Re: Proposal: pg_confcheck - syntactic & semantic validation of postgresql configuration files

On 10/14/15 1:50 PM, Andres Freund wrote:
> On October 14, 2015 7:45:53 PM GMT+02:00, Alvaro Herrera <alvherre(at)2ndquadrant(dot)com> wrote:
>> Amir Rohan wrote:
>>
>>> it does fail the "dependent options" test:
>>> $ postgres -C "archive_mode"
>>> on
>>> $ postgres -C wal_level
>>> minimal
>>>
>>> no errors, great, let's try it:
>>> $ pg_ctl restart
>>>
>>> FATAL: WAL archival cannot be enabled when wal_level is "minimal"
>>
>> This complaint could be fixed we had a --check-config that runs the
>> check hook for every variable, I think.

I think that would be widely useful and fairly uncontroversial.

> The problem is that this, and some others, invariant is checked outside the GUC framework. Which we should probably change, which IIRC will require some new infrastructure.

In the extreme, this problem is not solvable (halting problem). If we
had a dry-run checking functionality, there would probably be more
incentive to normalize many of the common dependency cases into a
declarative system.