6.1 pg_dump core dump - Mailing list pgsql-hackers

From bower@image.Kodak.COM (J Christopher Bower)
Subject 6.1 pg_dump core dump
Date
Msg-id 199807271531.LAA10550@coyote.image.Kodak.COM
Whole thread Raw
Responses Re: [HACKERS] 6.1 pg_dump core dump
List pgsql-hackers
I have posted to the general news group but need a more definitive response.

I am running Postgresql 6.1 on SunOS 5.5.1.  We are presently looking into
upgrading to Postgresql 6.3.2, but are having a problem with pg_dump core
dumping. It looks like the output is OK, but I would like to know for sure.

I read that pg_dump 6.2 should be able to dump a 6.1 database but can not
locate the 6.2 source to try it.

I have tried dumping my database with the pg_dumpall provided from the 6.3.2
tar file,but this also core dumps.

I have tried dumping the 6.1 database with 6.3.2 pg_dump, but get a user
authentication error.

I decided to debug the 6.1 pg_dump source and determined that the core dump
occured dumping tblinfo cleanup.  Specifically when structures allocated for
sequence tables are being cleaned up. I made the following changes in the
getTables function.

    for (i=0;i<ntups;i++) {
        tblinfo[i].oid = strdup(PQgetvalue(res,i,i_oid));
        tblinfo[i].relname = strdup(PQgetvalue(res,i,i_relname));
        tblinfo[i].relarch = strdup(PQgetvalue(res,i,i_relarch));
        tblinfo[i].relacl = strdup(PQgetvalue(res,i,i_relacl));
        tblinfo[i].sequence = (strcmp (PQgetvalue(res,i,i_relkind), "S") == 0);

        /* Local fix - needs to be initialized to zero for sequence tables.
        */

        tblinfo[i].numatts = 0;
        tblinfo[i].attlen = 0;
        tblinfo[i].attlen = NULL;
        tblinfo[i].inhAttrs = NULL;
        tblinfo[i].attnames = NULL;
        tblinfo[i].typnames = NULL;

    }


I would like to know if my changes are appropriate, and how can I get the
correct patch if one is available.  Any input would be appreciated.  Thanks
in advance.

Chris Bower
Software Engineer
Eastman Kodak

pgsql-hackers by date:

Previous
From: Bruce Momjian
Date:
Subject: OR clause issues
Next
From: Vadim Mikheev
Date:
Subject: Re: [HACKERS] OR clause issues