Kyotaro HORIGUCHI <horiguchi.kyotaro@lab.ntt.co.jp> writes:
> At Wed, 25 May 2016 10:11:28 -0400, Tom Lane <tgl@sss.pgh.pa.us> wrote in <24577.1464185488@sss.pgh.pa.us>
>> The only case
>> that is certain to work is switches before non-switch arguments, and so
>> we should not give any documentation examples in the other order. On a
>> platform using GNU getopt(), getopt() will rearrange the argv[] array to
>> make such cases work, but non-GNU getopt() doesn't do that (and I don't
>> really trust the GNU version not to get confused, either).
> Yeah, I knew it after reading port/getopt_long.c. But the error
> message seems saying nothing about that (at least to me). And
> those accumstomed to the GNU getopt's behavior will fail to get
> the point of the message. The documentation also doesn't
> explicitly saying about the restriction; it is only implicitly
> shown in synopsis. How about something like the following
> message? (The patch attached looks too dependent on the specific
> behavior of our getopt_long.c, but doing it in getopt_long.c also
> seems to be wrong..)
It's not a bad idea to try to improve our response to this situation,
but I think this patch needs more work. I wonder in particular why
you're not basing the variant error message on the first unprocessed
argument starting with '-', rather than looking at the word before it.
Another thought is that the message is fairly unhelpful because it does
not show where in the arguments things went wrong. Maybe something
more like
if (argv[optind][0] == '-') fprintf(stderr, _("%s: misplaced option \"%s\": options must appear before non-option
arguments\n"), progname, argv[optind]);else // existing message
In any case, if we wanted to do something about this scenario, we should
do it consistently across all our programs, not just pg_dump. I count
25 appearances of that "too many command-line arguments" message...
regards, tom lane