (TREE_OPERAND (bbody, 1));
  
 @@ -11898,16 +11903,31 @@ java_complete_lhs (node)
  	  
  	  value = fold_constant_for_init (nn, nn);
  
 +	  /* When we have a primitype type, or a string and we're not
 +             emitting a class file, we actually don't want to generate
 +             anything for the assignment. */
  	  if (value != NULL_TREE &&
  	      (JPRIMITIVE_TYPE_P (TREE_TYPE (value)) || 
  	       (TREE_TYPE (value) == string_ptr_type_node &&
  		! flag_emit_class_files)))
  	    {
 +	      /* Prepare node for patch_assignment */
  	      TREE_OPERAND (node, 1) = value;
 +	      /* Call patch assignment to verify the assignment */
  	      if (patch_assignment (node, wfl_op1, value) == error_mark_node)
  		return error_mark_node;
 +	      /* Set DECL_INITIAL properly (a conversion might have
 +                 been decided by patch_assignment) and return the
 +                 empty statement. */
  	      else
 -		return empty_stmt_node;
 +		{
 +		  tree patched = patch_string (TREE_OPERAND (node, 1));
 +		  if (patched)
 +		    DECL_INITIAL (nn) = patched;
 +		  else
 +		    DECL_INITIAL (nn) = TREE_OPERAND (node, 1);
 +		  return empty_stmt_node;
 +		}
  	    }
  	  if (! flag_emit_class_files)
  	    DECL_INITIAL (nn) = NULL_TREE;
 @@ -11999,7 +12019,18 @@ java_complete_lhs (node)
  		       || JSTRING_P (TREE_TYPE (node))))
  	    node = java_refold (node);
  	}
 -      
 +
 +      /* Seek to set DECL_INITIAL to a proper value, since it might have
 +	 undergone a conversion in patch_assignment. We do that only when
 +	 it's necessary to have DECL_INITIAL properly set. */
 +      nn = TREE_OPERAND (node, 0);
 +      if (TREE_CODE (nn) == VAR_DECL 
 +	  && DECL_INITIAL (nn) && CONSTANT_VALUE_P (DECL_INITIAL (nn))
 +	  && FIELD_STATIC (nn) && FIELD_FINAL (nn) 
 +	  && (JPRIMITIVE_TYPE_P (TREE_TYPE (nn))
 +	      || TREE_TYPE (nn) == string_ptr_type_node))
 +	DECL_INITIAL (nn) = TREE_OPERAND (node, 1);
 +
        CAN_COMPLETE_NORMALLY (node) = 1;
        return node;
  

Re: java/4230: gcj -C segfaults on static string continuation '+' in gcj 3.1 expermental.

[rodrigc]

Synopsis: gcj -C segfaults on static string continuation '+' in gcj 3.1 expermental.

State-Changed-From-To: feedback->closed
State-Changed-By: rodrigc
State-Changed-When: Fri Jan  4 19:15:34 2002
State-Changed-Why:
    Patch applied:
     http://gcc.gnu.org/ml/gcc-patches/2001-09/msg00247.html 

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=4230

Re: java/4230: gcj -C segfaults on static string continuation '+' in gcj 3.1 expermental.

[apbianco]

Synopsis: gcj -C segfaults on static string continuation '+' in gcj 3.1 expermental.

Responsible-Changed-From-To: unassigned->apbianco
Responsible-Changed-By: apbianco
Responsible-Changed-When: Thu Sep  6 00:35:27 2001
Responsible-Changed-Why:
    Mine.
State-Changed-From-To: open->feedback
State-Changed-By: apbianco
State-Changed-When: Thu Sep  6 00:35:27 2001
State-Changed-Why:
    I posted a patch.

http://sources.redhat.com/cgi-bin/gnatsweb.pl?cmd=view&pr=4230&database=gcc

java/4230: gcj -C segfaults on static string continuation '+' in gcj 3.1 expermental.

