Re: Errors when restoring backup created by pg_dumpall

On 12/9/24 14:14, PopeRigby wrote:
> On 12/7/24 11:58, David G. Johnston wrote:
>> On Sat, Dec 7, 2024 at 12:25 PM PopeRigby <poperigby(at)mailbox(dot)org> wrote:
>>
>>
>> It actually looks like setting those all to have public fixed all the
>> errors, including the one with lldap. So, how can I get it to not put
>> public there automatically for next time?
>>
>>
>> I assume you mean "get it to put public there" (i.e., the "not" is a typo)
>>
>> You cannot. The security team has decided to not permit an opt-in
>> bypass of the lock-downs implemented to fix CVE-2018-1058.
>>
>> Your only real choice at the moment is to replace the function call in
>> the generated expression with a custom function and in that custom
>> function's create function command attach a "set search_path to
>> public" clause.  That will prevent inlining and also ensure the public
>> schema is in the search_path when executing the public.ll_to_earth
>> function call.  With that in place the empty search_path in the dump
>> file will no longer matter.
>>
>> David J.
>>
> Yeah, that was a typo. It seems weird that this behavior would be broken
> by default though, is there anything that could fix it upstream?
>

You could file an issue here:

https://www.postgresql.org/account/login/?next=/account/submitbug/

Ask if the developers could use the mechanisms available here:

https://www.postgresql.org/docs/current/extend-extensions.html#EXTEND-EXTENSIONS-RELOCATION

to schema qualify the objects in the extension.

Not sure if that will fly or not, but it is worth a shot.

--
Adrian Klaver
adrian(dot)klaver(at)aklaver(dot)com