Обсуждение: pg_restore -F bug

This patch fixes a bug in the error message emitted by pg_restore on an
incorrect -F argument: write_msg() expects its first parameter to be a
"module name", not the format string. Patch applied to HEAD. Is this
worth backporting?

BTW, is there a reason that pg_restore seems to accept 'f', 't', or 'c'
parameters to this argument, but the documentation only documents 't'
and 'c'?

-Neil
Index: src/bin/pg_dump/pg_restore.c
===================================================================
RCS file: /var/lib/cvs/pgsql/src/bin/pg_dump/pg_restore.c,v
retrieving revision 1.69
diff -c -r1.69 pg_restore.c
*** src/bin/pg_dump/pg_restore.c    22 Feb 2005 04:39:38 -0000    1.69
--- src/bin/pg_dump/pg_restore.c    29 Apr 2005 06:46:34 -0000
***************
*** 313,319 ****
                  break;

              default:
!                 write_msg("unrecognized archive format '%s'; please specify 't' or 'c'\n",
                            opts->formatName);
                  exit(1);
          }
--- 313,319 ----
                  break;

              default:
!                 write_msg(NULL, "unrecognized archive format '%s'; please specify 't' or 'c'\n",
                            opts->formatName);
                  exit(1);
          }

Neil Conway wrote:
> This patch fixes a bug in the error message emitted by pg_restore on an
> incorrect -F argument: write_msg() expects its first parameter to be a
> "module name", not the format string. Patch applied to HEAD. Is this
> worth backporting?
>
> BTW, is there a reason that pg_restore seems to accept 'f', 't', or 'c'
> parameters to this argument, but the documentation only documents 't'
> and 'c'?

I didn't think pg_restore could use a text file, and I think it emits an
error if you try.  Maybe they accept the flag so they can throw a
meaningful error later.

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Neil Conway <neilc@samurai.com> writes:
> This patch fixes a bug in the error message emitted by pg_restore on an
> incorrect -F argument: write_msg() expects its first parameter to be a
> "module name", not the format string. Patch applied to HEAD. Is this
> worth backporting?

Certainly --- particularly if the error makes it dump core, as seems
likely (haven't tried it).

> BTW, is there a reason that pg_restore seems to accept 'f', 't', or 'c'
> parameters to this argument, but the documentation only documents 't'
> and 'c'?

I think the 'f' option is only intended for debugging purposes ...
Philip would know ...

            regards, tom lane

Tom Lane wrote:
> Certainly --- particularly if the error makes it dump core, as seems
> likely (haven't tried it).

Ok, backpatched to stable branches back to 7.2

-Neil