Обсуждение: dumpall prob

CREATE UNIQUE INDEX "ethernet_ip_key" on "ethernet" using btree ( "ip" "network_ops" );

was generated by dumpall. "network_ops" apparently don't exist (not sure what
is should be called). Changing to
using btree ( "ip" )
was sufficient to fix, but I don't know what it should be to fix dumpall.

patrimoine=> create unique index "ethernet_ip_key" on "ethernet" using btree ( "ip" );
CREATE
patrimoine=> \d ethernet_ip_key
Index "ethernet_ip_key"Attribute | Type 
-----------+------ip        | inet
unique btree


Cheers,

Patrick

Patrick Welche wrote:
> 
> CREATE UNIQUE INDEX "ethernet_ip_key" on "ethernet"
>   using btree ( "ip" "network_ops" );
> was generated by dumpall.

What version? If your example above is verbatim, then there seems to
be a missing comma in the arguments to the USING clause; if you are
using a recent/current version of pg_dump, then it is not likely
fixed...
                 - Thomas

-- 
Thomas Lockhart                lockhart@alumni.caltech.edu
South Pasadena, California

Thomas Lockhart wrote:
> 
> Patrick Welche wrote:
> > 
> > CREATE UNIQUE INDEX "ethernet_ip_key" on "ethernet"
> >   using btree ( "ip" "network_ops" );
> > was generated by dumpall.
> 
> What version? If your example above is verbatim, then there seems to
> be a missing comma in the arguments to the USING clause; if you are
> using a recent/current version of pg_dump, then it is not likely
> fixed...

Oops - my fault. Obviously I pg_dump'd before installing the new one, so
it's the old one that had the "network_ops" problem.
Funnily enough I thought I would pg_dump with the new one and diff it
against the new, but

% pg_dumpall > db.out2
Connection to database 'List' failed.
pqReadData() -- backend closed the channel unexpectedly.       This probably means the backend terminated abnormally
  before or while processing the request.
 

pg_dump failed on List, exiting


... investigating ...

Cheers,

Patrick

"Patrick Welche" <prlw1@newn.cam.ac.uk> writes:
> CREATE UNIQUE INDEX "ethernet_ip_key" on "ethernet" using btree ( "ip" "network_ops" );
> was generated by dumpall. "network_ops" apparently don't exist (not sure what
> is should be called).

Current sources dump this as "inet_ops".  I think the problem in 6.5.*
was caused by bogus entries in the pg_opclass table (same classname for
inet and cidr types).  You could try reaching in and changing the system
table entries, but it might be safer just to live with manually patching
the dump file until the next release.
        regards, tom lane

> % pg_dumpall > db.out2
> Connection to database 'List' failed.
> pqReadData() -- backend closed the channel unexpectedly.
>         This probably means the backend terminated abnormally
>         before or while processing the request.
> 
> pg_dump failed on List, exiting
> 
> 
> ... investigating ...

Reason:

In pg_dumpall, line 50 is:

psql -l -A -q -t| tr '|' ' ' | grep -v '^template1 ' | \

which outputs:

% psql -l -A -q -t| tr '|' ' ' | grep -v '^template1 '
List of databases
Database Owner
darwin prlw1
...
(5 rows)


so presumably, it tries to open a connection to "List" as user "of"
this implies that "-q" isn't quite as quiet as it could be...

I tried changing it to

psql -l -A -q -t| tr '|' ' ' | egrep -v '(^template1 |^List of databases|^Database Owner| rows)' | \ 

as a work around, but then:

\connect template1 ERROR:  attribute 'prlw1' not found
create database darwin;
\connect darwin ERROR:  attribute 'prlw1' not found

etc

and this is because line 56 wants to read
               where usesysid = \'$DBUSERID\'; \" | \
rather than               where usesysid = $DBUSERID; \" | \


Cheers,

Patrick

> "Patrick Welche" <prlw1@newn.cam.ac.uk> writes:
> > CREATE UNIQUE INDEX "ethernet_ip_key" on "ethernet" using btree ( "ip" "network_ops" );
> > was generated by dumpall. "network_ops" apparently don't exist (not sure what
> > is should be called).
> 
> Current sources dump this as "inet_ops".  I think the problem in 6.5.*
> was caused by bogus entries in the pg_opclass table (same classname for
> inet and cidr types).  You could try reaching in and changing the system
> table entries, but it might be safer just to live with manually patching
> the dump file until the next release.

Fixed in the current source tree.

--  Bruce Momjian                        |  http://www.op.net/~candle maillist@candle.pha.pa.us            |  (610)
853-3000+  If your life is a hard drive,     |  830 Blythe Avenue +  Christ can be your backup.        |  Drexel Hill,
Pennsylvania19026