v0.0.2

Other

• composite SPU into RGBA overlay (palette + contrast + BT.601)
• add DVD Sub-Picture Unit decoder
• scrub enumerated-denial / decorative-attribution prose (r131 disclaimer-hygiene sweep follow-up)
• add Phase-3c-precursor NavInstruction disassembler
• decode NAV-pack DSI typed sub-sections (SML_PBI / SML_AGLI / SYNCI)
• decode NAV-pack PCI highlight (HLI_GI + SL_COLI + BTN_IT)
• re-export PaletteEntry / NavCommand / PgcCommandTable at crate root
• decode PGC palette colour-LUT + pre/post/cell command table
• Phase 3b: VOB → MKV mux glue + convert_dvd_to_mkv pipeline
• Phase 3a: VOB demuxer — MPEG-PS pack + nav-pack + DVD substream router
• Phase 2: IFO body parser — VMGI/VTSI MAT + TT_SRPT + VTS_PTT_SRPT + VTS_PGCI + VTS_C_ADT

Added

• SPU RGBA compositor — SubPictureUnit::composite(buf, palette)
  turns a parsed sub-picture plus the PGC's 16-entry PaletteEntry
  colour-LUT into a finished SpuBitmap overlay (row-major
  [R, G, B, A] pixels + on-screen rectangle), completing the
  "final framebuffer left to the caller" gap entirely inside the
  crate. Clean-room per docs/container/dvd/application/mpucoder-spu.html
  (SET_COLOR/SET_CONTR semantics, 0x0 transparent … 0xF opaque,
  top/bottom field interleave) + stnsoft-color_pick.html (BT.601
  studio-swing luma scale Y = 16 0 % … Y = 235 100 %). No
  libdvdread / libdvdnav / libdvdcss / FFmpeg / VLC / mpv / xine
  source consulted; no web search.

  • ycbcr_to_rgb — standalone BT.601 studio-swing
    (Y, Cb, Cr) -> (R, G, B) inverse-matrix conversion in fixed
    point (1.164 / 1.596 / 0.391 / 0.813 / 2.018 coefficients scaled
    by 1<<16, round-half-up, clamped to 0..=255).
  • SpuBitmap — { x, y, width, height, rgba } overlay: the
    SET_DAREA rectangle plus the composited pixels, ready to blend
    onto the decoded MPEG-2 frame by the player.
  • The four 2-bit pixel codes are resolved through the unit's own
    SET_COLOR (→ 0..=15 palette index) and SET_CONTR
    (→ 0..=15 alpha, expanded to 8-bit by nibble replication);
    a unit lacking those uses well-defined fallbacks (background
    index, fully-opaque). Returns None when SET_DAREA /
    SET_DSPXA are absent (a malformed unit per the spec).
  • +5 unit tests (BT.601 known-point conversion incl. clamp +
    red/blue dominance, contrast-nibble expansion, full solid-rect
    composite round-trip, missing-SET_DAREA → None).
    126 lib tests (was 121 after the SPU decoder landed).

• spu module — DVD Sub-Picture Unit decoder, the overlay
  graphics stream that carries subtitles, menu button highlights,
  and karaoke captions. Pure-bytes decoder clean-room per
  docs/container/dvd/application/mpucoder-spu.html (sole 160-line
  source; no libdvdread / libdvdnav / libdvdcss / FFmpeg / VLC / mpv
  / xine / HandBrake source consulted; no web search).

