v0.3.0

Built against fighters-codex v0.8.7 (extern/fx_lib).

Added

• build link fx_lib unconditionally and retire FA_WITH_FX_LIB — the codex root
  embeds as a subproject (fx::lib + fx::render) and this repo's 3-OS CI is the
  proving ground for the remaining upstream macOS gap (codex#155) (#32)
• bridge FA install discovery, persistence, and configure() first run — env var →
  persisted config → Windows registry/drive probes, then a native folder-picker flow on
  IWindow::showFolderDialog (engine#665) with per-OS config under the platform config
  dir (#11)
• bridge mount LIB archives and answer hasAsset and listAssets — memory-mapped,
  case-insensitive flat name index in the shape of FA's own LibStartUp hint index
  (last registration wins, deterministic mount order); load*() stays Phase 3 (#32)
• bridge on-disk translation cache keyed by source fingerprint — self-invalidating
  per-user store under the platform cache dir; readWithCache() is the shared read
  path for the Phase 3 transcoders (#13)

Changed

• test Catch2 v3.7.1 adopted as the unit-test framework (FetchContent, the
  fighters-codex precedent); test_plugin_load deliberately stays a plain
  ModLoader-rehearsal executable (#38)
• build fl-headers re-vendored at engine v0.3.6 — implements the grown
  IContentPack contract (namespaceId() = "fa", sensor/weapon/manual loaders,
  resolveTilePath) and vendors the IWindow platform header set for configure()
• readme refresh fighters-codex relationship (fx_lib on Linux, fx_render, fxc) (#61)
• docs asset support matrix refreshed against fx_lib v0.8.x (pin now v0.8.7) (music now in-lib, T2
  full band read/write, SH state/articulation, campaign/UI/video parsers) and the
  engine's v0.3.x asset-type split; README fxc misnomer corrected to fxe

v0.2.1

Built against fighters-codex v0.3.0 (extern/fx_lib).

Fixed

• The release workflow published releases with empty notes and a failing
  provenance step: #26's action-gh-release v3 bump had rewired the notes body
  to a nonexistent git-cliff step and added a broken attestation step. Notes
  wiring restored; attestation removed until artifact packaging lands (#51).
  v0.2.0's published notes were repaired by hand (#50)

v0.2.0

Built against fighters-codex v0.3.0 (extern/fx_lib).

Added

• Plugin skeleton (roadmap Phase 1, #31): bridge/ builds the plugin shared
  library (libfa-bridge.so / fa-bridge.dll / libfa-bridge.dylib) with a
  stub fa::FaContentPack implementing the engine's IContentPack —
  priority 100 (overrides fl-base-pack), init() keyed on FA_INSTALL_DIR,
  every load*() empty until the transcode pipelines land — and exports
  fighters_legacy_create_pack() for the engine's ModLoader (#10)
• extern/fl-headers/ — the engine's three interface headers vendored verbatim
  at engine v0.2.6, with PIN.md recording tag/SHA, per-file sha256 hashes, and
  the update procedure
• packaging/manifest.toml.in + build-time staging: every build produces a
  drop-in mod directory at build/<preset>/stage/mods/fa-bridge/
• Assert-based smoke tests: test_fa_content_pack (stub contract, all init()
  states) and test_plugin_load (loads the built plugin exactly as the
  engine's ModLoader does — dlopen/LoadLibrary, factory symbol, manifest
  consistency)
• FA_BUILD_TESTS CMake option (default ON); MSVC test presets; ctest step
  restored in CI with noTestsAction: error; sanitizer and coverage workflows
  restored to run on every push and pull request
• .github/workflows/add-to-project.yml — auto-adds newly opened (or reopened)
  issues and pull requests to the org-wide
  Fighters Ecosystem project board
  via actions/add-to-project; uses pull_request_target so fork PRs can read
  the ADD_TO_PROJECT_PAT org secret without checking out untrusted code

Fixed

• asan and coverage presets now pass their instrumentation flags to shared
  library links (CMAKE_SHARED_LINKER_FLAGS) — required for a SHARED plugin
• debug-msvc / release-msvc test presets exist, so the
  ctest --preset debug-msvc instruction in docs/development.md works

v0.1.0

Built against fighters-codex v0.3.0 (extern/fx_lib).

Added

• docs/roadmap.md — charter, the 1.0 definition, seven constraint-gated phases
  (P0 Program Reset → P6 Full Coverage & 1.0) with external gates wired to live
  tracker issues in fighters-codex and fighters-legacy (#44)
• docs/asset-support-matrix.md — the living per-format matrix of which FA content
  the bridge can deliver today: aircraft flight models, entity definitions, ~95% of
  3D models, all 2D art, and sound effects are ready-now; missions partial; terrain,
  music, video, and campaign blocked upstream (#28)
• extern/fx_lib git submodule, registered and pinned to fighters-codex v0.3.0,
  with the FA_WITH_FX_LIB CMake option (default OFF until the codex cross-platform
  port lands) and dependabot watching the pin (#27)
• Portable stdlib-only Python release tooling: scripts/release.py (verifies the
  fx_lib pin is a codex release tag before tagging) and scripts/draft-changelog.py,
  replacing git-cliff (#29)
• epic label and epic issue form; milestones Phase 0–6 with epics #30–#36 (#44)

Fixed

• Windows CI had been failing on every PR: the MSVC presets pinned the
  Visual Studio 17 2022 generator, which windows-latest no longer ships (#29)
• The DCO check rejected GitHub-generated merge commits, which can never carry
  Signed-off-by (#43)
• The FA-asset .gitignore blocklist ignored non-FA extensions (*.FLT, *.SHP,
  *.VXA) while missing the real FA set; now cross-checked against the codex format
  docs, both case variants (#27)
• Repository links pointed at the pre-migration jomkz org (#29, #44)

Changed

• Repository renamed fa-content → fa-bridge — it provides no content; it
  bridges the end-user's own FA installation into locally cached engine assets (#45)
• Documentation truth pass: everything previously presented as existing is now
  explicitly planned-state with its roadmap phase, and the undefined "Phase 1
  Workstream B" vocabulary is gone. docs/architecture.md now states the real
  engine contract — packs return raw bytes in canonical open formats past
  AssetValidator, so every load*() is a transcode (#44)
• The planned assets/ module is renamed transcode/ across scopes, labels, and
  issue forms — before the directory exists (#44)
• CI is truthful: the vacuous ctest step is gone and the sanitizer/coverage
  workflows are manual-only until the Phase 1 skeleton restores them with real
  tests; the release workflow extracts notes from CHANGELOG.md (#29)