Обсуждение: Adding clarification on extension_control_path when extension is present in multiple folders
When playing with the new GUC extension_control_path, I found that the paths are evaluated one by one (like search_path)and if an extension is present in multiple paths, only the on in the first path will be considered.I think this is totally fine, but I'd like to adda documentation patch to explain this (attached file). What do you think ? Pierrick, Here is an example to understand that I can only create extension in version 2.0. postgres=# show extension_control_path ; extension_control_path ----------------------------------------------------------------------------- $system:/extensions/meilleureextension/share:/extensions/monextension/share (1 row) postgres=# select * from pg_available_extensions where name like '%mon%'; name | default_version | installed_version | comment --------------+-----------------+-------------------+--------------- monextension | 2.0 | | Meilleure monextension | 1.0 | | Mon extension (2 rows) postgres=# select * from pg_available_extensions where name like '%mon%'; name | default_version | installed_version | comment --------------+-----------------+-------------------+--------------- monextension | 2.0 | | Meilleure monextension | 1.0 | | Mon extension (2 rows) postgres=# create extension monextension version "1.0" ; ERROR: extension "monextension" has no installation script nor update path for version "1.0" postgres=# create extension monextension ; CREATE EXTENSION postgres=# select * from pg_available_extensions where name like '%mon%'; name | default_version | installed_version | comment --------------+-----------------+-------------------+--------------- monextension | 2.0 | 2.0 | Meilleure monextension | 1.0 | 2.0 | Mon extension (2 rows)
Вложения
Re: Adding clarification on extension_control_path when extension is present in multiple folders
От
"Matheus Alcantara"
Дата:
On Fri Sep 5, 2025 at 10:53 AM -03, Pierrick wrote: > When playing with the new GUC extension_control_path, I found that the > paths are evaluated one by one > (like search_path)and if an extension is present in multiple paths, only > the on in the first path > will be considered.I think this is totally fine, but I'd like to adda > documentation patch to explain > this (attached file). > > What do you think ? > Hi, thanks for working on this! Just to mention that I have this on my plate, I'm just waiting for a conclusion on discussion at [1] so that I can make a proper review here. Thanks! [1] https://www.postgresql.org/message-id/8f5a0517-1cb8-4085-ae89-77e7454e27ba@dalibo.com -- Matheus Alcantara