Обсуждение: fix for pg_dump (caused by array output syntax changes)

don't know if it's too simple, and didn't solved for quite a period of time.
But I still need pg_dump to do my backup, so do a simple(but ugly) hack.
Wish it helps.

regards laser
*** pg_dump.c    2004-09-06 23:52:08.000000000 +0800
--- pg_dump.c.new    2004-09-06 23:55:30.000000000 +0800
***************
*** 1527,1533 ****
      {
          appendPQExpBuffer(query, "SELECT tableoid, oid, nspname, "
                            "(select usename from pg_user where nspowner = usesysid) as usename, "
!                           "nspacl, "
                            "(SELECT spcname FROM pg_tablespace t WHERE t.oid = nsptablespace) AS nsptablespace "
                            "FROM pg_namespace");
      }
--- 1527,1533 ----
      {
          appendPQExpBuffer(query, "SELECT tableoid, oid, nspname, "
                            "(select usename from pg_user where nspowner = usesysid) as usename, "
!                           "nspacl[0:2], "
                            "(SELECT spcname FROM pg_tablespace t WHERE t.oid = nsptablespace) AS nsptablespace "
                            "FROM pg_namespace");
      }
***************
*** 2380,2386 ****
           */
          appendPQExpBuffer(query,
                            "SELECT c.tableoid, c.oid, relname, "
!                           "relacl, relkind, relnamespace, "
                            "(select usename from pg_user where relowner = usesysid) as usename, "
                            "relchecks, reltriggers, "
                            "relhasindex, relhasrules, relhasoids, "
--- 2380,2386 ----
           */
          appendPQExpBuffer(query,
                            "SELECT c.tableoid, c.oid, relname, "
!                           "relacl[0:1], relkind, relnamespace, "
                            "(select usename from pg_user where relowner = usesysid) as usename, "
                            "relchecks, reltriggers, "
                            "relhasindex, relhasrules, relhasoids, "

Weiping <laser@qmail.zhengmai.net.cn> writes:
> don't know if it's too simple, and didn't solved for quite a period of time.
> But I still need pg_dump to do my backup, so do a simple(but ugly) hack.

This isn't going to be applied because pg_dump is not broken ... at
least not against CVS tip.  You apparently are running a pre-beta
backend that numbered ACL entries from zero.

            regards, tom lane

>This isn't going to be applied because pg_dump is not broken ... at
>least not against CVS tip.  You apparently are running a pre-beta
>backend that numbered ACL entries from zero.
>
>
>
Got it: The code are up-to-date, but I didn't initdb,
so the system catalog didn't update thus the pg_dump broken.

thanks and regards laser