v0.18.9: bump version

Complements the wedge-skip backstop (fbdb50c) with proactive
avoidance so we don't HIT the wedge in the first place. User's
take after seeing the Dune Pt 2 rip wedge at 48%: 'we shouldn't be
wedging.'

Empirical observations from the 23:09:12-23:09:55 wedge timeline:

  5 read errors over 43 s, ~8 s apart (drive's own ECC recovery
  takes 5-10 s per failure). Not 'hammering' in any usual sense,
  but cumulative firmware-state buildup over 5 in-cluster errors
  was enough to tip the BU40N into wedge mode at the 5th error.

  Damage cluster spanned ~140 MB (LBAs 19.898M-19.965M). Current
  damage-jump base of 256 sectors × batch=32 = 16 MB first jump,
  doubling to 32 MB, 64 MB... Each jump landed BACK INSIDE the
  140 MB cluster, exposing the drive to MORE in-cluster errors.

Two avoidance levers:

1. Inter-error pause on Pass 1 (PASS_1_FAIL_PAUSE_SECS = 5 s):
   pre-fix Pass 1 ran pause_secs=0 on all errors to 'zoom past'
   damage zones. Successful reads still zoom at zero pause — the
   pause applies only to FAILED reads, giving the drive's firmware
   cool-down between cluster exposures. Cost: ~5 s per scattered
   failure (~30-60 s total on a damage cluster); trivial vs.
   crashing the rip at 48%.

2. Larger damage-jump base (JUMP_BASE_SECTORS = 1024, up from
   256): first jump at batch=32 now covers 64 MB instead of 16 MB,
   second jump 128 MB instead of 32 MB. Two jumps clear 192 MB —
   well past most single-cluster damage patterns. Smaller jumps
   were landing inside the cluster and adding to the wedge counter.

Plus a halt-aware sleep helper (sleep_secs_or_halt) so the new
inter-error pause doesn't degrade halt response time. Halt poll
granularity 100 ms — halt fires within ~100 ms regardless of
remaining pause time. Updated three sleep call sites in disc/mod.rs
(SkipBlock pause, JumpAhead post-pause, Retry pause).

The wedge-SKIP backstop (fbdb50c) stays — combined with this
avoidance work, the flow becomes:
  damage cluster encountered →
    pause 5 s, mark NonTrimmed →
    second failure →
    pause 5 s, mark NonTrimmed →
    ...
    threshold hit →
    damage-jump 64 MB (clears 95% of clusters) →
    if jump lands in another cluster: 128 MB next jump →
    only if drive STILL wedges after all this:
      wedge-skip kicks in (1 GB jump + 30 s cooldown × 16 budget)

Tests:
  pass_1_pauses_briefly_on_skip_for_wedge_avoidance — locks the
    new 5 s pause behavior in place (replaces the old pause=0 test).
  integration test threshold bumped from 5 s to 60 s with comment
    explaining the new bound is 'not infinite' rather than
    'milliseconds-fast'.
  All 433+ tests green on cargo +1.86 fmt + clippy + test.

Precommit green.

.github/workflows/ci.yml

@@ -14,6 +14,10 @@ jobs:
         with:
           components: clippy, rustfmt
       - run: cargo fmt --check
+      # libfreemkv is a library — Cargo.lock is gitignored. --locked
+      # would always fail on a fresh runner because there's no committed
+      # lockfile to lock against. The binary crates (freemkv, autorip,
+      # bdemu) track Cargo.lock and DO use --locked.
       - run: cargo clippy -- -D warnings
 
   test:

.github/workflows/release.yml

@@ -28,6 +28,8 @@ jobs:
     steps:
       - uses: actions/checkout@v5
       - uses: dtolnay/rust-toolchain@1.86.0
+      # libfreemkv is a library — Cargo.lock isn't tracked, so --locked
+      # would always fail (no lockfile to lock against on a fresh runner).
       - run: cargo test
 
   publish:

