Re: Add Postgres module info

Andrei Lepikhov <lepihov@gmail.com>

From: Andrei Lepikhov <lepihov@gmail.com>
To: Alexander Korotkov <aekorotkov@gmail.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Yurii Rashkovskii <yrashk@omnigres.com>, PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2025-03-02T18:35:35Z
Lists: pgsql-hackers
On 17/2/2025 02:41, Alexander Korotkov wrote:
> On Mon, Dec 16, 2024 at 7:02 AM Andrei Lepikhov <lepihov@gmail.com> wrote:
>> On 12/13/24 10:17, Tom Lane wrote:
>>> There's nothing stopping a field of the magic block from being
>>> a "const char *" pointer to a string literal.
>> Ok, See v.2 in attachment.
> 
> Generally, the patch looks good to me.  I have couple of questions.
> 
> 1) Is it intended to switch all in-core libraries to use PG_MODULE_MAGIC_EXT()?
I haven't such intention. Just wanted to demonstrate how it might work.

> 2) Once we have module version information, it looks natural to
> specify the required version for dependant objects, e.g. SQL-funcions
> implemented in shared libraries.  For instance,
> CREATE FUNCTION ... AS 'MODULE_PATHNAME' LANGUAGE C module_version >= '1.0';
Just to be clear. You want this stuff to let the core manage situations 
of stale binaries and throw an error like the following:
"No function matches the given name, argument types and module version"
Do I understand you correctly?
It may make sense, but I can't figure out a use case. Could you describe 
at least one example?

-- 
regards, Andrei Lepikhov



Commits

  1. Avoid mixing designated and non-designated field initializers.

  2. Use PG_MODULE_MAGIC_EXT in our installable shared libraries.

  3. Introduce PG_MODULE_MAGIC_EXT macro.