Elena Zannoni wrote:
> Adam Fedor writes:
>  > 2003-01-03  Adam Fedor  <fedor@gnu.org>
>  > 
>  > 	* symtab.c (symbol_init_demangled_name): Check for and demangle
>  > 	ObjC symbols.
>  > 	(make_symbol_completion_list): Look for ObjC symbols
>  > 
>  > Index: symtab.c
>  > ===================================================================
>  > RCS file: /cvs/src/src/gdb/symtab.c,v
>  > retrieving revision 1.84
>  > diff -u -p -r1.84 symtab.c
[...snip...]
>  > +      ALL_OBJFILE_MSYMBOLS (objfile, msymbol)
>  > +	{
>  > +	  static char *tmp = NULL;
>  > +	  static unsigned int tmplen = 0;
>  > +	  
>  > +	  char *method, *category, *selector;
>  > +	  char *tmp2 = NULL;
>  > +	  
>  > +	  QUIT;
>  > +	  
>  > +	  method = SYMBOL_DEMANGLED_NAME (msymbol);
>  > +	  if (method == NULL)
>  > +	    method = SYMBOL_NAME (msymbol);
>  > +	  if (method == NULL)
>  > +	    continue;
>  > +
>  > +	  /* add the minimal symbol no matter what */
>  > +	  completion_list_add_name (method, sym_text, sym_text_len, text, word);
>  > +     
> 
> Why don't you use the (yes, awful) macro COMPLETION_LIST_ADD_SYMBOL ? 
> 

I guess I'm not sure that SYMBOL_NAME(msymbol) is always non-NULL. Is 
that the case?

>  
>  > +	  /* is it a method? */
>  > +	  if ((method[0] != '-') && (method[0] != '+'))
>  > +	    continue;
>  > +	  if (sym_text[0] == '[')
>  > +	    /* complete on shortened method method */
>  > +	    completion_list_add_name (method + 1, sym_text, sym_text_len, text, word);
> 
> Do you add 2 methods for a symbol that starts with '['?
> 
> 
>  > +
>  > +	  while ((strlen (method) + 1) >= tmplen)
>  > +	    {
>  > +	      tmplen = (tmplen == 0) ? 1024 : tmplen * 2;
> 
> Conditional expressions are generally frowned upon.
> 
>  > +	      tmp = xrealloc (tmp, tmplen);
>  > +	    }
>  > +	  selector = strchr (method, ' ');
>  > +	  if (selector != NULL)
>  > +	    selector++;
>  > +
>  > +	  category = strchr (method, '(');
>  > +
>  > +	  if ((category != NULL) && (selector != NULL))
>  > +	    {
>  > +	      memcpy (tmp, method, (category - method));
>  > +	      tmp[category - method] = ' ';
>  > +	      memcpy (tmp + (category - method) + 1, selector, strlen (selector) + 1);
>  > +	      completion_list_add_name (tmp, sym_text, sym_text_len, text, word);
>  > +	      if (sym_text[0] == '[')
>  > +		completion_list_add_name (tmp + 1, sym_text, sym_text_len, text, word);
>  > +	    }
>  > +
>  > +	  if (selector != NULL)
>  > +	    {
>  > +	      /* complete on selector only */
>  > +	      strcpy (tmp, selector);
>  > +	      tmp2 = strchr (tmp, ']');
>  > +	      if (tmp2 != NULL)
>  > +		*tmp2 = '\0';
>  > +
>  > +	      completion_list_add_name (tmp, sym_text, sym_text_len, text, word);
>  > +	    }
> 
> I am not sure I understand: how many variations of the same name get
> added to the completion list?
> 
>  > +	}
>

For a particular completion symbol, which could be something like

isFlip or [MyObject isFlip

You could potentially complete on any of

-[MyObject(PrivateCategory) isFlipped]
-[MyObject isFlipped]
[MyObject isFlipped]
isFlipped


All of these symbols COULD be unique, although in practice this is 
unlikely. Attached is an updated patch (second half only).