  • SpuHeader — the 4-byte SPUH (SPDSZ total size +
    SP_DCSQTA offset to the Sub-Picture Display Control Sequence
    Table).
  • SpuCommand — typed enum for the eight SP_DCSQ command
    codes: ForcedStartDisplay (0x00) / StartDisplay (0x01)
    / StopDisplay (0x02) / SetColor (0x03, four 4-bit
    palette indices) / SetContrast (0x04, four 4-bit alpha
    values) / SetDisplayArea (0x05, four 12-bit coordinates)
    / SetPixelDataAddresses (0x06, top/bottom field offsets) /
    ChangeColorContrast (0x07, raw LN_CTLI / PX_CTLI
    parameter blob preserved for the caller) / EndOfSequence
    (0xFF, the CMD_END terminator).
  • SpDcSq — one display-control sequence: a 4-byte header
    (90 kHz/1024 start_time + next_offset chain pointer) plus
    the decoded command list. Chain-walk validates per-block
    forward progress and rejects loops.
  • SubPictureUnit::parse — top-level entry that walks SPUH
    • every chained DCSQ from the SP_DCSQTA offset until a
      terminal block (whose next_offset points back at itself).
      Convenience accessors pixel_data_offsets() / display_dimensions()
      pull the PXDtf/PXDbf offsets and rectangle width/height out of
      the command stream.
  • decode_rle_field — the 2-bit / four-form PXD run-length
    decoder. Implements the nested-prefix encoding (n n c c /
    0 0 n n n n c c / 0 0 0 0 n n n n n n c c /
    0 0 0 0 0 0 n n n n n n n n c c) and the 16-bit
    "count=0 = until end of line" terminator, with byte alignment
    at every row boundary per mpucoder-spu.html §PXDtf.
  • render_field — flattens the run vector into a row-major
    Vec<u8> of palette indices (0..=3), one byte per pixel,
    ready for blending against the PGC's 16-entry PaletteEntry
    table.
  • spdcsq_stm_to_ms — converts an SP_DCSQ_STM 90 kHz/1024
    delay to integer milliseconds via the inverse of the
    mpucoder-spu.html conversion table.

  Producing a final framebuffer (YCrCb + alpha) is intentionally
  left to the caller — that step needs the PGC PaletteEntry
  table (already exposed by crate::ifo) plus the renderer's
  preferred pixel format, both outside the SPU bitstream itself.

  +13 unit tests (header parse / delay conversion / one-run RLE
  for all four forms / end-of-line marker / EOL row padding /
  full-unit round-trip with six commands / CHG_COLCON raw
  round-trip / DCSQTA-out-of-range rejection / runaway-DCSQ
  rejection / opcode table). 132 tests total (was 119).

• nav module — typed VM instruction decoder (Phase 3c precursor).
  The previous NavCommand surface exposed only an 8-byte raw word
  plus the 3-bit command_type classifier; the new NavCommand::decode() -> NavInstruction returns a typed-enum disassembly tree clean-room
  per docs/container/dvd/application/mpucoder-vmi.html +
  mpucoder-vmi-sum.html + mpucoder-vmi-jmp.html +
  mpucoder-sprm.html (no libdvdread / libdvdnav / libdvdcss /
  FFmpeg / VLC / mpv / xine / HandBrake source consulted; no web
  search). No execution — an interpreter that owns
  SPRMs / GPRMs / PC / RSM stack is the bulk of Phase 3c proper;
  decoding the stream is the prerequisite step shared by a future
  executor, an analyser, and a disc debugger.

  • Register — 8-bit operand classifier: Gprm(0..=15) /
    Sprm(0..=23) / Invalid(raw) per the asterisk note on the VMI
    spec page (only 0x00..=0x0F and 0x80..=0x97 are valid).
  • SetOp + CmpOp — the SET (12 named codes: mov,
    swp, add, sub, mul, div, mod, rnd, and, or,
    xor) and CMP (7 named codes: BC, EQ, NE, GE, GT,
    LE, LT) sub-op tables from the same page.
  • LinkSubset — the 13-entry inner table for the Type-1 0x20 0x01 Link command: LinkTopCell / LinkNextCell /
    LinkPrevCell / LinkTopPG / LinkNextPG / LinkPrevPG /
    LinkTopPGC / LinkNextPGC / LinkPrevPGC / LinkGoupPGC /
    LinkTailPGC / Rsm + Nop, with the spec's invalid bag
    (0x04, 0x08, 0x0E, 0x0F, 0x11..0x1F) preserved via
    Invalid(raw).
  • JumpSSTarget + CallSSTarget — the four-way
    destination selector (FirstPlay / VmgmMenu { menu } /
    VtsmMenu { vts, ttn, menu } / VmgmPgcn { pgcn }) from the
    JumpSS / CallSS rows in mpucoder-vmi.html. CallSSTarget
    additionally carries the rsm_cell resume-cell byte shared by
    all four CallSS variants.
  • NavInstruction — top-level decode enum. Variants for the
    well-defined opcodes: Nop, Goto { line }, Break,
    SetTmpPml { level, line }, LinkSub { subset, hl_bn },
    LinkPgcn { pgcn }, LinkPttn { pttn, hl_bn },
    LinkPgn { pgn, hl_bn }, LinkCn { cn, hl_bn }, Exit,
    JumpTT { ttn }, JumpVtsTt { ttn },
    JumpVtsPtt { ttn, pttn }, JumpSs(JumpSSTarget),
    CallSs(CallSSTarget), SetStn (with af/sf/nf flag
    bits and per-channel register-or-immediate source), SetNvtmr,
    SetGprmMd (with counter mode bit), SetAmxMd, SetHlBtnn,
    Set { op, dst, src }. Compound Type 4..6 forms surface their
    classifier SetOp + CmpOp sub-ops via SetCLnk (Type 4),
    CSetCLnk (Type 5), CmpSetLnk (Type 6); the per-operand
    sub-decode is deferred to the executor. Type 7 returns
    Unknown (the VMI page documents the family has never been
    observed in real-world streams); structurally-impossible
    encodings return Invalid.
  • 42 new unit tests covering the Register GPRM / SPRM /
    invalid-hole classifier, the full SetOp / CmpOp /
    LinkSubset named-code tables, the spec's named-but-invalid
    sub-codes (SetSystem sub=5, Set sub=0/C/F, Type 0 cmd
    nibble 4..F, Link cmd nibble 2), the round-trip from a
    NavCommand::default() (all zero) decoding to Nop, and one
    decoded form per NavInstruction variant including the
    JumpSS four-way target selector and the CallSS rsm_cell field.

• NAV-pack DSI typed sub-section decode (Data Search Information):
  the DsiPacket decoder previously surfaced only the DSI_GI preamble
  and a flat 43-entry VOBU_SRI array; it now returns a typed
  DsiPacket { general_info, sml_pbi, sml_agli, vobu_sri, synci }
  with every spec-listed field exposed by name, clean-room per
  mpucoder-dsi_pkt.html (no libdvdread / libdvdnav / FFmpeg / VLC /
  mpv / xine source consulted).