CHANGELOG.md

@@ -1,5 +1,23 @@
 # Changelog
 
+## 0.18.4 (2026-05-09)
+
+### Build / CI hardening — no library code changes
+
+- All `cargo` invocations in `.github/workflows/*.yml` now use
+  `--locked`. Previously `cargo build` / `cargo test` could silently
+  re-resolve `Cargo.lock` if a lock-pinned version wasn't yet on
+  crates.io — masking dependency races between same-tag releases of
+  sibling crates. The 0.18.3 release hit this: autorip's docker image
+  built `libfreemkv v0.18.2` because libfreemkv 0.18.3 hadn't
+  published yet at the time autorip's CI ran, and cargo silently fell
+  back to the previous version. With `--locked`, that race now hard-
+  fails the build with a clear "package X is not in registry" error
+  and we retrigger after the upstream lands.
+
+- `cargo publish --locked` in the release workflow guards against
+  publishing a lockfile-mismatched release.
+
 ## 0.18.3 (2026-05-09)
 
 ### Behaviour change

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "libfreemkv"
-version = "0.18.3"
+version = "0.18.9"
 edition = "2024"
 rust-version = "1.86"
 license = "AGPL-3.0-only"

src/disc/mod.rs

@@ -1634,9 +1634,7 @@ impl Disc {
 
                         match action {
                             read_error::ReadAction::Retry { pause_secs } => {
-                                if pause_secs > 0 {
-                                    std::thread::sleep(std::time::Duration::from_secs(pause_secs));
-                                }
+                                sleep_secs_or_halt(pause_secs, opts.halt.as_ref());
                             }
                             read_error::ReadAction::Bisect => {
                                 read_ctx.bisecting = true;
@@ -1713,9 +1711,7 @@ impl Disc {
                                     break 'outer;
                                 }
                                 bytes_done = bytes_done.saturating_add(block_bytes);
-                                if pause_secs > 0 {
-                                    std::thread::sleep(std::time::Duration::from_secs(pause_secs));
-                                }
+                                sleep_secs_or_halt(pause_secs, opts.halt.as_ref());
                                 pos += block_bytes;
                             }
                             read_error::ReadAction::JumpAhead {
@@ -1770,9 +1766,7 @@ impl Disc {
                                     "damage-jump"
                                 );
                                 pos = jump_pos;
-                                if pause_secs > 0 {
-                                    std::thread::sleep(std::time::Duration::from_secs(pause_secs));
-                                }
+                                sleep_secs_or_halt(pause_secs, opts.halt.as_ref());
                             }
                             read_error::ReadAction::AbortPass => {
                                 let (status, sense) = extract_scsi_context(&err);
@@ -1974,6 +1968,34 @@ pub struct PatchOutcome {
     pub wedged_threshold: u64,
 }
 
+/// Sleep `secs` seconds, but break early if `halt` flips to true.
+/// Used by Pass 1's wedge-avoidance inter-error pause so halt
+/// remains responsive regardless of how long the pause is.
+/// Polling granularity 100 ms — bounded latency on halt regardless
+/// of pause length.
+pub(crate) fn sleep_secs_or_halt(
+    secs: u64,
+    halt: Option<&std::sync::Arc<std::sync::atomic::AtomicBool>>,
+) {
+    if secs == 0 {
+        return;
+    }
+    let Some(h) = halt else {
+        std::thread::sleep(std::time::Duration::from_secs(secs));
+        return;
+    };
+    let total = std::time::Duration::from_secs(secs);
+    let slice = std::time::Duration::from_millis(100);
+    let start = std::time::Instant::now();
+    while start.elapsed() < total {
+        if h.load(std::sync::atomic::Ordering::Relaxed) {
+            return;
+        }
+        let remaining = total.saturating_sub(start.elapsed());
+        std::thread::sleep(remaining.min(slice));
+    }
+}
+
 pub fn mapfile_path_for(iso_path: &std::path::Path) -> std::path::PathBuf {
     let mut s = iso_path.as_os_str().to_os_string();
     s.push(".mapfile");
@@ -3290,7 +3312,12 @@ mod tests {
 
     /// Build a DiscTitle with full control over the fields the title
     /// sorter cares about. