Review main Merged

96798[TASK] Pin Extbase workspace reads and writes in the conformance suite

Alexander Schnitzler CR +2 V +2 Normal +209 −2 updated 10/18/26, 14:32

Commit message

Nothing pinned what a query returns for a translated or versioned
record inside a workspace, or which row an update actually writes to
when the object was loaded in one. Bug fixes for the LIMIT-after-
overlay guess and for count() in workspaces would otherwise change
observable results without a test noticing.

This commit extends BehaviourConformanceTest with the workspace group:
a workspace query overlays live rows with their versions, drops hidden
or delete-placeholder versions and moved records whose move pointer is
not itself selected, and lists a record new only in the workspace under
its own uid; the same, with enable fields ignored, returns hidden
versions and hidden live rows while still dropping placeholders and
moved records; a single moved record resolves through its move pointer
when fetched directly, which is explained inline rather than with a new
test, since the existing single-fetch case already exercises the same
code path; relations under a workspace resolve through their own
versions, including MM rows of a version replacing those of the live
record. Count in a workspace is marked for its fix, LIMIT in a
workspace is pinned as an approximation of LIMIT after the overlay, and
a live case pins the full limit after language filtering. Updating an
object loaded in a workspace is pinned as a known hole with no prior
test: it writes the live row, not the version, because Extbase never
writes to versions.

This does not change any class under Classes/; it only adds tests.

This is step 2 of 4 of "Pin Extbase persistence behaviour in an
outcome-based conformance suite", which gives the refactorings and bug
fixes that follow an oracle for what Extbase does today in terms of
records and rows, not SQL strings.

Review

  • Code review CR +2 Done
    • Benni Mack +2
  • Verified V +2 Done
    • Benni Mack +2
    • Oliver Klee +1
    • core-ci +1 CI

Issues 3

Files 1

Review log 911

  1. Alexander Schnitzler uploaded patch set 1 10/17/26, 16:44

  2. core-ci voted on patch set 1 V +1 10/17/26, 16:55 Core CI is happy: https://git.typo3.org/typo3/CI/cms/-/pipelines/111479

  3. Alexander Schnitzler commented on patch set 1 10/17/26, 18:10

    This change is ready for review.
  4. Oliver Klee voted on patch set 1 V +1 10/17/26, 19:15

  5. Alexander Schnitzler wrote 10/17/26, 19:47

    Topic set to extbase
  6. Alexander Schnitzler uploaded patch set 2 V +1 10/18/26, 13:07

  7. Alexander Schnitzler commented on patch set 2 10/18/26, 13:10

    Patch set 2 only touches LIMIT; the workspace cases are unchanged. - The LIMIT @todo in `workspaceCountAndLimit()` now states the goal: LIMIT applies after the overlay. Fetching twice the limit in a workspace (#105159, change 86370) is an approximation of that goal and can still return fewer rows than the limit. This follows Christian's analysis in #85284: TypoScript `CONTENT` applies LIMIT in SQL before the overlay, which is a known limitation and not a yardstick for Extbase. - New test `limitWithRespectSysLanguageIsAppliedAfterLanguageFiltering()`: with `respectSysLanguage`, the SQL already drops default rows that have no visible translation, so Extbase returns the full limit. It is pinned without a @todo, because this is the intended behaviour. - The @todo for `count()` in a workspace stays as it was.
  8. core-ci voted on patch set 2 V +1 10/18/26, 13:14 Core CI is happy: https://git.typo3.org/typo3/CI/cms/-/pipelines/111519

  9. Oliver Klee voted on patch set 2 V +1 10/18/26, 13:15

  10. Benni Mack voted on patch set 2 CR +2 V +2 10/18/26, 14:32

  11. Benni Mack merged patch set 3 10/18/26, 14:32