Decreasing priority as no wheelwar will exist on db-A: php83-pdo_pgsql isn't declared there.

Tests to detect this issue in the future: D3878 / 905fee1ce5b9

This change breaks tests repaired in T2123:

Confirmed for recent package:

OK for WindRiver, will need careful deployment on db-A-001.

Taking this, we've PostgreSQL deinstalled on WindRiver; as we're on a UNIX system, it still runs without any issue as long as it doesn't need to load dynamically a library, accepts connection, serves traffic but software isn't there anymore.

In T1713#33906, @dereckson wrote:

Plan is to create Keruald Elements with bricks to build frameworks, including the Session class.

But with last requirement on Obsidian to split concerns between User and UserRepository, we need to devise an base abstract class or an interface for the session code.

Plan is to create Keruald Elements with bricks to build frameworks, including the Session class.

This blocks new attributes use for other tests.

In T2185#33876, @dereckson wrote:

Each test run produces this:

Nov 11 02:29:33 windriver postgres[63389]: [19-1] 2025-11-11 02:29:33.804 UTC [63389] ERROR:  relation "nonexisting" does not exist at character 13
Nov 11 02:29:33 windriver postgres[63389]: [19-2] 2025-11-11 02:29:33.804 UTC [63389] STATEMENT:  DELETE FROM nonexisting
Nov 11 02:29:34 windriver postgres[63402]: [19-1] 2025-11-11 02:29:34.181 UTC [63402] ERROR:  relation "nonexisting" does not exist at character 15
Nov 11 02:29:34 windriver postgres[63402]: [19-2] 2025-11-11 02:29:34.181 UTC [63402] STATEMENT:  SELECT * FROM nonexisting
Nov 11 02:29:34 windriver postgres[63403]: [19-1] 2025-11-11 02:29:34.209 UTC [63403] ERROR:  relation "nonexisting" does not exist at character 15
Nov 11 02:29:34 windriver postgres[63403]: [19-2] 2025-11-11 02:29:34.209 UTC [63403] STATEMENT:  SELECT * FROM nonexisting
Nov 11 02:29:34 windriver postgres[63408]: [19-1] 2025-11-11 02:29:34.357 UTC [63408] ERROR:  relation "not_existing_table" does not exist
Nov 11 02:29:34 windriver postgres[63408]: [19-2] 2025-11-11 02:29:34.357 UTC [63408] STATEMENT:  TRUNCATE not_existing_table
Nov 11 02:29:34 windriver postgres[63409]: [19-1] 2025-11-11 02:29:34.385 UTC [63409] ERROR:  relation "not_existing" does not exist
Nov 11 02:29:34 windriver postgres[63409]: [19-2] 2025-11-11 02:29:34.385 UTC [63409] STATEMENT:  TRUNCATE not_existing
Nov 11 02:29:34 windriver postgres[63416]: [19-1] 2025-11-11 02:29:34.591 UTC [63416] FATAL:  password authentication failed for user "notexisting"
Nov 11 02:29:34 windriver postgres[63416]: [19-2] 2025-11-11 02:29:34.591 UTC [63416] DETAIL:  Role "notexisting" does not exist.
Nov 11 02:29:34 windriver postgres[63416]: [19-3]       Connection matched pg_hba.conf line 22: "host    all             all             127.0.0.1/32            scram-sha-256"

I guess we could avoid to post PostgreSQL logs directly in /var/log/messages.

Each test run produces this:

Already deployed on db-A.

rOBSIDIAN329bc358fc01 refers to table orbeon_form_data_attach, permission is needed for that one too, but missing in rOPS (was a hotfix on Complector)

Actually, we use the same database name than for MariaDB / MySQl: