Date: Sat, 22 Sep 2012 00:15:53 -0400
From: Bruce Momjian <bruce@momjian.us>
To: Rural Hunter <ruralhunter@gmail.com>
Cc: PostgreSQL-development <pgsql-hackers@postgresql.org>,
	Tom Lane <tgl@sss.pgh.pa.us>
Subject: Re: [ADMIN] pg_upgrade from 9.1.3 to 9.2 failed
Message-ID: <20120922041553.GA22183@momjian.us>
References: <5053F896.60005@gmail.com> <20120915180603.GA20907@momjian.us>
	<505557CD.8090304@gmail.com> <20120916171746.GA24907@momjian.us>
	<28666.1347833056@sss.pgh.pa.us>
	<20120917043236.GA20123@momjian.us> <5056AF29.3040903@gmail.com>
	<20120917210723.GA30394@momjian.us>
	<20120918232239.GA2014@momjian.us> <505C307E.20501@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <505C307E.20501@gmail.com>
User-Agent: Mutt/1.5.20 (2009-06-14)

On Fri, Sep 21, 2012 at 05:16:46PM +0800, Rural Hunter wrote:
> >>I am thinking this query needs to be split apart into a UNION where the
> >>second part handles TOAST tables and looks at the schema of the _owner_
> >>of the TOAST table.  Needs to be backpatched too.
> >OK, I am at a conference now so will not be able to write-up a patch
> >until perhaps next week.  You can drop the information schema in the old
> >database and pg_upgrade should run fine.  I will test your failure once
> >I create a patch.
> >
> OK. I will try. I also found some problems on initdb when re-init my
> pg9.2 db.
> 1. initdb doesn't create the pg_log dir so pg can not be started
> after initdb before I create the dir manually.
> 2. The case issue of db charset name. I installed pg9.1 and pg9.2
> with zh_CN.UTF8. But somehow it seems the actual chaset name is
> stored with lowercase 'zh_CN.utf8' during the install. In this case,
> I can run the pg_upgrade without problem since they are both
> lowercase. But when I re-init pg9.2 with option '-E zh_CN.UTF8',
> pg_upgrade will fail and report that encoding/charset mis-match: one
> is uppercase and another is lowercase. If I run initdb with '-E
> zh_CN.utf8', it will tell me there is no such charset in the system.
> I found a workaround to run initdb with '--lc-collate=zh_CN.utf8
> --lc-ctype=zh_CN.utf8 --lc-messages=zh_CN.utf8
> --lc-monetary=zh_CN.utf8 --lc-numeric=zh_CN.utf8
> --lc-time=zh_CN.utf8'. But the case problem is really confusing.

Yes, it sounds very confusing.  I wonder if pg_upgrade should do a
case-insentive comprison of encodings?  Comments?

-- 
  Bruce Momjian  <bruce@momjian.us>        http://momjian.us
  EnterpriseDB                             http://enterprisedb.com

  + It's impossible for everything to be true. +