Bug 456293

Indeed, must be related to my recent fix of other issues with option parsing :O(

It might be worth using http://www.boost.org/doc/libs/1_37_0/doc/html/program_options.html instead of getopt.

*** Bug 463255 has been marked as a duplicate of this bug. ***

This is in fact a bug in glibc. If the optstring starts with "+:", it does not return ':' for 'missing argument' error.

I did some debugging and i suspect this happens because the 'optstring' variable is not incremented on each call to getopt/getopt_long, but only on the first call:

getopt.c:405:

  if (d->optind == 0 || !d->__initialized)
    {
      if (d->optind == 0)
        d->optind = 1;  /* Don't scan ARGV[0], the program name.  */
      optstring = _getopt_initialize (argc, argv, optstring, d,
                                      posixly_correct);
      d->__initialized = 1;
    }

(there's ++optstring in _getopt_initialize)

And as a consequence, the code around getopt.c:1097 returns '?' for optstring == "+:something":

    if (optstring[0] == ':')
      c = ':';
    else
      c = '?';

c is returned later.

Pasky, can you take over from here? :O)

FWIW, i've found similar report also in debian's BTS: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=514137

As explained in the debian bugreport and manual section 25.2.1, this is documented and expected behavior - : is returned for invalid option only if the option string starts with :.

Please read my comment #3 again. I'm pretty sure this _is_ a bug. The debian report was about something else after all, and the comment explaining it is not a bug in their case is also from me.

The bottom line is that you need to be able to use the +/- mode selector together with the colon in the opt string. The code in c#3 suggests it as well.

From my comment in the debian bugreport:
"getopt/getopt_long is actually not supposed to return ':' if the
optstring does not _begin_ with a colon (e.g. ":g:") or a +/- sign
followed by a colon (e.g. "+:g" - this actually has another bug, but
that's not what this report is about)."

Aha, I see!

This is already tracked and fixed upstream: http://sourceware.org/bugzilla/show_bug.cgi?id=11039

I'm backporting that fix to glibc-2.11 branch, I plan to tag another version update very soon and get that to 11.3 yet.

Thanks for the report!