[RFA]: Patch to fix the SEGV error when printing f77 array element

[RFA]: Patch to fix the SEGV error when printing f77 array element

[Wu Zhou]

Daniel, 

I had cleaned up the patch to fix the SEGV error we discussed on gdb@ the
days before and tested it on ppc64 arch with the latest CVS tree.  It 
works ok.  Please review and comment.  Thanks.

2005-06-9  Wu Zhou  <woodzltc@cn.ibm.com>

	* eval.c (evaluate_subexp_standard): Add code to check the target
	type of a TYPE_CODE_PTR value when we encounter a f77 undetermined
	arglist.  If it is TYPE_CODE_ARRAY or TYPE_CODE_STRING, jump to 
	relevant code path, else go on with the original path.

Index: eval.c
===================================================================
RCS file: /cvs/src/src/gdb/eval.c,v
retrieving revision 1.55
diff -c -p -r1.55 eval.c
*** eval.c	26 May 2005 20:48:58 -0000	1.55
--- eval.c	9 Jun 2005 09:23:02 -0000
*************** evaluate_subexp_standard (struct type *e
*** 1254,1260 ****
  	case TYPE_CODE_STRING:
  	  goto op_f77_substr;
  
- 	case TYPE_CODE_PTR:
  	case TYPE_CODE_FUNC:
  	  /* It's a function call. */
  	  /* Allocate arg vector, including space for the function to be
--- 1254,1259 ----
*************** evaluate_subexp_standard (struct type *e
*** 1267,1272 ****
--- 1266,1298 ----
  	  argvec[tem] = 0;	/* signal end of arglist */
  	  goto do_call_it;
  
+ 	case TYPE_CODE_PTR:
+ 	  /* Fortran always passes variables to subroutines as pointer. 
+ 	     So we need to look into its target type to see if it is 
+ 	     array subscript or substring operations.  */
+ 	  switch (TYPE_CODE (TYPE_TARGET_TYPE (type)))
+ 	    {
+ 	    case TYPE_CODE_ARRAY:
+ 	      arg1 = value_ind (arg1);
+ 	      type = check_typedef (value_type (arg1));
+ 	      goto multi_f77_subscript;
+ 
+ 	    case TYPE_CODE_STRING:
+ 	      arg1 = value_ind (arg1);
+ 	      type = check_typedef (value_type (arg1));
+ 	      goto op_f77_substr;
+ 
+ 	    default:
+ 	      /* It's a function call.  */
+ 	      argvec = (struct value **) alloca (sizeof (struct value *) * (nargs + 2));
+ 	      argvec[0] = arg1;
+ 	      tem = 1;
+ 	      for (; tem <= nargs; tem++)
+ 		argvec[tem] = evaluate_subexp_with_coercion (exp, pos, noside);
+ 	      argvec[tem] = 0;	/* signal end of arglist */
+ 	      goto do_call_it;
+ 	    }
+ 
  	default:
  	  error (_("Cannot perform substring on this type"));
  	}


Cheers
- Wu Zhou

Re: [RFA]: Patch to fix the SEGV error when printing f77 array element

[Daniel Jacobowitz]

On Thu, Jun 09, 2005 at 06:02:03PM +0800, Wu Zhou wrote:
> Daniel, 
> 
> I had cleaned up the patch to fix the SEGV error we discussed on gdb@ the
> days before and tested it on ppc64 arch with the latest CVS tree.  It 
> works ok.  Please review and comment.  Thanks.

Can you handle TYPE_CODE_PTR before the switch statement, instead?

Also, rather than default, you can probably check that the target type
of the pointer is TYPE_CODE_FUNC.

Do you feel up to adding our first compilable fortran test to the
testsuite?

-- 
Daniel Jacobowitz
CodeSourcery, LLC

Re: [RFA]: Patch to fix the SEGV error when printing f77 array element

[Wu Zhou]

On Thu, 9 Jun 2005, Daniel Jacobowitz wrote:

> On Thu, Jun 09, 2005 at 06:02:03PM +0800, Wu Zhou wrote:
> > Daniel, 
> > 
> > I had cleaned up the patch to fix the SEGV error we discussed on gdb@ the
> > days before and tested it on ppc64 arch with the latest CVS tree.  It 
> > works ok.  Please review and comment.  Thanks.
> 
> Can you handle TYPE_CODE_PTR before the switch statement, instead?
> 
> Also, rather than default, you can probably check that the target type
> of the pointer is TYPE_CODE_FUNC.

Good point.  What about the following re-worked patch?  Thanks.

2005-06-10  Wu Zhou  <woodzltc@cn.ibm.com>

	* eval.c (evaluate_subexp_standard): Add code to check the target
	type of a TYPE_CODE_PTR value when we encounter a f77 undetermined
	arglist.  If it is array, string or function, work on the target 
	value instead. 

Index: eval.c
===================================================================
RCS file: /cvs/src/src/gdb/eval.c,v
retrieving revision 1.55
diff -c -p -r1.55 eval.c
*** eval.c	26 May 2005 20:48:58 -0000	1.55
--- eval.c	10 Jun 2005 07:04:12 -0000
*************** evaluate_subexp_standard (struct type *e
*** 1246,1251 ****
--- 1246,1269 ----
        type = check_typedef (value_type (arg1));
        code = TYPE_CODE (type);
  
+       if (code == TYPE_CODE_PTR)
+ 	{
+ 	  /* Fortran always passes variable to subroutines as pointer.
+ 	     So we need to look into its target type to see if it is
+ 	     array, string or function.  If it is, we need to switch
+ 	     to the target value the original one points to.  */ 
+ 	  struct type *target_type = check_typedef (TYPE_TARGET_TYPE (type));
+ 
+ 	  if (TYPE_CODE (target_type) == TYPE_CODE_ARRAY
+ 	      || TYPE_CODE (target_type) == TYPE_CODE_STRING
+ 	      || TYPE_CODE (target_type) == TYPE_CODE_FUNC)
+ 	    {
+ 	      arg1 = value_ind (arg1);
+ 	      type = check_typedef (value_type (arg1));
+ 	      code = TYPE_CODE (type);
+ 	    }
+ 	} 
+ 
        switch (code)
  	{
  	case TYPE_CODE_ARRAY:

> Do you feel up to adding our first compilable fortran test to the
> testsuite?

Yes, I am thinking of adding a testcase to verify this.  But do I need
to keep some principles/tips in mind when adding a new testcase to the
testsuite?  As you might know, I didn't do that before.  TIA.

Cheers
- Wu Zhou

Re: [RFA]: Patch to fix the SEGV error when printing f77 array element

[Daniel Jacobowitz]

On Fri, Jun 10, 2005 at 03:22:50PM +0800, Wu Zhou wrote:
> 2005-06-10  Wu Zhou  <woodzltc@cn.ibm.com>
> 
> 	* eval.c (evaluate_subexp_standard): Add code to check the target
> 	type of a TYPE_CODE_PTR value when we encounter a f77 undetermined
> 	arglist.  If it is array, string or function, work on the target 
> 	value instead. 

This version is fine.  Thanks!

-- 
Daniel Jacobowitz
CodeSourcery, LLC

Re: [RFA]: Patch to fix the SEGV error when printing f77 array element

[Wu Zhou]

Quoting Daniel Jacobowitz <drow@false.org>:

> This version is fine.  Thanks!

Commited.  Thanks for reviewing this.

Cheers
- Wu Zhou