[patch] Fix optional arguments in getopt.

[Peter Rosin]

A while back Peter Rosin wrote:
> I noticed that the getopt implementation does not handle optional
> arguments very well. E.g. ./foo --listen, where listen is an optional
> long argument, is interpreted to have an isten argument. Also, it is
> broken if other short options precede an optional short option. I.e.
> when -l is optional, ./foo -l works, but ./foo -il does not.
> 
> Here is a small fix for that, please apply...
> 
> I'm not on the list, please CC me. Thanks!
> 
> Cheers,
> Peter
> 
> 
> 	* libc/stdlib/getopt.c (getopt_internal): Handle optional
> 	arguments better for long options and short options not
> 	appearing as the first option in a sequence.
> 
> --- newlib/libc/stdlib/getopt.c	2008-02-04 11:21:50.242125000 +0100
> +++ newlib/libc/stdlib/getopt.c	2008-02-04 11:25:11.867125000 +0100
> @@ -308,13 +308,8 @@
>      case OPTIONAL_ARG:
>        if (*possible_arg == '=')
>          possible_arg++;
> -      if (*possible_arg != '\0')
> -        {
> -          optarg = possible_arg;
> -          optwhere = 1;
> -        }
> -      else
> -        optarg = NULL;
> +      optarg = (*possible_arg != '\0') ? possible_arg : NULL;
> +      optwhere = 1;
>        break;
>      case REQUIRED_ARG:
>        if (*possible_arg == '=')

Ping? The patch is simple enough, always reset optwhere to 1 after handling
an option with an optional argument.

Cheers,
Peter