Skip to content

verify-media-dlc (plan7): scripted checks for H.264, DLC and the flaky guest Break

All runs: Mac, headless, scripted (no one played). Runtime built from this branch (base 272aeda) in $TMPDIR; ~/Desktop/NappleLATEST untouched. Game data: main checkout Napple/work/games (LAS 01006BB00C6F0000, SMO 0100000000010000, built 2026-10-22). Default save root unless noted.

1. SMO / H.264 (VideoToolbox)

  • SMO never creates a movie::Decoder in any scripted path I tried (log has zero movie: lines): idle 150 s on the title menu; Action Guide (LS_DOWN x3, A); New Game (A every 4 s for 110 s, fresh save root with system/ copied in) which reaches the in-engine opening cutscene (screenshots at frames 3300 and 4500), still no movie. So no scripted path in SMO reaches a movie in the first ~2 minutes; SMO's early flow looks to use in-engine cutscenes only. SMO modules do link the movie library (subsdk0), so a later scene may.
  • LAS does create decoders at ~0.5 s: type 3 (VP9, texture mode 2, format 3, runs on the host, 1080p, ~0.6-0.9 ms/frame) and a second one, type 5, format 1, buffer mode 1. That type-5 object is the only unknown decoder any title makes in the first seconds. Default run: "left to the library".
  • With NAPPLE_MOVIE_H264_TYPE=5 the VideoToolbox path IS reached on LAS: movie: H.264 decoder 0x... on the host (VideoToolbox), mode 1 (buffer), output format 1 then output format 1 not implemented on the host (NV12 = 4 only) then decoder ... destroyed after 0 frames; the game keeps running (exit 0, VP9 movie unaffected). So the hook, the type override and the VT decoder creation work against a real guest; no H.264 frame was decoded (format 1 is refused, and type 5 may not be AVC at all: it is created next to the VP9 decoder, so it may be an audio or auxiliary decoder). Evidence logs in the shared notes folder (las-h264-0.log, las-h264-5.log).
  • Fix needed (not mine, runtime/movie.c): confirm the real AVC DecoderType value (a title that actually plays an H.264 movie is needed; TOTK/other titles use VP9) and decide what format 1 means.

2. Animal Crossing, updates and DLC

  • Command: napple build <3 nsp> -k prod.keys --title-keys ... --firmware firmware.zip --updates all --dlc all --data-only -o $TMPDIR/vmd/ac (2m15s, 26 GB, 3 build dirs, one per version).
  • There is no DLC in Testing Material/AnimalCrossing+Updates+DLC. The three files are the base (01006F8002326000 v0) and two updates (01006F8002326800 v2097152 = 3.0.1 and v2228224 = 3.0.3). Plan log: 2 update(s), 0 DLC selected / 3 version(s), 0 DLC. The [01006F8002326800] file looks like DLC by its id but is the update title id (base + 0x800). So aoc:u cannot be shown with real content; this needs an AOC .nsp.
  • Boot, 45 s headless, each build dir: none boots.
  • 3.0.3 (ff1d1c05670db602) and 3.0.1 (8f2cb7a9774959c8): exit 70 at 0.43 s, ipc: TIPC request (cmd 0) is not implemented / FAIL: TIPC not implemented, right after boot: loader callback: modules linked, before any aoc/fs call. NAPPLE_IPC_LENIENT=1 does not help (ipc.c fails on type >= 16 unconditionally, ipc.c:314).
  • v0 (7fc1baff976aeca4): SIGSEGV fault=0x20 pc=main+0x19a6250 at startup (exit 139).
  • So no aoc:u / fs cmd 30 lines were produced. The unit-level coverage of those (cleanup-small's fs cmd 30 test) is the only DLC evidence.
  • Fix needed (runtime/ipc.c, not mine): a TIPC (tiny IPC, command type = 16 + id) dispatch path, starting with the sm: commands the SDK sends at init. AC's SDK is a newer libnx/NX SDK that uses TIPC.

3. "guest Break ... sdk+0xd5f6c4, exit 70" flakiness

  • 5 runs x 12 s on each of LAS and SMO with the branch build and with ~/Desktop/NappleLATEST/bin/napple-runtime: 20/20 exit 0, no FAIL/CRASH line (twice: once on an older base, once on this base).
  • Not reproduced, so I cannot call it flaky or build-specific from these data. Observations:
  • The reported failure timing (0.43 s) equals the AC TIPC failure timing, i.e. the early-boot class. The binary identity matters less than the game data: the report may have come from a game built with a different (Studio) build than work/games.
  • An empty --save-root (no system/fonts) crashes both games at ~0.47 s with CRASH signal=11 fault=0x38 pc=main+0x111e000 after pl: no shared fonts .... That is exactly the earlier "unreproducible boot-crash" signature (fault 0x38, main+0x111e000): its cause is a missing <save-root>/system/fonts (fonts-fallback owns the fix in services/pl.c). Worth checking that the exit-70 run also used a save root without fonts.

Files

Napple/tools/flaky_boot.sh (new; N sequential headless boots of one runtime on one data dir, prints exit code and first FAIL/CRASH line; SAVE_ROOT= to override), this note.