Bug in -c CLI option of pg_dump/pg_restore

From: Guillaume Lelarge <guillaume(at)lelarge(dot)info>
To: pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>
Cc: "Jehan-Guillaume (ioguix) de Rorthais" <jgdr(at)dalibo(dot)com>, Gilles Darold <gilles(dot)darold(at)dalibo(dot)com>
Subject: Bug in -c CLI option of pg_dump/pg_restore
Date: 2012-10-13 14:47:06
Message-ID: 1350139626.3053.9.camel@localhost.localdomain
Views: Raw Message | Whole Thread | Download mbox
Lists: pgsql-hackers


One of my colleagues, Jehan-Guillaume de Rorthais, found a weird
behaviour of the "-c" command line option in the pg_restore tool while
doing a training. Here is the following steps he followed:

createdb foo
<adds a few objets in foo>
pg_dump -Fc foo > foo.dump
createdb bar
pg_restore -c -d bar foo.dump

bar contains the same objects as foo (nothing unusual here), but... foo
is no longer present. Actually, if you use the "-c" command line option,
you get a "DROP DATABASE" statement. To me, it feels like a quite
terrible bug.

It's quite easy to reproduce. Just create a database, and use pg_dump
with the "-c" option:

createdb foo
pg_dump -s -c foo | grep DATABASE

and you end up with this:


I tried from 8.3 till 9.2, and only 9.2 has this behaviour.

You'll find attached a patch that fixes this issue. Another colleague,
Gilles Darold, tried it in every possible way, and it works. I'm not
sure the test I added makes it a very good patch, but it fixes the bug.



Attachment Content-Type Size
bug_dropping_database.patch text/x-patch 709 bytes


Browse pgsql-hackers by date

  From Date Subject
Next Message Bruce Momjian 2012-10-13 15:17:32 Re: Adding comments for system table/column names
Previous Message Satoshi Nagayasu 2012-10-13 14:34:12 Re: pg_stat_lwlocks view - lwlocks statistics, round 2