[Patch] -var-evaluate-expression NAME [FORMAT]

[Marc Khouzam]

> > I kind of cheated and used NULL to indicate that the format used should be
> > the one stored in the varobj structure.
> 
> I would use a new function, e.g., something like 
> 
> char *
> varobj_get_formatted_value (struct varobj *var,
> 			    enum varobj_display_formats format)
> {
>   enum varobj_display_formats oldformat;
>   char* value;
>   oldformat = var->format;
>   var->format = format;
>   value = varobj_get_value (var);
>   var->format = oldformat;
>   return value;
> }
> 
> and call it when argc == 2 in mi_cmd_var_evaluate_expression so that you
> don't have to change all calls to varobj_get_value and related functions.
> (Note: not tested)

I hadn't thought of temporarily setting var->format, instead of passing 'format'
everywhere.  Much easier!  Is there any way that varob_get_value() could not return?
Like a failed gdb_assert?  Or are we sure that var->format will be reset to
the old format properly no matter what happens to varobj_get_value?

> > +/* Parse a string argument into a format value.  */
> > +
> > +static enum varobj_display_formats
> > +mi_parse_format (const char *arg)
> > +{
> > +  int len;
> > +         
> > +  len = strlen (arg);
> > +
> > +  if (strncmp (arg, "natural", len) == 0)
> > +    return FORMAT_NATURAL;
> > +  else if (strncmp (arg, "binary", len) == 0)
> > +       return FORMAT_BINARY;
> > +  else if (strncmp (arg, "decimal", len) == 0)
> > +       return FORMAT_DECIMAL;
> > +  else if (strncmp (arg, "hexadecimal", len) == 0)
> > +       return FORMAT_HEXADECIMAL;
> > +  else if (strncmp (arg, "octal", len) == 0)
> > +       return FORMAT_OCTAL;
> > +  else
> > +    error (_("Unknown display format: must be: \"natural\", \"binary\", \"decimal\", \"hexadecimal\", or \"octal\""));
> > +}
> > +
> 
> and I would refactor this out of mi_cmd_var_set_format.

Didn't I do that already?