  • DsiGi — DSI_GI general information (packet 0x00..0x20):
    nv_pck_scr, nv_pck_lbn, vobu_ea, the 1st/2nd/3rd reference-
    frame end-address triplet, the (vobu_vob_idn, vobu_c_idn)
    identifier pair, and the BCD c_eltm cell-elapsed-time + frame-
    rate bits field. Convenience getters
    (DsiPacket::nv_pck_scr() etc.) mirror the pre-refactor flat-field
    accessors so the bump stays source-compatible for call-sites that
    only read DSI_GI.
  • SmlPbi + SmlAudioGap — SML_PBI seamless-playback info
    (packet 0x20..0xB4, 148 bytes): the 16-bit ilvu flag word with
    preu() / is_ilvu() / unit_start() / unit_end() bit
    decoders, the (ilvu_ea, nxt_ilvu_sa, nxt_ilvu_sz) interleaved-
    block jump pointers, the VOB-span video PTM pair, and the 8 ×
    16-byte per-audio-stream gap table (stp_ptm1, stp_ptm2,
    gap_len1, gap_len2 per stream).
  • SmlAgli + SmlAngleCell — SML_AGLI seamless-angle info
    (packet 0xB4..0xEA, 54 bytes): 9 angle cells, each 6 bytes wide
    (dsta: u32 with bit-31 direction flag + sentinel values for
    "absent" and "no more video"; sz: u16 ILVU size in sectors).
  • VobuSri — VOBU search-information table (packet 0xEA..0x192,
    168 bytes = 42 × 4): sri_nvwv (next-VOBU-with-video), 19 forward
    scaled-distance entries, sri_nv + sri_pv brackets, 19 backward
    entries, sri_pvwv (previous-VOBU-with-video). The bit-31
    VALID_BIT, bit-30 INTERMEDIATE_BIT, and 30-bit OFFSET_MASK
    constants make sentinel handling explicit. (Previous flat-array
    decode over-read by 4 bytes into SYNCI; the typed layout fixes
    that.)
  • Synci — SYNCI A/V-sync pointer table (packet 0x192..0x222,
    144 bytes): a_synca: [u16; 8] audio + sp_synca: [u32; 32]
    subpicture per-stream first-packet offsets. AUDIO_DIRECTION_BIT
    (bit 15) and SP_DIRECTION_BIT (bit 31) constants surface the
    spec-defined direction flag.
  • 9 new unit tests (dsi_section_offsets_match_spec,
    dsi_parses_general_info_block,
    dsi_parses_sml_pbi_block_and_ilvu_flags,
    dsi_pbi_ilvu_flag_decoders_isolate_bits,
    dsi_parses_sml_agli_block,
    dsi_parses_vobu_sri_block_and_brackets,
    dsi_parses_synci_block, dsi_rejects_short_buffer,
    dsi_nav_pack_round_trip_through_full_sector) and a new
    build_dsi_body helper that emits a fully-populated 546-byte DSI
    body so every per-section offset is pinned exactly.

Changed

• Breaking — DsiPacket's public field layout. The previous flat
  { nv_pck_scr, nv_pck_lbn, vobu_ea, vobu_1stref_ea, vobu_2ndref_ea, vobu_3rdref_ea, vobu_vob_idn, vobu_c_idn, c_eltm, vobu_sri: Box<[u32; 43]> } shape was replaced by the typed sub-section struct described
  above. Source-compatible getters (nv_pck_scr() etc.) are provided
  for the DSI_GI fields; the vobu_sri field is now a VobuSri struct
  rather than a flat boxed array. Pre-0.0.2 release — no published
  consumer to break.

• NAV-pack PCI highlight information (menu buttons): the
  PciPacket decoder previously read only hli_ss; it now
  materialises the full HLI_GI / SL_COLI / BTN_IT sub-structure when
  a VOBU declares buttons, clean-room per mpucoder-pci_pkt.html (no
  libdvdread / libdvdnav / libdvdcss / FFmpeg / VLC / mpv / xine
  source consulted).

  • HighlightInfo + PciPacket::highlight: Option<HighlightInfo>
    — the HLI_GI general-information block (hli_s_ptm, hli_e_ptm,
    btn_sl_e_ptm, raw btn_md grouping word, btn_sn, btn_ns,
    nsl_btn_ns, fosl_btnn, foac_btnn). None when the VOBU
    declares no buttons (btn_ns == 0) — the common case, not an
    error.
  • SlColi + SlColiCell — the three SL_COLI_1..3
    selection/action colour-and-contrast schemes. Each 8-byte scheme
    is decoded into selection + action arrays of four
    { color, contrast } cells, indexed by emphasis code
    (0 = background, 1 = pattern, 2 = emphasis1, 3 =
    emphasis2). color is a 4-bit PGC colour-LUT index; contrast
    is the 4-bit blend weight a subpicture/menu renderer applies.
  • ButtonInfo — one 18-byte BTN_IT entry: btn_coln
    colour-scheme selector, the 10-bit X/Y rectangular region
    (start_x/end_x/start_y/end_y), the auto-action flag, the
    four up/down/left/right D-pad adjacency selectors, and the
    raw 8-byte VM command (executing it is Phase 3c VM work per
    mpucoder-vmi.html). The button table holds exactly btn_ns
    entries; an over-long count (> 36) or a body too short to carry
    the declared table raises Error::InvalidUdf.
  • 4 new unit tests (pci_without_buttons_yields_no_highlight,
    pci_decodes_single_button_highlight,
    pci_rejects_overlong_btn_ns,
    pci_rejects_truncated_button_table); a new add_one_button_hli
    test helper injects a known single-button HLI block into the
    synthetic nav sector so every decoded field is asserted exactly.
  • Note: PciPacket and NavPack no longer derive Copy
    (HighlightInfo owns a Vec<ButtonInfo>); they remain Clone.

