Re: pg_dump multi VALUES INSERT

On Fri, 22 Feb 2019 at 14:40, David Rowley  wrote:
>
> On Tue, 19 Feb 2019 at 02:34, Surafel Temesgen  wrote:
> > I see that there are already a test for zero column table in test_fourth_table_zero_col
> > and if am not wrong table_index_stats is empty table
>
> Maybe Fabien would like to see a test that dumps that table with
> --rows-per-insert= to ensure the output remains
> as the other test.  I think it might be a good idea too.

This patch was failing to build due to the new extra_float_digits
option that's been added to pg_dump.  It was adding an additional case
for 8 in the getopt_long switch statement. In the attached, I've
changed it to use value 9 and 10 for the new options.

I also went ahead and added the zero column test that Fabien
mentioned. Also added the missing backslash from the other test that
had been added.

Fabien also complained about some wording in the docs. I ended up
changing this a little bit as I thought the change was a little
uninformative about what rows won't be restored when an INSERT fails.
I've changed this so that it mentions that all rows which are part of
the same INSERT command will fail in the restore.

I think this can be marked as ready for committer now, but I'll defer
to Fabien to see if he's any other comments.

-- 
 David Rowley                   http://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services