0.6.0 - 2026-08-14

Changed

• ⚠️ Requires phoenix_kit ~> 2.4. PublishingGroup.changeset/2 calls
  PhoenixKit.Utils.Slug.put_slug/3, which core added in 2.4.0. The previous
  ~> 2.0 let a host resolve a core without that function: the build succeeded
  and then raised UndefinedFunctionError on every group create and update, in
  the host's application. Nothing else about the requirement changed — it stays
  two-segment, so every later 2.x core still resolves.

Fixed

• Creating a group whose name matched a TRASHED group failed with
  "already exists" (PR #41 follow-up). idx_publishing_groups_slug is a plain
  UNIQUE index with no status predicate, so a trashed group still owns its slug,
  but add_group/2 built its uniqueness probe from the ACTIVE groups only. It
  reported the slug free, the insert died on the constraint, and the admin saw
  "already exists" for a group no longer visible anywhere. The probe now covers
  every group regardless of status: the name yields news-2, and an explicitly
  typed slug that a trashed group holds is refused at the context layer instead
  of crashing into the index.

• The group slug suffix nested instead of counting. A third group named alike
  got same-name-2-3 (and a fourth same-name-2-3-4) because the loop recursed
  on the already-suffixed slug. It also appended the suffix without regard for
  the length cap validate_length(:slug, max: ...) enforces. Now uses core's
  Slug.ensure_unique/3, which counts from the base and trims it to make room —
  the third is same-name-3.

• A slug DERIVED from the group name skipped valid_slug?/1. Only the
  explicitly-typed branch ran the check, so nothing validated what the generator
  itself produced. A disallowed derived slug is now suffixed past rather than
  rejecting an otherwise legitimate group name.

Changed (internal)

• The group-create uniqueness probe is a single select slug query
  (DBStorage.all_group_slugs/0) instead of list_groups/0, which ran
  StaleFixer.fix_stale_group/1 — a potential write — over every group.

• Retired the :needs_unreleased_core test exclusion. The two tests behind it
  (sibling-dialect locale acceptance, media-selector lock_file_type) pass
  against the released core, so the tag was silently skipping working coverage.
  1570 tests → 1575, none excluded.