• PGC palette + command-table parse (richer Phase 2 IFO body): the
  Pgc materialiser now decodes the two PGC-header tables it
  previously skipped, both clean-room per mpucoder-pgc.html (no
  libdvdread / libdvdnav / libdvdcss / FFmpeg / VLC / mpv / xine
  source consulted).

  • PaletteEntry + Pgc::palette: [PaletteEntry; 16] — the
    subpicture/highlight colour-LUT at PGC offset 0x00A4, sixteen
    (0, Y, Cr, Cb) cells (leading reserved byte dropped) surfaced
    as { y, cr, cb }. This is the table an SPU display-control
    sequence indexes into via its 4-bit colour codes
    (mpucoder-spu.html), so a subtitle/menu renderer needs it to
    resolve a pixel to an actual YCrCb value.
  • NavCommand + PgcCommandTable + Pgc::commands: Option<PgcCommandTable> — the command table at
    offset_commands (previously only the offset was read). The
    8-byte header (pre/post/cell counts + end_address) is decoded
    and each list is carved into fixed 8-byte NavCommand words.
    The pre + post + cell <= 128 spec invariant is enforced;
    truncated lists and over-long counts raise Error::InvalidUdf.
    Executing the words is deferred to the Phase 3c VM
    (mpucoder-vmi.html); at the container layer we expose the raw
    words plus a NavCommand::command_type() convenience (top three
    bits of byte 0 — the VMI command-group selector) so a downstream
    interpreter has what it needs without a full opcode model here.
  • 6 new unit tests (palette_entry_skips_reserved_byte,
    command_table_carves_three_lists,
    command_table_rejects_overlong_count,
    command_table_rejects_truncated_list,
    pgc_without_command_table_yields_none, plus extended
    palette/command assertions in
    pgci_parses_one_pgc_with_three_cells). The
    build_pgc_with_cells test helper now emits a real palette +
    4-word command table so the existing round-trip exercises the
    new fields.

• Phase 3b (VOB → MKV mux): clean-room glue between the Phase 3a VOB
  demuxer and oxideav-mkv's MkvMuxer::{add_chapter,write_packet, write_trailer}. Gated behind a default-off mkv-output cargo
  feature so the dvd crate stays useful for chapter-introspection
  consumers and so default-feature CI doesn't have to pull the
  (still-unreleased at time of writing) MKV chapter API in. No
  external library source (libdvdread, libdvdnav, libdvdcss, FFmpeg,
  VLC, mpv, xine, HandBrake) was consulted.

