Re: BUG #5487: dblink failed with 63 bytes connection names

On 01/06/10 05:55, Takahiro Itagaki wrote:
> "Takahiro Itagaki"<itagaki(dot)takahiro(at)oss(dot)ntt(dot)co(dot)jp> wrote:
>>
>> Contib/dblink module seems to have a bug in handling
>> connection names in NAMEDATALEN-1 bytes.
>
> Here is a patch to fix the bug. I think it comes from wrong usage
> of snprintf(NAMEDATALEN - 1). It just copies 62 bytes + \0.
>
> In addition, it should be safe to use pg_mbcliplen() to truncate
> extra bytes in connection names because we might return invalid
> text when a multibyte character is at 62 or 63 bytes.

Hmm, seems that dblink should call truncate_identifier() for the
truncation, to be consistent with truncation of table names etc.

I also spotted this in dblink.c:

> /* first gather the server connstr options */
> if (strlen(servername) < NAMEDATALEN)
> foreign_server = GetForeignServerByName(servername, true);

I think that's wrong. We normally consistently truncate identifiers at
creation and at use, so that if you create an object with a very long
name and it's truncated, you can still refer to it with the untruncated
name because all such references are truncated too.

--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com