At Thu, 15 Jun 2023 21:58:28 -0700, Nathan Bossart <nathandbossart@gmail.com> wrote in
> On Fri, Jun 16, 2023 at 10:30:09AM +0900, Michael Paquier wrote:
> > On Thu, Jun 15, 2023 at 05:09:59PM -0700, Nathan Bossart wrote:
> >> I've attached a new version of the patch that omits the
> >> POSIXLY_CORRECT stuff.
> >
> > This looks OK at quick glance, though you may want to document at the
> > top of getopt_long() the reordering business with the non-options?
>
> I added a comment to this effect in v3. I also noticed that '-' wasn't
> properly handled as a non-option, so I've tried to fix that as well.
(Honestly, the rearrangement code looks somewhat tricky to grasp..)
It doesn't work properly if '--' is involved.
For example, consider the following options (even though they don't
work for the command).
psql -t -T hoho -a hoge -- -1 hage hige huge
After the function returns -1, the argv array looks like this. The
"[..]" indicates the position of optind.
psql -t -T hoho -a -- [-1] hage hige huge hoge
This is clearly incorrect since "hoge" gets moved to the end. The
rearrangement logic needs to take into account the '--'. For your
information, the glibc version yields the following result for the
same options.
psql -t -T hoho -a -- [hoge] -1 hage hige huge
regards.
--
Kyotaro Horiguchi
NTT Open Source Software Center