  • pgc_time_to_ns(PgcTime) -> u64 — RFC 9559 §5.1.7 needs
    ChapterTimeStart / ChapterTimeEnd in nanoseconds; DVD's BCD
    hh:mm:ss:ff field uses 30 fps (NTSC) or 25 fps (PAL) per
    mpucoder-pgc.html. Conversion is exact rational math so
    0:0:1.15 @ 30 fps becomes the spec-exact 1_500_000_000 ns
    (truncating with the obvious 1e9 / 30 constant would have
    rounded to 1_499_999_995 ns — see the regression test).
  • mkv_writer::write_title_to_mkv(disc, title_idx, image_path, out_path) — two-pass DVD → MKV converter. Pass 1 probes the
    title's cells to enumerate the (video, AC-3 × N, DTS × N, LPCM
    × N, subpicture × N) stream set so MKV's mandatory upfront
    Tracks element can be sized correctly. Pass 2 re-walks the
    cells and forwards each PES packet to MkvMuxer::write_packet
    with PTS preserved verbatim in the PES's 90 kHz time base; the
    muxer rescales to its internal 1 ms TimecodeScale. Chapter
    atoms are queued via MkvMuxer::add_chapter before
    write_header, one per DvdChapter from the PGC's PTT list,
    titled "Chapter N".
  • pipeline::convert_dvd_to_mkv(source, title_idx, out_path)
    — high-level front door accepting either a dvd://... URI or a
    bare filesystem path. Auto-detect (dvd://) is rejected as a
    Phase-2 followup, matching the existing source-driver semantics.
  • pipeline::list_titles(source) — convenience wrapper around
    DvdDisc::enumerate_titles for CLI front-ends that want to
    surface the title list before letting the user pick one.
  • Sector walker — walk_cell_sectors re-uses the constants
    from vob::{SC_*, looks_like_nav_pack, PackHeader, PesPacket}
    so the pack-header + system-header + nav-pack + padding
    transitions are decoded once across the round. Nav-packs are
    consumed (validated via NavPack::parse) but not surfaced
    further; subpicture / DTS / LPCM PES payloads' first byte (the
    substream ID) is stripped before the body lands in MKV.

• Phase 3b tests (10 cases, all gated behind --features mkv-output): PgcTime → ns for NTSC 30 fps / PAL 25 fps / illegal-
  frame-rate / hour-boundary; stream classification (codec id +
  media type) for video / AC-3 / DTS / LPCM / subpicture; sort
  determinism for the stream set; plus three pipeline::resolve_*
  tests covering URI parsing (dvd:///abs, bare path, dvd://
  auto-detect rejection).

• Phase 3a (VOB demuxer): clean-room MPEG-PS pack + nav-pack walker

  • DVD-substream router per mpucoder-{packhdr,pes-hdr,mpeghdrs,
    pci_pkt,dsi_pkt,dvdmpeg}.html + stnsoft-{vobov,sys_hdr}.html. No
    libdvdread, libdvdnav, FFmpeg, VLC, mpv, or xine source consulted.
  • PackHeader — 14-byte MPEG-2 Program Stream pack header
    decoder: 00 00 01 BA sync + 33-bit SCR base + 9-bit SCR_ext +
    22-bit program_mux_rate + 3-bit pack_stuffing_length. All
    five marker bits are validated; bad sync / missing marker /
    mux_rate == 0 raise Error::InvalidUdf.
  • PesPacket<'a> — zero-copy PES decoder for the DVD subset:
    0xBA pack, 0xBB system header, 0xBD private_stream_1,
    0xBE padding, 0xBF private_stream_2 (no extension), and
    0xC0..=0xDF / 0xE0..=0xEF (MPEG-2 extension with 5-byte PTS
    or 10-byte PTS+DTS). PTS_DTS_flags == 01 is rejected per spec.
  • DvdSubstream — typed substream classifier for the first
    payload byte of a 0xBD packet: Subpicture(0x20..=0x3F),
    Ac3(0x80..=0x87), Dts(0x88..=0x8F), Lpcm(0xA0..=0xA7)
    with track() accessor normalising to 0..=7 (audio) / 0..=31
    (subpicture).
  • PciPacket — Presentation Control Information decoder for
    the DVD-Video NAV-pack's PCI half: nv_pck_lbn, vobu_cat,
    vobu_uop_ctl, vobu_s_ptm, vobu_e_ptm, vobu_se_e_ptm,
    c_eltm, hli_ss.
  • DsiPacket — Data Search Information decoder for the
    DSI half: nv_pck_scr, nv_pck_lbn, vobu_ea,
    vobu_{1,2,3}stref_ea, vobu_vob_idn, vobu_c_idn, c_eltm,
    and the 43-entry vobu_sri search-pointer table (0xEA..0x196)
    used for chapter-accurate forward/backward seek.
  • NavPack — 2048-byte sector-level decoder that validates
    pack header + system header + 0xBF/0x00 PCI prefix + 0xBF/0x01
    DSI prefix and surfaces (pci, dsi). A cheap looks_like_nav_pack
    probe skips the full parse on demux routing.
  • VobDemuxer — stateful walker that consumes 2048-byte
    sectors and routes packets into per-stream buffers. Nav-packs
    are consumed and stashed in VobStreams::nav_packs; video PES
    payloads append to video; private_stream_1 payloads are
    classified and routed to AC-3 / DTS / LPCM / subpicture
    BTreeMap<u8, Vec<u8>> per track. The first substream-ID byte
    is stripped before append so consumers see clean substream
    bytes. 0xC0..=0xC7 (MPEG audio) is pooled into ac3 so
    callers can probe codec from the first frame.
  • demux_vobs(&mut reader, &disc, ts, &cells) -> VobStreams
    • demux_vobs_path(...) convenience wrapper: resolves
      (VobId, CellId) pairs through VtsCAdt::lookup, translates
      title-relative sector positions to absolute LBA via
      VTS_xx_1.VOB's base LBA, then runs each cell's range through
      VobDemuxer.

