From: Andrei Lepikhov <lepihov@gmail.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Michael Paquier <michael@paquier.xyz>
Cc: Alexander Korotkov <aekorotkov@gmail.com>
Cc: Yurii Rashkovskii <yrashk@omnigres.com>
Cc: PostgreSQL-development <pgsql-hackers@postgresql.org>
Subject: Re: Add Postgres module info
Date: Sun, 23 Mar 2025 10:44:16 +0100
Message-ID: <d056f9fc-b29c-48c4-81ae-d81a5bb7b5f2@gmail.com> (raw)
In-Reply-To: <441661.1742683797@sss.pgh.pa.us>
References: <dd4d1b59-d0fe-49d5-b28f-1e463b68fa32@gmail.com>
<506788.1733941263@sss.pgh.pa.us>
<2c215c98-f713-4c5b-b21a-a4b8fa41890b@gmail.com>
<556341.1733967371@sss.pgh.pa.us>
<0844622e-9967-409a-8038-470a36f1039c@gmail.com>
<CAG=VW14ZmKDcxYs7irSvhVS+JeDg2W_gyR++L2GXEpzVVYQ+Yw@mail.gmail.com>
<8b81a3c0-761f-4cfb-97c6-68819fe4d282@gmail.com>
<982748.1734059835@sss.pgh.pa.us>
<0e82bf8c-ae70-498d-861e-dba2bb154cad@gmail.com>
<CAPpHfduL6=LdHco_+wCk=VQyZzEd5hjLi46jtV2gQbDrpXf5Gw@mail.gmail.com>
<Z7Kmann0SyoP7kDr@paquier.xyz>
<29d29c3e-ac9e-48cc-87b4-b1cfd8c2ede8@gmail.com>
<992ec366-5c0d-4f23-a02e-30e663613f44@gmail.com>
<4476e8cd-6427-4470-98a8-8d61b82d8b1e@gmail.com>
<441661.1742683797@sss.pgh.pa.us>
On 3/22/25 23:49, Tom Lane wrote:
> I spent awhile reviewing the v5 patch, and here's a proposed v6.
> Some notes:
>
> * I didn't like depending on offsetof(Pg_magic_struct, module_extra)
> to determine which parts of the struct are checked for compatibility.
> It just seems way too easy to break that with careless insertion
> of new fields, and such breakage might not cause obvious failures.
> I think the right thing is to break out the ABI-checking fields as
> their own sub-struct, rather than breaking out the new fields as a
> sub-struct.
Agree. It is a clear approach. I like it.
>
> * I renamed the inquiry function to pg_get_loaded_modules, since
> it only works on loaded modules but that's hardly clear from the
> previous name.
+1
>
> * It is not clear to me what permission restrictions we should put
> on pg_get_loaded_modules, ...
I vote for the idea of stripping the full path to just a filename. My
initial use cases were:
1. User reports the issue and need to provide me all loaded modules at
the moment of query execution. Higher privileges needs administrative
procedures that is a long way and not all the time possible.
2. A module needs to detect another loaded module - it is not a frequent
case so far, but concurrency on queryId with pg_stat_statements is at
least one of my examples happening sometimes.
Also, permissions here should be in agreement with permissions on
pg_available_extensions(), right?
>
> * I didn't like anything about the test setup. ...
Ok, thanks. I just played with alternatives.
>
> Still TBD:
>
> * I'm not happy with putting pg_get_loaded_modules into dfmgr.c.
> It feels like the wrong layer to have a SQL-callable function,
> and the large expansion in its #include list is evidence that we're
> adding functionality that doesn't belong there. But I'm not quite
> sure where to put it instead. Also, the naive way to do that would
> require exporting DynamicFileList which doesn't feel nice either.
> Maybe we could make dfmgr.c export some sort of iterator function?
I just attempted to reduce number of exported objects here. If it is ok
to introduce an iterator, the pg_get_loaded_modules() may live in
extension.c
>
> * Should we convert our existing modules to use PG_MODULE_MAGIC_EXT?
> I'm mildly in favor of that, but I think we'd need some automated way
> to manage their version strings, and I don't know what that ought to
> look like. Maybe it'd be enough to make all the in-core modules use
> PG_VERSION as their version string, but I think that might put a dent
> in the idea of the version strings following semantic versioning
> rules.
Yes, additional burden to bump version string was a stopper for me to
propose such a brave idea.
--
regards, Andrei Lepikhov
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients from your mail client; the link fills in
the headers below: reply via email
Some clients, webmail especially, drop In-Reply-To and References from
a link, and the reply then starts a new thread: use the mbox below.
To: pgsql-hackers@postgresql.org
Cc: lepihov@gmail.com, tgl@sss.pgh.pa.us, michael@paquier.xyz, aekorotkov@gmail.com, yrashk@omnigres.com
Subject: Re: Add Postgres module info
In-Reply-To: <d056f9fc-b29c-48c4-81ae-d81a5bb7b5f2@gmail.com>
References: <dd4d1b59-d0fe-49d5-b28f-1e463b68fa32@gmail.com> <506788.1733941263@sss.pgh.pa.us> <2c215c98-f713-4c5b-b21a-a4b8fa41890b@gmail.com> <556341.1733967371@sss.pgh.pa.us> <0844622e-9967-409a-8038-470a36f1039c@gmail.com> <CAG=VW14ZmKDcxYs7irSvhVS+JeDg2W_gyR++L2GXEpzVVYQ+Yw@mail.gmail.com> <8b81a3c0-761f-4cfb-97c6-68819fe4d282@gmail.com> <982748.1734059835@sss.pgh.pa.us> <0e82bf8c-ae70-498d-861e-dba2bb154cad@gmail.com> <CAPpHfduL6=LdHco_+wCk=VQyZzEd5hjLi46jtV2gQbDrpXf5Gw@mail.gmail.com> <Z7Kmann0SyoP7kDr@paquier.xyz> <29d29c3e-ac9e-48cc-87b4-b1cfd8c2ede8@gmail.com> <992ec366-5c0d-4f23-a02e-30e663613f44@gmail.com> <4476e8cd-6427-4470-98a8-8d61b82d8b1e@gmail.com> <441661.1742683797@sss.pgh.pa.us> <d056f9fc-b29c-48c4-81ae-d81a5bb7b5f2@gmail.com>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox