Bill McGonigle <mcgonigle@medicalmedia.com> writes:
> There is one discussion of NAMEDATALEN in the archives, pertaining to
> version 6.4.2. That discussion brings up a few questions/assumptions
> related to getting this to work reliably. I'd love to hear any
> comments/corrections/amplifications:
> 1) It was required that OIDNAMELEN be set to sizeof(Oid) + NAMEDATALEN.
OIDNAMELEN is long gone. You don't have to change anything except
NAMEDATALEN.
> 3) Will psql from another machine fail to work if that machine's pgsql
> hasn't been compiled with the modified MAXDATALEN? Is this the same
> question as (2)?
The reason NAMEDATALEN is in postgres_ext.h is that it's visible to (and
used by) clients as well as the backend. So yes, you'd better recompile
everything. I am not sure what problems you would have with mismatched
clients. If you're lucky, they'll merely truncate your longer names,
and not coredump ...
regards, tom lane