• Phase 3a tests (all synthetic, 12 cases): pack-header roundtrip

  • bad-sync + zero-mux-rate rejection, PES with/without PTS +
    bad-start-code rejection, DVD substream classification across all
    four substream families, NavPack parse + corrupt-DSI rejection,
    and an end-to-end synthetic VOBU (nav sector + video PES sector +
    AC-3 PES sector) showing the demuxer routes payloads correctly
    while preserving nav-pack metadata.

• Phase 2 (IFO body parsing): clean-room IFO structural decoder per
  mpucoder + stnsoft DVD-Video reference pages (no libdvdread /
  libdvdnav / libdvdcss / FFmpeg / VLC / mpv / xine source consulted).

  • VMGI_MAT — full VIDEO_TS.IFO Video Manager Information
    Management Table parse: last-sector + IFO-end + version + VMG
    category + provider ID + number-of-title-sets + sector pointers
    to FP_PGC / menu VOB / TT_SRPT / VMGM_PGCI_UT / VMG_PTL_MAIT /
    VMG_VTS_ATRT / TXTDT_MG / VMGM_C_ADT / VMGM_VOBU_ADMAP.
  • VTSI_MAT — full VTS_xx_0.IFO Video Title Set Information
    Management Table parse: title-set last sector + IFO-end + version
    • VTS category + sector pointers to PTT_SRPT / PGCI / VTSM_PGCI_UT
      / TMAPTI / VTSM_C_ADT / VTSM_VOBU_ADMAP / VTS_C_ADT / VTS_VOBU_ADMAP.
  • TT_SRPT — Title Search Pointer Table walker (8-byte header
    • N × 12-byte entries) exposing per-title (VTS_number, VTS_TTN, chapter_count, angle_count, parental_mask, vts_start_sector).
  • VTS_PTT_SRPT — Part-of-Title (chapter) search pointer table
    walker with per-title PTT body inferred from the offset list
    (boundaries derived from the next-title offset, or end_address + 1 for the last title).
  • VTS_PGCI — Program Chain Information table: 8-byte header +
    SRP list (per-PGC category + offset) + each PGC's 0xEC-byte
    header (nr_of_programs, nr_of_cells, BCD playback time +
    frame-rate bits, prohibited UOPs, next/prev/goup PGCN, still
    time, playback mode) + program map + Cell Playback Information
    Table (24 bytes per cell — category, restricted flag, still
    time, cell command, BCD playback time, first/last VOBU start +
    ILVU/last-VOBU end sectors) + Cell Position Information Table
    (4 bytes per cell — VOB ID + Cell ID).
  • VTS_C_ADT — Cell Address Table walker (shared format with
    VMGM_C_ADT + VTSM_C_ADT) — entry count recovered from the
    end_address header field, (vob_id, cell_id) → (start_sector, end_sector) lookup helper.
  • PgcTime — BCD playback-time decoder for the hh:mm:ss:ff
    field with FrameRate::{Pal25, Ntsc30, Illegal, Reserved}
    discrimination (bits 7+6 of frame byte per mpucoder-pgc.html).
  • VtsIfo materialiser — parse(buf, vts_number) -> VtsIfo
    pulls VTSI_MAT, PTT_SRPT, PGCI, and C_ADT into a single view
    and rebuilds per-title chapter lists (DvdTitle → Vec<DvdChapter>,
    each chapter carrying its first/last cell numbers derived from
    the PGC's program map).
  • DvdDisc Phase-2 API — parse_vmg(&reader) -> VmgIfo,
    parse_vts(&reader, ts_index) -> VtsIfo, enumerate_titles( &reader) -> Vec<DvdTitleEntry>, and a parse_vmg_tt_srpt
    convenience accessor.

