Re: How to check that selectors exist

[Stan Shebs]

Nicola Pero wrote:

> Here is a set of patches implementing -Wundeclared-selector as promised.
>
> NB: I will be leaving for two weeks on saturday (holidays finally :-).  
> If Stan approves it when I'm away, I'm happy with anyone else applying it
> (including any sort of reformatting and improvements which might be
> necessary).  Else, I'll apply it if approved before I leave (which I
> doubt, given how slow Stan is in approving :-), or when I am back.
Just for that :-), here's the OK.  (modulo a couple nits)

> +/*
> + * This function is called by the parser when a @selector() expression
> + * is found, in order to compile it.  It is only called by the parser
> + * and only to compile a @selector().
> + */
The usual style is not to have leading '*' in comment blocks.

> +	  warning ("unknown (undeclared) selector `%s'", 
> +		   IDENTIFIER_POINTER (selname));
I suggest phrasing this as just "undeclared selector `%s'".  The use of
both terms made me wonder if it were possible to have known but
undeclared selectors, or vice versa.

>  (METHOD_SEL_NAME (mchain) == key)
> -       return mchain;
> +       {
> +         return mchain;
> +       }
Coding style sez not to do this change.

Also, you need to document it in doc/invoke.texi, otherwise the users
might never learn of this nice new warning's existence!

Stan