"Dave Page" <dpage@postgresql.org> writes:
> On 18/01/2008, Peter Eisentraut <peter_e@gmx.net> wrote:
>> I didn't follow how the user got into this mess, so I don't know whether the
>> suggestion "you need to initdb" is appropriate.
> I would think not, as you almost certainly must be doing a file level
> restore of the data directory to get into this state and therefore
> probably want to keep your data.
You could argue that that's true if there's an incompatible pg_control
file there at all; I'm not sure why wrong-endianness is different from
wrong-version or wrong-datetime-option or anything else.
IIRC the HINT to initdb was originally kind of pointed at developers
who'd just downloaded a new CVS update with a new catversion number.
It might not be so appropriate for the field. The worst case scenario
would be someone blindly following the hint and blowing away their old
data ...
regards, tom lane