Re: [HACKERS] PostgreSQL 6.4.2: typeidTypeRelid error - Mailing list pgsql-hackers

From David R. Favor
Subject Re: [HACKERS] PostgreSQL 6.4.2: typeidTypeRelid error
Date
Msg-id 36E3CF4A.5CA7C23@austin.ibm.com
Whole thread Raw
In response to PostgreSQL 6.4.2: typeidTypeRelid error  (Erik Riedel <riedel+@CMU.EDU>)
List pgsql-hackers
Erik Riedel wrote:
> 
> I'd like to re-ask on this thread of a few days ago because I suspect
> there is a deeper problem here that changing compilers simply "hides" in
> some way.
> 
> I have a similar problem on Digital UNIX 4.0d with postgresql 6.4.2
> compiled with vendor CC (V5.6 using "cc -std1" in compile).
> 
> % mv ${PGDATA} ${PGDATA}.foo
> % mkdir $PGDATA
> % initdb
> % createdb test
> % psql test
> test=> create table val (v_num decimal);
> test=> \d val
> ERROR:  typeidTypeRelid: Invalid type - oid = 0
> 
> I have the same problem in an existing database that actually has data
> in it - queries (seem to) work ok, but using psql to examine tables and
> such is impossible.
> 
> The last discussion of this was on AIX 4.3.2 compiled with gcc, and the
> fix (if I read the resolution of that thread properly) was to use the
> vendor xlC on AIX instead.

This is correct. I changed two items.
  1) switched from gcc to xlc
  2) changed -qchars=unsigned (was -qchars=signed)

Most of the regression tests failed with -qchars=signed and all passed without
so much as a whimper when -qchars=unsigned. Don't know why this happened, or if
it makes any difference on your platform, but that's what AIX needs to work.


pgsql-hackers by date:

Previous
From: The Hermit Hacker
Date:
Subject: Re: [HACKERS] 6.4.3 release
Next
From: Vadim Valiakhmetov
Date:
Subject: Postgress case insensitive search on FreeBSD3.0