[Tony Knaus]

>Number:         4230
>Category:       java
>Synopsis:       gcj -C segfaults on static string continuation '+' in gcj 3.1 expermental.
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          ice-on-legal-code
>Submitter-Id:   net
>Arrival-Date:   Tue Sep 04 15:36:00 PDT 2001
>Closed-Date:
>Last-Modified:
>Originator:     awk@spinnakernet.com
>Release:        3.1 20010902 (experimental)
>Organization:
Spinnaker Networks Inc.
>Environment:
System: Linux panic 2.2.12-20 #1 Mon Sep 27 10:40:35 EDT 1999 i686 unknown
Architecture: i686

	
host: i686-pc-linux-gnu
build: i686-pc-linux-gnu
target: i686-pc-linux-gnu
configured with: ../gcc/configure --prefix=/usr/local/gcc-3.0 --enable-shared --enable-threads=posix --enable-version-specific-runtime-libs --enable-java-awt=xlib --enable-languages=c,c++,java
>Description:

gcj 3.1 generates a segmentation fault while building the xerces parser and
generating byte code. The compiler does not seg fault when compiling
to object code or just generating dependencies.

gcj -M -C --encoding=UTF-8 -CLASSPATH ":build/xerces/src:src/xerces/src:src/cimom:src/regexp/src/java:src/servlet/src:build/tomcat/src:src/tomcat/src:src/remotetea/src:src/j-dom/jdom/src/java:src/enhydra/SchemaMapper/src/java:src/tests:src/tools/formgen:src/tools/kbgen:src/tools/typegen:src/tools/utils:src/tools/logmessage:src/ui:src/providers:src/rpc:build/ui:build/tests" build/xerces/src/org/apache/xerces/utils/regex/Token.java > build/xerces/src/org/apache/xerces/utils/regex/.Token.u
build/xerces/src/org/apache/xerces/utils/regex/Token.java:1: Internal error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL: http://www.gnu.org/software/gcc/bugs.html > for instructions.
cons: *** [build/xerces/src/org/apache/xerces/utils/regex/.Token.u] Error 1
cons: errors constructing build/xerces/src/org/apache/xerces/utils/regex/.Token.u
> 

I have tracked the problem down to the initialization of 
static strings with line continuation '+' characters. Folling is a 
simple example that causes the seg-fault.

public class StringCrash {
    static final String viramaString = 
    "f"
    +"o"
    +"o"
    +"b"
    +"a"
    +"r";

    public static void main(String[] args) {
	System.out.println("viramaString= " + viramaString);
    }
};
	
>How-To-Repeat:

> /usr/local/gcc-3.0/bin/gcj -v -C StringCrash.java
Reading specs from /usr/local/gcc-3.0/lib/gcc-lib/i686-pc-linux-gnu/3.1/specs
Configured with: ../gcc/configure --prefix=/usr/local/gcc-3.0 --enable-shared --enable-threads=posix --enable-version-specific-runtime-libs --enable-java-awt=xlib --enable-languages=c,c++,java
Thread model: posix
gcc version 3.1 20010902 (experimental)
 /usr/local/gcc-3.0/lib/gcc-lib/i686-pc-linux-gnu/3.1/jc1 StringCrash.java -quiet -dumpbase StringCrash.java -g1 -version -fsyntax-only -femit-class-files -o /dev/null
GNU Java version 3.1 20010902 (experimental) (i686-pc-linux-gnu)
        compiled by GNU C version 3.1 20010902 (experimental).
Class path starts here:
    ./
    /usr/local/gcc-3.0/share/libgcj.jar/ (system) (zip)
StringCrash.java:1: Internal error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL: http://www.gnu.org/software/gcc/bugs.html > for instructions.
> /usr/local/gcc-3.0/bin/gcj -v -E -C StringCrash.java
Reading specs from /usr/local/gcc-3.0/lib/gcc-lib/i686-pc-linux-gnu/3.1/specs
Configured with: ../gcc/configure --prefix=/usr/local/gcc-3.0 --enable-shared --enable-threads=posix --enable-version-specific-runtime-libs --enable-java-awt=xlib --enable-languages=c,c++,java
Thread model: posix
gcc version 3.1 20010902 (experimental)
> 

Compile the above example with the '-C' option.

>Fix:

workarounds:

- don't compile to byte-code
- or don't use string continuation when compiling to byte code
>Release-Note:
>Audit-Trail:
>Unformatted: