Re: [RFA] More wrappers in varobj

Re: [RFA] More wrappers in varobj

[Andrew Cagney]

> Hi,
> 
> I've been trying to update varobj in light of what appears to be either
> bitrot or a bunch of v3 abi issues. We are seeing an extraordinary number
> of problems with C++.

Keith,

have a look at breakpoint.c:gdb_breakpoint for a more robust / current 
way of implementing these wrappers.

In particular:

- it uses catch_exceptions() which has better defined return values.

- the wrapped function has the correct function signature (not char *) 
meaning that the (illegal) function cast is not needed.

- it uses a dedicated struct to pass in the parameter list instead of 
relying on casts and a union.

enjoy,
Andrew


> Index: wrapper.h
> ===================================================================
> RCS file: /cvs/src/src/gdb/wrapper.h,v
> retrieving revision 1.10
> diff -u -p -r1.10 wrapper.h
> --- wrapper.h	2001/03/06 08:21:18	1.10
> +++ wrapper.h	2001/10/15 19:45:21
> @@ -37,6 +37,8 @@ extern int gdb_value_subscript (value_pt
> 
>  extern int gdb_value_ind (value_ptr val, value_ptr * rval);
> 
> +extern int gdb_value_struct_elt (value_ptr *argp, value_ptr *args, char *name, int *static_memfuncp, char *err, value_ptr * rval);
> +
>  extern int gdb_parse_and_eval_type (char *, int, struct type **);
> 
>  #endif /* WRAPPER_H */
> Index: wrapper.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/wrapper.c,v
> retrieving revision 1.12
> diff -u -p -r1.12 wrapper.c
> --- wrapper.c	2001/03/27 20:36:24	1.12
> +++ wrapper.c	2001/10/15 19:45:21
> @@ -55,6 +55,8 @@ static int wrap_value_subscript (char *)
> 
>  static int wrap_value_ind (char *opaque_arg);
> 
> +static int wrap_value_struct_elt (char *opaque_arg);
> +
>  static int wrap_parse_and_eval_type (char *);
> 
>  int
> @@ -257,6 +259,47 @@ wrap_value_ind (char *opaque_arg)
> 
>    val = (value_ptr) (args)->args[0].pointer;
>    (args)->result.pointer = value_ind (val);
> +  return 1;
> +}
> +
> +int
> +gdb_value_struct_elt (value_ptr *argp, value_ptr *args, char *name,
> +		      int *static_memfuncp, char *err, value_ptr * rval)
> +{
> +  struct gdb_wrapper_arguments argss;
> +
> +  argss.args[0].pointer = argp;
> +  argss.args[1].pointer = args;
> +  argss.args[2].pointer = name;
> +  argss.args[3].pointer = static_memfuncp;
> +  argss.args[4].pointer = err;
> +
> +  if (!catch_errors ((catch_errors_ftype *) wrap_value_struct_elt, &argss,
> +		     "", RETURN_MASK_ERROR))
> +    {
> +      /* An error occurred */
> +      return 0;
> +    }
> +
> +  *rval = (value_ptr) argss.result.pointer;
> +  return 1;
> +}
> +
> +static int
> +wrap_value_struct_elt (char *opaque_arg)
> +{
> +  char *err, *name;
> +  value_ptr *argp, *args;
> +  int *static_memfuncp;
> +  struct gdb_wrapper_arguments *argss = (struct gdb_wrapper_arguments *) opaque_arg;
> +
> +  argp = (value_ptr *) argss->args[0].pointer;
> +  args = (value_ptr *) argss->args[1].pointer;
> +  name = (char *) argss->args[2].pointer; 
> +  static_memfuncp = argss->args[3].pointer;
> +  err  = (char *) argss->args[4].pointer;
> +
> +  (argss)->result.pointer = value_struct_elt (argp, args, name, static_memfuncp, err);
>    return 1;
>  }
> 
> 
> 
> 

Re: [RFA] More wrappers in varobj

[Keith Seitz]

On Mon, 15 Oct 2001, Andrew Cagney wrote:

> have a look at breakpoint.c:gdb_breakpoint for a more robust / current
> way of implementing these wrappers.

Am I to presume that we are deprecating wrapper.[ch] in favor of this new
mechanism? If so, I could just work to get rid of wrapper.[ch]
altogether...

Just say the word.
Keith

Re: [RFA] More wrappers in varobj

[Andrew Cagney]

> On Mon, 15 Oct 2001, Andrew Cagney wrote:
> 
> 
>> have a look at breakpoint.c:gdb_breakpoint for a more robust / current
>> way of implementing these wrappers.
> 
> 
> Am I to presume that we are deprecating wrapper.[ch] in favor of this new
> mechanism? If so, I could just work to get rid of wrapper.[ch]
> altogether...
> 
> Just say the word.
> Keith

Should the code live in wrapper.[hc]?  It can for now.

