On Jul 20, 2018, at 10:12 AM, Jaap Keuter <[email protected]> wrote:
> While having reviewed extcap documentation [Ref1] I was triggered by bug
> 14982 [Ref2] (usbpcap no longer recognized in 2.9) concerning the addition of
> --extcap-version parameter to the extcap command line. This stated that the
> command line parameter "--extcap-version=2.9” causes a problem with usbpcap.
> The documentation states that in Wireshark 2.9 the command line parameter
> “--extcap-version x.x” has been added. Note the lack of equal sign. Looking
> at the code this was originally so, until change 26772 [Ref3] where the equal
> sign was added due to a problem with some variations of getopt_long().
Yes - if you have a command-line option for which the value is optional, if you
want to work with versions of getopt_long() other than the GNU one, you *must*
specify the option with =:
$ cat optiontest.c
#include <getopt.h>
#include <stdio.h>
static const struct option longopts[] = {
{ "noargument", no_argument, NULL, 'n' },
{ "requiredargument", required_argument, NULL, 'r' },
{ "optionalargument", optional_argument, NULL, 'o' },
{ NULL, 0, NULL, 0 }
};
int
main(int argc, char **argv)
{
int opt;
int i;
while ((opt = getopt_long(argc, argv, "nr:o:", longopts, NULL)) != -1) {
switch (opt) {
case 'n':
printf("noargument specified\n");
break;
case 'r':
printf("requiredargment specified, argument is
\"%s\"\n",
optarg);
break;
case 'o':
if (optarg == NULL)
printf("optionalargument specified, no
argument\n");
else
printf("optionalargment specified, argument is
\"%s\"\n",
optarg);
break;
case ':':
printf("Missing option argument\n");
break;
case '?':
printf("Unknown/ambiguous option\n");
break;
default:
printf("getopt_long returns '%c' (%d)\n", opt, opt);
break;
}
}
argc -= optind;
argv += optind;
for (i = 0; i < argc; i++)
printf("argument %d is \"%s\"\n", i + 1, argv[i]);
return 0;
}
$ gcc -o optiontest optiontest.c
$ ./optiontest
$ ./optiontest --noargument
noargument specified
$ ./optiontest --noargument --requiredargument
noargument specified
optiontest: option `--requiredargument' requires an argument
Unknown/ambiguous option
$ ./optiontest --noargument --requiredargument foo
noargument specified
requiredargment specified, argument is "foo"
$ ./optiontest --noargument --requiredargument=foo
noargument specified
requiredargment specified, argument is "foo"
$ ./optiontest --noargument --optionalargument
noargument specified
optionalargument specified, no argument
$ ./optiontest --noargument --optionalargument foo
noargument specified
optionalargument specified, no argument
argument 1 is "foo"
$ ./optiontest --noargument --optionalargument=foo
noargument specified
optionalargment specified, argument is "foo"
so if we want to allow both --extcap-version without an argument and
--extcap-version with an argument, and we don't want to have to check for
whether, if we have a system-supplied getopt_long(), that getopt_long() parses
--optionalargument foo as "--optionalargument with an argument foo" rather than
"--optionalargument with no argument" and, if it doesn't do that, use our copy
of the GNU getopt_long() (or just use it even if it does), we'll have to
require that the argument to --extcap-version be specified with an =.
> Doesn’t this mean that all other command line parameters with values are to
> be generated with equal sign?
It's a requirement for any parameter where the value is optional.
It's not a requirement for any parameter where the value is required.
> If so this could require changes across the board, changes to documentation
> (docbook, man page) in the least.
Yes.
___________________________________________________________________________
Sent via: Wireshark-dev mailing list <[email protected]>
Archives: https://www.wireshark.org/lists/wireshark-dev
Unsubscribe: https://www.wireshark.org/mailman/options/wireshark-dev
mailto:[email protected]?subject=unsubscribe
