Tom Lane wrote:
> Robert Treat <xzilla@users.sourceforge.net> writes:
>
>> ISTR being unconvinced by the pg_restore arguments, but as I think about it
>> some more, for someone to set statement_timeout on a production system, and
>> then have that be blindly overridden by any random pg_dump user seems a bit
>> unfair. pg_dump is not only used as a backup tool, it is also used as a
>> general user tool (for example, pgadmin calls pg_dump if you want to see a
>> tables schema).
>>
>
> So? In those usages, it's not going to run long enough to have a
> statement_timeout problem anyway.
>
> When there is a data dump involved, you still have to defend the
> proposition that it's okay for pg_dump to deliver a bad dump if
> statement_timeout hits it. I can't accept that.
>
>
>
I agree.
What is more, the solution to the non-dump uses of pg_dump is to put
that functionality in a library where clients can call it directly
rather than using pg_dump.
cheers
andrew