Bart Blanquart wrote:
> On 02/19/09 18:25, Glenn Faden wrote:
>> This looks good to me. It probably could have been coded differently
>> using getopt, but since that wasn't used in the original code, I think
>> your changes are acceptable.
>
> I originally coded it using getopt, but that fails to handle "auths -ab
> username" correctly, as it can't tell if the "b" is the argument to "-a"
> or a (non-existant) option.
This is a significant issue. Whether or not you use getopt() in the
implementation, the command syntax should be getopt-compliant. In other
words, the example above should check if the user has the authorization
called "b". The easiest way to achieve getopt-compliance, of course, is
to use getopt().
>
> This also meant it becomes tricky to handle "auths -asolaris.something
> username" correctly: the switch(argc) would fall to the wrong case (as
> argc==3 but it should fall to case 4).
That's why you want to use optind to decrement argc and increment argv
*after* you call getopt(). That way you get correct results with or
without the space between the option and argument.
>
> If there's an easy way to handle this I can swap out lines 82-92 for a
> getopt based command line parser (and update 93-99 to match).
>
> Bart
I didn't review your code; my comments are based on the email thread.
But I'm 99% sure you should be using getopt().
Scott
--
Scott Rotondo
Principal Engineer, Solaris Security Technologies
President, Trusted Computing Group
Phone/FAX: +1 408 850 3655 (Internal x68278)
