Re: extension_control_path: ignore empty elements and deduplicate directories

From: Andres Freund <andres(at)anarazel(dot)de>
To: Chao Li <li(dot)evan(dot)chao(at)gmail(dot)com>
Cc: PostgreSQL-development <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: extension_control_path: ignore empty elements and deduplicate directories
Date: 2026-11-16 15:10:44
Message-ID: f7dfhd43tuzdcgbitfbnfgqt3tjfbm2t3jg27sryuzbnuj4uuq@ubiahyeigfcs
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi,

On 2026-11-03 13:21:17 +0800, Chao Li wrote:
> 1. Empty path elements are silently interpreted as "/extension"
>
> Say a user sets extension_control_path = '$system:', it will be interpreted as two paths: $system and an empty string, and the empty string will be expanded to "/extension". Since "/extension" is an absolute path, it will be searched for extensions. If the host machine doesn't have "/extension", the path will be silently ignored; but if the directory happens to exist, it will be searched for extensions, which doesn't sound desirable, because the user might not be aware of that at all.
>
> Adding a trailing colon is an easy mistake to make, but silently searching "/extension" is unintended behavior. So I think we should silently discard empty path elements. An entirely empty string ('') is interpreted as $system as documented, and that behavior should be retained.

I don't like silently ignoring user mistakes. Either we should make it an
error or accept it. Which user would be served by silently reinterpreting
what they might have meant?

> 2. It may read the same directory more than once
>
> Say a user sets extension_control_path = '$system:$system', that's an
> obvious duplicate path, but the current implementation will scan $system
> twice when listing available extensions, and only deduplicate extension
> names later.

> There is no correctness problem here, but deduplicating search paths may
> avoid unnecessary directory reads. I think this is a relatively low-priority
> but useful optimization.

Don't do that then. It's not like this is anywhere near performance relevant
paths.

Greetings,

Andres Freund

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Tom Lane 2026-11-16 15:20:34 Re: [PATCH] Normalize numeric/bigint scientific notation after rounding
Previous Message Fujii Masao 2026-11-16 14:54:45 Re: [PATCH] Check interrupts while draining received WAL