Re: Re: [COMMITTERS] pgsql: Don't override arguments set via options with positional argumen

On 04/18/2012 10:03 AM, Peter Eisentraut wrote:
> On ons, 2012-04-18 at 09:53 -0400, Tom Lane wrote:
>> Peter Eisentraut<peter_e(at)gmx(dot)net> writes:
>>> My vote is to revert this altogether and leave it be. In the
>>> alternative, make it an error.
>> You mean in HEAD too? I don't agree with that, for sure. What this
>> patch is accomplishing is to make sure that the less-commonly-used
>> programs have similar command-line-parsing behavior to psql and pg_dump,
>> where we long ago realized that failure to check this carefully could
>> result in very confusing behavior. (Especially on machines where
>> getopt is willing to rearrange the command line.)
> OK, if you care strongly about that, make it an error. But don't just
> ignore things.

It won't be ignored. It will be caught by the "too many arguments" logic.

The case where repeated arguments should be disallowed is a similar but
different case that probably demands a much larger patch. I don't think
its existence militates against this fix, however.

>
>> I agree with Andrew that this is a bug fix. I can see the argument
>> for not applying it to back branches, but not for declaring that it's
>> not a bug.
> We shouldn't be backpatching things that are merely confusing. It works
> as designed at the time, after all. Improvements belong in master.
>

If it was really intended to work this way then that's a piece of very
poor design, IMNSHO. It looks to me much more like it was just an
oversight.

I don't have terribly strong feelings about this, since we've not had
lots of complaints over the years, so I'll revert it in the back branches.

cheers

andrew