Tom wrote:
> Hmm, I'd expect it to work --- as you say, it's not hugely surprising if
> there are bugs in there, but I wasn't aware of any. Can you look more
> closely and find out what's going on --- for example, what shows up in
> the postmaster log with this case:
>
> > pg_restore: restoring data for table d_ref
> > pg_restore: [archiver (db)] error returned by PQputline
> > pg_restore: *** aborted because of error
>
The last dozen lines or so are... (sorry if there are any typos,
copy/paste not working).
COPY d_ref (ref_id, ref_host, ref_path, ref_query) from STDIN;
IN: pg_parse_query (postgres.c:393)
LOG: Checkpoint segments are being created too frequently (3 secs)
Consider increasing CHECKPOINT SEGMENTS
IN: sigusr1_handler (postmaster.c:2504)
ERROR: CopyReadAttribute: Literal carriage return data value
found in input that has newline termination; use \r
CONTEXT: COPY FROM, line 146178
IN: CopyReadAttribute (copy.c:1650)
FATAL: Socket command type
unknown
IN: SocketBackend (postgres.c:294)
> One thing to watch out for is that cvs-tip pg_dump won't talk to a
> pre-7.4 server at all (if it does, you've got some shared-library
> misalignment). That will have to be fixed, but it's not on
> the critical path at the moment.
That could explain everything. I think (but can't promise) that
I had the cvs-tip pg_dump. However it's complaining about '\r's
and I see that dumputils.c in cvs-tip looks like it escapes '\r's.
OTOH, I'm happy to run it again if you need to be sure (takes
about 10 hours, so I'd have results next day).
Ron