Re: variable objects and registers

[Nick Roberts <nickrob@snap.net.nz>]

Vladimir Prus writes:
> This patch adds new command -var-registers that creates and returns a list of
> variable objects for all registers gdb knows. The command takes one option --
> the frame, which is specified just like for -var-create. While not all
> registers are saved, and so gdb might not know values of some registers in
> parent frames, for some registers it's possible, and frontends might want to
> access those values.


> Here's example output:

>      -var-registers *
>      ^done,registers={
       ^done,registers=[

>		{name="var1",exp="$eax",numchild="0",value="16",type="int"},
>		............
>                {name="var10",exp="$eflags",numchild="0",value="[ SF IF ID ]",


Since this command creates new variable objects, perhaps it should be called
something like -var-create-registers.  

Alternatively, taking things further, instead of reusing some code for each
command through create_varobj_in_frame, perhaps you could have just one
command:

"-var-create" and "-var-create -r" for registers ("-var-create -m" for
memory-mapped registers).

I've not thought it through, I'm just brainstorming.

--- gdb/mi/mi-cmd-var.c	(/patches/gdb/varobj_printing/gdb_mainline)	(revision 2338)
+++ gdb/mi/mi-cmd-var.c	(/patches/gdb/varobj_for_registers/gdb_mainline)	(revision 2338)
@@ -68,16 +68,39 @@
 
 /* VAROBJ operations */
 
+static struct varobj *
+create_varobj_in_frame (char *name, char *expression, char *frame)
+{
+  CORE_ADDR frameaddr = 0;
+  struct cleanup *cleanup;
+  enum varobj_type var_type;
+
+  if (strcmp (frame, "*") == 0)
+    var_type = USE_CURRENT_FRAME;
+  else if (strcmp (frame, "@") == 0)
+    var_type = USE_SELECTED_FRAME;  
+  else
+    {
+      var_type = USE_SPECIFIED_FRAME;
+      frameaddr = string_to_core_addr (frame);
+    }
+
+  if (varobjdebug)
+    fprintf_unfiltered (gdb_stdlog,
+		    "Name=\"%s\", Frame=\"%s\" (0x%s), Expression=\"%s\"\n",
+			name, frame, paddr (frameaddr), expression);
+
+  return varobj_create (name, expression, frameaddr, var_type);
+}
+

I think the function above should go in varobj.c


...
+  numregs = NUM_REGS + NUM_PSEUDO_REGS;
+
+  make_cleanup_ui_out_tuple_begin_end (uiout, "registers");

I think this should be

make_cleanup_ui_out_list_begin_end (uiout, "registers");


+
+  for (regnum = 0; regnum < numregs; regnum++)
+    {
+      if (REGISTER_NAME (regnum) != NULL
+	  && *(REGISTER_NAME (regnum)) != '\0')
+	{
+	  char *name;
+	  char *expression;
+	  struct varobj *var;
+	  struct cleanup *cleanup_child;
+
+	  name = varobj_gen_name ();
+	  make_cleanup (free_current_contents, &name);
+
+	  expression = xstrprintf ("$%s", REGISTER_NAME (regnum));
+	  make_cleanup (xfree, expression);
+
+	  var = create_varobj_in_frame (name, expression, frame);
+
+	  cleanup_child = make_cleanup_ui_out_tuple_begin_end (uiout, NULL);
+	  print_varobj (var, PRINT_ALL_VALUES, 1 /* print expression */);
+	  do_cleanups (cleanup_child);
+	}
+    }
+
+  do_cleanups (cleanup);
+  return MI_CMD_DONE;
+}
+
+
+enum mi_cmd_result
 mi_cmd_var_delete (char *command, char **argv, int argc)
 {
   char *name;


Daniel Jacobowitz writes:

> No one else had any comments, and it looks fine to me.  The patch looks
> OK too.  It'll want a testcase, naturally.

And documentation too, hopefully!

-- 
Nick                                           http://www.inet.net.nz/~nickrob