This is the mail archive of the gdb-patches@sources.redhat.com mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH]: Add expected type to hand_function_call


Adam Fedor writes:
 > 
 > 
 > Elena Zannoni wrote:
 > > Adam Fedor writes:
 > >  > This patch adds a parameter to hand_function_call that can be used to 
 > >  > check the type of the function return. This is used often with 
 > >  > Objective-C as this information is known by the runtime and can be used 
 > >  > to check the validity of the function.
 > >  > 
 > >  > 2002-11-24  Adam Fedor  <fedor at gnu dot org>
 > >  > 
 > >  > 	* valops.c (find_function_addr): Make non-static.
 > >  > 	(hand_function_call): Add expect_type arg, use it to check
 > >  > 	return type of function (updated by Klee Dienes  <kdienes at apple dot com>).
 > >  > 	(call_function_by_hand): Update for change in hand_function_call.
 > >  > 	(call_function_by_hand_expecting_type): New function.
 > >  > 	* value.h (find_function_addr,
 > >  > 	call_function_by_hand_expecting_type): Declare.
 > >  > 
 > >  > Index: valops.c
 > >  > ===================================================================
 > >  > RCS file: /cvs/src/src/gdb/valops.c,v
 > >  > retrieving revision 1.79
 > >  > diff -u -p -r1.79 valops.c
 > >  > --- valops.c	7 Nov 2002 02:45:27 -0000	1.79
 > >  > +++ valops.c	25 Nov 2002 02:50:17 -0000
 > >  > @@ -48,7 +48,6 @@ extern int overload_debug;
 > >  >  static int typecmp (int staticp, int varargs, int nargs,
 > >  >  		    struct field t1[], struct value *t2[]);
 > >  >  
 > >  > -static CORE_ADDR find_function_addr (struct value *, struct type **);
 > >  >  static struct value *value_arg_coerce (struct value *, struct type *, int);
 > >  >  
 > >  >  
 > >  > @@ -1228,7 +1227,7 @@ value_arg_coerce (struct value *arg, str
 > >  >  /* Determine a function's address and its return type from its value.
 > >  >     Calls error() if the function is not valid for calling.  */
 > >  >  
 > >  > -static CORE_ADDR
 > >  > +CORE_ADDR
 > >  >  find_function_addr (struct value *function, struct type **retval_type)
 > >  >  {
 > >  >    register struct type *ftype = check_typedef (VALUE_TYPE (function));
 > >  > @@ -1296,7 +1295,8 @@ find_function_addr (struct value *functi
 > >  >     ARGS is modified to contain coerced values. */
 > >  >  
 > >  >  static struct value *
 > >  > -hand_function_call (struct value *function, int nargs, struct value **args)
 > >  > +hand_function_call (struct value *function, struct type *expect_type,
 > >  > +                    int nargs, struct value **args)
 > >  >  {
 > >  >    register CORE_ADDR sp;
 > >  >    register int i;
 > >  > @@ -1342,6 +1342,17 @@ hand_function_call (struct value *functi
 > >  >    if (!target_has_execution)
 > >  >      noprocess ();
 > >  >  
 > >  > +  funaddr = find_function_addr (function, &value_type);
 > >  > +  CHECK_TYPEDEF (value_type);
 > >  > +
 > >  > +  if ((value_type == NULL) || (TYPE_CODE (value_type) == TYPE_CODE_ERROR))
 > >  > +    value_type = expect_type;
 > >  > +
 > >  > +  if ((value_type == NULL) || (TYPE_CODE (value_type) == TYPE_CODE_ERROR))
 > >  > +    error ("Unable to call function at 0x%lx: no return type information available.\n"
 > >  > +           "To call this function anyway, you can cast the return type explicitly (e.g. 'print (float) fabs (3.0)')",
 > >  > +           (unsigned long) funaddr);
 > >  > +
 > > 
 > > Could this code be moved into the new function? This way you could leave
 > > hand_function_call alone, and just have 2 wrappers call_function_by_hand and
 > > call_function_by_hand_expecting_type.
 > > 
 > 
 > It's possible, but doing that would probably make it more contrived, and 
 > I would still have to change hand_function_call to pass in the 
 > value_type. Really, if I just remove the error message for NULL 
 > value_type, the entire change just consists of one if statement:
 > 
 > if ((value_type == NULL) || (TYPE_CODE (value_type) == TYPE_CODE_ERROR))
 >      value_type = expect_type;
 > 
 > Which is pretty benign.
 > 

Oh wait. Ignore my comments. I went and reread the old thread when
this was submitted by apple.  There was some agreement that the error
message was a good thing. So it should stay.

Also, the new behavior should be documented.

Did you submit the corresponding change to the eval.c file? I don't
see it.

I think new testcases should be added with the cast syntax as well.

elena



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]