• Phase 2 tests (all synthetic): VMGI_MAT parse + bad-magic
  rejection, VTSI_MAT parse, TT_SRPT walk (3 titles), PGCI with one
  PGC + 3 cells, VTS_PTT_SRPT walking 2 titles × 5 chapters, VTS_C_ADT
  with 4 cell entries, PgcTime decode (NTSC 30 fps + PAL 25 fps), and
  a full hand-built 4-sector composite VTS_xx_0.IFO image (VTSI_MAT

  • PTT_SRPT + PGCI + C_ADT) round-tripped through VtsIfo::parse
    with chapter-cell-range assertions.

• Bootstrap (Phase 1 — filesystem + disc detection): clean-room
  read-only DVD-Video support per ECMA-267 (DVD-ROM) + ECMA-268
  (DVD-ROM file system) + OSTA UDF 1.02 + the ECMA-167 UDF base
  standard. NO libdvdread, libdvdnav, libdvdcss, FFmpeg, xine, mpv,
  or VLC source consulted.

  • ISO 9660 reader (iso9660 module) — Primary Volume Descriptor
    at sector 16, root directory record + path-table walk, A-string /
    D-string decode, recursive directory enumeration.
  • UDF 1.02 mount (udf module) — Anchor Volume Descriptor
    Pointer probing (sectors 256 / 512 / N-256), Volume Descriptor
    Sequence (Primary VD, Partition Descriptor, Logical Volume
    Descriptor, Terminating Descriptor), File Set Descriptor, root
    File Identifier Descriptor walk, File Entry / ICB with Short_ad
    / Long_ad / Ext_ad allocation descriptors, OSTA compressed
    Unicode (compression IDs 8 + 16) per UDF 1.02 §2.1.3.
  • DVD-Video disc detection (disc module) — sniff for ISO 9660
    PVD + UDF AVDP on a file or block device, require a top-level
    VIDEO_TS/ directory containing VIDEO_TS.IFO, enumerate
    VIDEO_TS.IFO + .VOB + .BUP + per-VTS VTS_xx_0.IFO /
    VTS_xx_0.VOB (menu) / VTS_xx_1..9.VOB (title) / VTS_xx_0.BUP.
  • dvd:// source driver (source module, default-on registry
    feature) — registers under oxideav_core::SourceRegistry so a
    dvd:///path/to/disc.iso URL surfaces a typed DvdDiscSource
    that carries the file enumeration + byte-range read.

• Tests (all synthetic, no real disc data): ISO 9660 PVD / d-string /
  path-table / dir walk / nested dir / EOF rejection; UDF 1.02 AVDP /
  tag checksum / LVID / FSD / FID iteration / Short_ad / Long_ad /
  Ext_ad / d-string compression-mode 8 vs 16; DVD-Video single-VTS,
  multi-VTS, AUDIO_TS-empty, rejection-when-no-VIDEO_TS; one full
  round-trip against a hand-assembled ~64 KB synthetic disc image
  under tests/data/.