While it is a hack (the code it calls should always unwind the stack 
cleanly) it, or something like it, is going to remain around for a long 
time.  It is also an internal interface (varobj | core-gdb) rather than 
a libgdb interface.

Regarding the problems, the worst is the cast in:

	catch_errors ((catch_errors_ftype*)wrap_function, ...)

with

	wrap_function (char *a)

it isn't valid C - you can't assume that a ()(void*) function is called 
the same way as a ()(char*) function.

The other two are not so much of a concern (although I must admit a 
desire to avoid the unions and use structs so that the compiler can be 
used to check that parameter passing).  catch_exceptions() makes it 
possible for the wrapped functions to return well defined values - 
something not possible with catch_errors().

Andrew

Re: [RFA] More wrappers in varobj

[Keith Seitz]

Sorry for the delay. I got a little caught up in debugging some
other stuff...

On Mon, 15 Oct 2001, Andrew Cagney wrote:

> have a look at breakpoint.c:gdb_breakpoint for a more robust / current
> way of implementing these wrappers.

Is this better?

Slightly off-topic: We are going to be calling externally visible
functions something like "gdb_FOO", right? Example: gdb_breakpoint and
gdb_breakpoint_query. How are we going to name internal functions? Leading
"_" or something?

I ask because I don't really want to have this patch introduce
internal variations of gdb functions into the "libgdb" namespace.

Keith

Index: wrapper.h
===================================================================
RCS file: /cvs/src/src/gdb/wrapper.h,v
retrieving revision 1.10
diff -u -p -r1.10 wrapper.h
--- wrapper.h	2001/03/06 08:21:18	1.10
+++ wrapper.h	2001/10/19 18:36:50
@@ -18,6 +18,7 @@

 #ifndef WRAPPER_H
 #define WRAPPER_H 1
+#include "gdb.h"

 /* Use this struct to pass arguments to wrapper routines. */
 struct gdb_wrapper_arguments;
@@ -36,6 +37,10 @@ extern int gdb_value_assign (value_ptr,
 extern int gdb_value_subscript (value_ptr, value_ptr, value_ptr *);

 extern int gdb_value_ind (value_ptr val, value_ptr * rval);
+
+extern enum gdb_rc gdb_value_struct_elt (struct ui_out *uiout, value_ptr *result_ptr,
+					 value_ptr *argp, value_ptr *args,
+					 char *name, int *static_memfuncp, char *err);

 extern int gdb_parse_and_eval_type (char *, int, struct type **);

Index: wrapper.c
===================================================================
RCS file: /cvs/src/src/gdb/wrapper.c,v
retrieving revision 1.12
diff -u -p -r1.12 wrapper.c
--- wrapper.c	2001/03/27 20:36:24	1.12
+++ wrapper.c	2001/10/19 18:36:50
@@ -41,6 +41,16 @@ struct gdb_wrapper_arguments
       } args[10];
   };

+struct captured_value_struct_elt_args
+{
+  value_ptr *argp;
+  value_ptr *args;
+  char *name;
+  int *static_memfuncp;
+  char *err;
+  value_ptr *result_ptr;
+};
+
 static int wrap_parse_exp_1 (char *);

 static int wrap_evaluate_expression (char *);
@@ -55,6 +65,8 @@ static int wrap_value_subscript (char *)

 static int wrap_value_ind (char *opaque_arg);

+static int do_captured_value_struct_elt (struct ui_out *uiout, void *data);
+
 static int wrap_parse_and_eval_type (char *);

 int
@@ -290,3 +302,29 @@ wrap_parse_and_eval_type (char *a)

   return 1;
 }
+
+enum gdb_rc
+gdb_value_struct_elt (struct ui_out *uiout, value_ptr *result, value_ptr *argp,
+		      value_ptr *args, char *name, int *static_memfuncp,
+		      char *err)
+{
+  struct captured_value_struct_elt_args cargs;
+  cargs.argp = argp;
+  cargs.args = args;
+  cargs.name = name;
+  cargs.static_memfuncp = static_memfuncp;
+  cargs.err = err;
+  cargs.result_ptr = result;
+  return catch_exceptions (uiout, do_captured_value_struct_elt, &cargs,
+			   NULL, RETURN_MASK_ALL);
+}
+
+static int
+do_captured_value_struct_elt (struct ui_out *uiout, void *data)
+{
+  struct captured_value_struct_elt_args *cargs = data;
+  *cargs->result_ptr = value_struct_elt (cargs->argp, cargs->args, cargs->name,
+			     cargs->static_memfuncp, cargs->err);
+  return GDB_RC_OK;
+}
+