CREATE EXTENSION forces an library initialization - is it bug?

Pavel Stehule <pavel.stehule@gmail.com> writes:

I had to fix plpgsql_check issue
https://github.com/okbob/plpgsql_check/issues/155

The problem is in execution of _PG_init() in CREATE EXTENSION time.

It is a problem for any extension that uses plpgsql debug API, because it
is quietly activated.

Is it necessary?

Yes, I think so. If the extension has any C functions, then when its
script executes those CREATE FUNCTION commands then the underlying
library will be loaded (so we can check that the library is loadable
and the functions really exist). That's always happened and I do not
think it is negotiable.

regards, tom lane

pá 29. 9. 2023 v 20:14 odesílatel Tom Lane <tgl@sss.pgh.pa.us> napsal:

Pavel Stehule <pavel.stehule@gmail.com> writes:

I had to fix plpgsql_check issue
https://github.com/okbob/plpgsql_check/issues/155

The problem is in execution of _PG_init() in CREATE EXTENSION time.

It is a problem for any extension that uses plpgsql debug API, because it
is quietly activated.

Is it necessary?

Yes, I think so. If the extension has any C functions, then when its
script executes those CREATE FUNCTION commands then the underlying
library will be loaded (so we can check that the library is loadable
and the functions really exist). That's always happened and I do not
think it is negotiable.

ok

thank you for info