[TMP] Run regression for IBX-12043 (DBAL 4)

[ViniTou]

ibexa/core#800
ibexa/core-persistence#39
ibexa/doctrine-schema#45
https://github.com/ibexa/installer/pull/217
ibexa/doctrine-migrations#6
https://github.com/ibexa/migrations/pull/447
https://github.com/ibexa/taxonomy/pull/438
https://github.com/ibexa/product-catalog/pull/1563
https://github.com/ibexa/fieldtype-page/pull/218
https://github.com/ibexa/connector-payum/pull/44
https://github.com/ibexa/measurement/pull/141
https://github.com/ibexa/workflow/pull/202
ibexa/system-info#90
https://github.com/ibexa/discounts/pull/352
https://github.com/ibexa/site-factory/pull/185
https://github.com/ibexa/order-management/pull/196
https://github.com/ibexa/page-builder/pull/581
ibexa/fieldtype-matrix#100
https://github.com/ibexa/shopping-list/pull/68
https://github.com/ibexa/site-context/pull/149
ibexa/user#138
https://github.com/ibexa/discounts-codes/pull/54
https://github.com/ibexa/activity-log/pull/175
https://github.com/ibexa/collaboration/pull/119

Not listed, deliberately:

• [TMP] IBX-12043: Upgraded to Doctrine DBAL 4 test-core#35 — development tooling, prepare_project_edition.sh adds every listed package to the project's require, and this one does not belong in an installed project.
• https://github.com/ibexa/data-intelligence-layer/pull/45 — no edition declares it, so forcing it in would exercise a setup nobody ships. Covered by its own pipeline.
• https://github.com/ibexa/oauth2-server/pull/51 — not required by any edition installed here, so it cannot be exercised by this run. Its own pipeline fails on a pre-existing minimum-stability problem unrelated to DBAL.

Each package is flagged shouldBeAddedAsVCS so Composer resolves the branch straight from GitHub. Satis lagged behind a force push more than once while this work was in progress, and a regression run silently resolving a stale commit is worse than a slower one.

What this build is expected to catch that the package pipelines cannot:

• The per-package Browser tests jobs fail across the set with Your requirements could not be resolved, because an inline alias only applies in the root package and ibexa/oss pulls these transitively at ~6.0.x-dev. Here the branches are added to the project's own require, so the alias holds and the stack resolves as one.
• MariaDB. No package pipeline in the organisation runs it, and DBAL 4 reparents MariaDBPlatform off MySQLPlatform, which already broke the Random sort clause once during this work. Setup 2 covers MariaDB 11.4.
• A real installation rather than a test fixture: schema generated by the installer, on MySQL 8.4, MariaDB 11.4 and PostgreSQL 18, with Solr and Elasticsearch.

Databases are created fresh here, so this exercises the new utf8mb4_unicode_520_ci DDL rather than the upgrade path. Converting an existing database is a separate procedure — two console commands plus the SQL script in ibexa/installer — and is not covered by this run.

[bnowak]

in browser tests (beside of timeouts) I can spot some failing tests, like:

Then I should see a Base price with "12.00 $" value         # Ibexa\ProductCatalog\Behat\Context\ProductsContext::iVerifyBasePrice()
      Webmozart\Assert\InvalidArgumentException: Failed asserting that expected string '12.00 $' is equal to actual '12.00 US$' for css locator 'basePricePreview': '.ibexa-pc-product-prices__main-price .ibexa-pc-product-item-preview__value' in vendor/webmozart/assert/src/Assert.php:2604

And I set product attributes                                                          # Ibexa\ProductCatalog\Behat\Context\ProductsContext::iSetAttributes()
      | label | value | attributeType |
      | Year  | 2026  | datetime      |
      RuntimeException: Field Year not found. Found:  in vendor/ibexa/admin-ui/src/lib/Behat/Page/ContentUpdateItemPage.php:201

Are they dbal update-related? or some independent product-catalog topics?

anyway, if would be nice to have full regression set pass in the end before merging 😉

[ViniTou]

@bnowak
those are either failing on main branch as well, or single PR needs rebased after the fix. I'm waiting for 6.0 to lit green before restart as it dosent make sense otherwise.

(those were failing at 6.0 at the time creating this as well, for now, only the selector one is failing afaik).