[PATCH] PR java/15769 infinite loop

[PATCH] PR java/15769 infinite loop

[Andrew Pinski]

Just like C/C++ front-end change which Zack committed/tested for me,
the Java front-end had forgot to add the unordered tree codes to the
list of tree codes which act as comparisons to truthvalue_conversion
language hook when the new tree codes were added.
The C/C++ change is referenced here:
<http://gcc.gnu.org/ml/gcc-patches/2004-05/msg01458.html>

Thanks,
Andrew Pinski

Testcase:
class bug15769 {
	private boolean foo () { return false; }

	public boolean bar (double blaz)
	{	
		return (Double.POSITIVE_INFINITY != blaz) && foo ();
	}
}


ChangeLog:
         * expr.c (java_truthvalue_conversion): Handle
         UNEQ_EXPR, UNLE_EXPR, UNGE_EXPR, UNLT_EXPR, UNGT_EXPR,
         ORDERED_EXPR, and UNORDERED_EXPR as comparison operators,
         i.e. return the expression.


Patch:
Index: expr.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/java/expr.c,v
retrieving revision 1.191
diff -u -p -r1.191 expr.c
--- expr.c	14 May 2004 02:29:32 -0000	1.191
+++ expr.c	3 Jun 2004 17:33:41 -0000
@@ -160,12 +160,16 @@ java_truthvalue_conversion (tree expr)

    switch (TREE_CODE (expr))
      {
-    case EQ_EXPR:
-    case NE_EXPR: case LE_EXPR: case GE_EXPR: case LT_EXPR: case 
GT_EXPR:
+    case EQ_EXPR:   case NE_EXPR:   case UNEQ_EXPR: case LTGT_EXPR:
+    case LE_EXPR:   case GE_EXPR:   case LT_EXPR:   case GT_EXPR:
+    case UNLE_EXPR: case UNGE_EXPR: case UNLT_EXPR: case UNGT_EXPR:
+    case ORDERED_EXPR: case UNORDERED_EXPR:
      case TRUTH_ANDIF_EXPR:
      case TRUTH_ORIF_EXPR:
      case TRUTH_AND_EXPR:
      case TRUTH_OR_EXPR:
+    case TRUTH_XOR_EXPR:
+    case TRUTH_NOT_EXPR:
      case ERROR_MARK:
        return expr;

Re: [PATCH] PR java/15769 infinite loop

[Andrew Pinski]

On Jun 3, 2004, at 13:43, Andrew Pinski wrote:

> Just like C/C++ front-end change which Zack committed/tested for me,
> the Java front-end had forgot to add the unordered tree codes to the
> list of tree codes which act as comparisons to truthvalue_conversion
> language hook when the new tree codes were added.
> The C/C++ change is referenced here:
> <http://gcc.gnu.org/ml/gcc-patches/2004-05/msg01458.html>
>
> ChangeLog:
>         * expr.c (java_truthvalue_conversion): Handle
>         UNEQ_EXPR, UNLE_EXPR, UNGE_EXPR, UNLT_EXPR, UNGT_EXPR,
>         ORDERED_EXPR, and UNORDERED_EXPR as comparison operators,
>         i.e. return the expression.

Forgot to say bootstrapped on i686-pc-linux-gnu with no regressions.

Re: [PATCH] PR java/15769 infinite loop

[Tom Tromey]

>>>>> "Andrew" == Andrew Pinski <pinskia@physics.uc.edu> writes:

Andrew> Just like C/C++ front-end change which Zack committed/tested for me,
Andrew> the Java front-end had forgot to add the unordered tree codes to the
Andrew> list of tree codes which act as comparisons to truthvalue_conversion
Andrew> language hook when the new tree codes were added.

This looks fine to me, but Andrew or Per will have to approve.

The test case is fine.  I couldn't tell whether this is an execute
test or just a compilation test, but just go ahead and put it in the
appropriate place.

Tom

Re: [PATCH] PR java/15769 infinite loop

[Andrew Pinski]

> 
> >>>>> "Andrew" == Andrew Pinski <pinskia@physics.uc.edu> writes:
> 
> Andrew> Just like C/C++ front-end change which Zack committed/tested for me,
> Andrew> the Java front-end had forgot to add the unordered tree codes to the
> Andrew> list of tree codes which act as comparisons to truthvalue_conversion
> Andrew> language hook when the new tree codes were added.
> 
> This looks fine to me, but Andrew or Per will have to approve.
> 
> The test case is fine.  I couldn't tell whether this is an execute
> test or just a compilation test, but just go ahead and put it in the
> appropriate place.

It is just an compilation test, because before the patch gimplification
would go into an infinite loop.

Thanks,
Andrew Pinski

Re: [PATCH] PR java/15769 infinite loop

[Andrew Haley]

Tom Tromey writes:
 > >>>>> "Andrew" == Andrew Pinski <pinskia@physics.uc.edu> writes:
 > 
 > Andrew> Just like C/C++ front-end change which Zack committed/tested for me,
 > Andrew> the Java front-end had forgot to add the unordered tree codes to the
 > Andrew> list of tree codes which act as comparisons to truthvalue_conversion
 > Andrew> language hook when the new tree codes were added.
 > 
 > This looks fine to me, but Andrew or Per will have to approve.

Looks fine to me too.

Andrew.