[RFA] fix gdb.ada/null_record.exp

[RFA] fix gdb.ada/null_record.exp

[Joel Brobecker]

[-- Attachment #1: Type: text/plain, Size: 1899 bytes --]

Hello,

Daniel reported that null_record.exp was failing. The test was failing
on my side as well, although not the same way (I was so intent on
finding out what the cause was that I didn't pay enough attention to
what Daniel posted - sorry).

Here is what I observed:

    ptype empty
    type = <void>
    (gdb) FAIL: gdb.ada/null_record.exp: ptype on null record

The output is not a surprise when you look at what ada_evaluate_subexp
returns for OP_TYPE expressions:

    case OP_TYPE:
      /* The value is not supposed to be used.  This is here to make it
         easier to accommodate expressions that contain types.  */
      (*pos) += 2;
      if (noside == EVAL_SKIP)
        goto nosideret;
      else if (noside == EVAL_AVOID_SIDE_EFFECTS)
!!!->   return allocate_value (builtin_type_void);
      else
        error (_("Attempt to use a type name as an expression"));

The reason this was good enough is that, in GDB 6.4 (we are still
using this as our baseline), ptype_command was not doing a full
evaluation of the expression to get the type description. Rather,
it used a short-cut:

        static struct type *
        ptype_eval (struct expression *exp)
        {
          if (exp->elts[0].opcode == OP_TYPE)
            {
              return (exp->elts[1].type);
            }
          else
          [...]
        }

This was changed since then to call evaluate_type (expr) instead,

Knowing this, I discovered that calling ptype on probably any type
name would return this "<void>" type. I reproduced this with a non-
null record too, for instance.

The proposed fix is attached:

2007-01-02  Joel Brobecker  <brobecker@adacore.com>

        * ada-lang.c (ada_evaluate_subexp) [OP_TYPE]: Return a value with
        the appropriate type rather than a bogus void type.

Tested on x86-linux, no regression. Fixes the null_record.exp on my side.
OK to apply?

Thanks,
-- 
Joel

[-- Attachment #2: ada-lang.c.diff --]
[-- Type: text/plain, Size: 628 bytes --]

Index: ada-lang.c
===================================================================
RCS file: /cvs/src/src/gdb/ada-lang.c,v
retrieving revision 1.85
diff -u -p -r1.85 ada-lang.c
--- ada-lang.c	1 Dec 2006 00:32:29 -0000	1.85
+++ ada-lang.c	2 Jan 2007 10:52:36 -0000
@@ -8576,7 +8576,7 @@ ada_evaluate_subexp (struct type *expect
       if (noside == EVAL_SKIP)
         goto nosideret;
       else if (noside == EVAL_AVOID_SIDE_EFFECTS)
-        return allocate_value (builtin_type_void);
+        return allocate_value (exp->elts[pc + 1].type);
       else
         error (_("Attempt to use a type name as an expression"));
 

Re: [RFA] fix gdb.ada/null_record.exp

[Daniel Jacobowitz]

On Tue, Jan 02, 2007 at 03:31:02PM +0400, Joel Brobecker wrote:
> 2007-01-02  Joel Brobecker  <brobecker@adacore.com>
> 
>         * ada-lang.c (ada_evaluate_subexp) [OP_TYPE]: Return a value with
>         the appropriate type rather than a bogus void type.

OK.  Won't fix it here, I think, but I strongly suspect bad debug info.

-- 
Daniel Jacobowitz
CodeSourcery

Re: [RFA] fix gdb.ada/null_record.exp

[Joel Brobecker]

> > 2007-01-02  Joel Brobecker  <brobecker@adacore.com>
> > 
> >         * ada-lang.c (ada_evaluate_subexp) [OP_TYPE]: Return a value with
> >         the appropriate type rather than a bogus void type.
> 
> OK.  Won't fix it here, I think, but I strongly suspect bad debug info.

Thanks Daniel. Checked in.
(we'll get your test to pass...)

-- 
Joel