Re: [HACKERS] odd pg_dump output?

> >Thanks Tatsuo! But these are older problems, not new breakage from me,
> >right? Have you had a chance to test the latest pg_dump with your
> >patches? I'm still concerned that I didn't test pg_dump with a
> >regression test dump/reload/dump sequence (by comparing the two dump
> >files).
>
> I ran pg_dump with the regression DB, then did psql to reload.
> seems there are some problems in the output from pg_dump.
>
> BTW, this is FreeBSD 2.2.6 and one of the regression gets dump core
> (select_having).
> --
> Tatsuo Ishii
> t-ishii(at)sra(dot)co(dot)jp
>
> CREATE TYPE widget ( internallength = 24, externallength = -1, input = 144673(widget_in), output = 144674(widget_out), send = 144674(widget_out), receive = 144673(widget_in), default = '-');
> ERROR: parser: parse error at or near "("
> CREATE TYPE _widget ( internallength = -1, externallength = -1, input = 750(array_in), output = 751(array_out), send = 751(array_out), receive = 750(array_in), default = '-');
> ERROR: parser: parse error at or near "("

Ah, I see now. It is even more complicated that I thought. Creating a
type requires you to specify a valud regproc value WITHOUT QUOTES.
Tricky. My fix should do it. It will appear as widget_in_144673. You
can optionally just specify the object id. Output will always show
both. I have to use an undercore, because that is the only way to make
them valid identifiers.

--
Bruce Momjian | 830 Blythe Avenue
maillist(at)candle(dot)pha(dot)pa(dot)us | Drexel Hill, Pennsylvania 19026
+ If your life is a hard drive, | (610) 353-9879(w)
+ Christ can be your backup. | (610) 853-3000(h)