Review: preserve CUE track metadata at launch

[Alexbeav]

Selecting a same-stem CUE currently replaces it with the BIN before mounting, so the runtime loses the CUE track layout. Use the existing resolve_disc_path(path).mount result at the launch adapter. Raw-only images, CHD and the resolver's unusable-CUE fallback retain their existing behavior.

The source-owned regression extracts the actual adapter and compiles it with the production resolver and parser. The unchanged adapter loses track metadata in four of eight synthetic cases; the corrected adapter passes all eight, including 12-track mixed media and eight-track audio-only media through both CUE and BIN entry points. It checks track numbers, types and starts.

Run python runtime/tests/test_launch_disc_path.py --help for the configured-compiler entry point. CTest registers the regression for GNU/Clang C++ builds.

Validation and limits

• Exact base: 8d56cf659fd84367c3f778e17851674ac7896371; review head: 7c11d57ea12812aa01587dcb5c313093ed1123d5.
• Native Windows GCC 16 runtime builds and passes 69/69 enabled runtime tests; two additional tests are disabled in the source configuration. The build has UI, setup wizard, debug tools, netplay, rewind and Vulkan disabled.
• Owned Wipeout XL (USA), canonical 12-track CUE with owned SCPH1001: verification, current-source generation, native build and an isolated 50-second startup pass; normal SDL window close at frame 3092, process exit 0. Executable SHA-256 49f53d0ef9bbf63a3a03f064534901434bf93bf601d8b4de811768ea84eac630. This is bounded startup and normal shutdown, not gameplay completion or listening acceptance. Audio used SDL dummy output. No native Linux/macOS game route is claimed.
• The unchanged current recompiler builds and has 58/61 enabled tests passing, with three pre-existing baseline failures (dirty_text_continuation_guards, aot_overlay_discovery, release_zip) and three disabled tests. This branch does not change that recompiler source. Those baseline failures are not reported as passes.
• Diff audit: general source repair, synthetic tests and documentation only; no retail disc/BIOS, generated game code, player data or private tooling.

Scope and fork review

This review targets immutable review-base/mstan-8d56cf659fd8. It contains 1 commit and 4 changed files. Do not merge it into the fork default branch; it is the review record before upstream submission.

• docs/LAUNCH_DISC_PATH.md: behavior and verification documentation.
• runtime/CMakeLists.txt: source-owned regression and test registration.
• runtime/src/main.cpp: bounded correction.
• runtime/tests/test_launch_disc_path.py: source-owned regression and test registration.

Developed with AI assistance; validated as described (test evidence in PR body). AI writes the code and the PR, but I always test before I send something up. Happy to iterate on this process with your feedback.

---

Summary by cubic

Fixes the launch adapter so selecting a same-stem CUE keeps its track layout instead of swapping it for the BIN before mounting. The adapter now returns resolve_disc_path(path).mount; raw-only images, CHD, and the unusable-CUE fallback behave as before.

Bug Fixes

• Adds a source-owned regression that compiles the actual adapter with the production resolver and parser and checks track numbers, types, and starts across eight synthetic cases (the old adapter lost track metadata in four).
• Registers the regression for GNU/Clang builds in CTest and documents the adapter behavior and limits in docs/LAUNCH_DISC_PATH.md.

^{Written for commit 7c11d57. Summary will update on new commits.}

[Alexbeav]

@cubic-dev-ai review this PR

[cubic-dev-ai]

@cubic-dev-ai review this PR

@Alexbeav I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

No issues found across 4 files

_{Re-trigger cubic}

[cubic-dev-ai]

No issues found across 4 files

_{Re-trigger cubic}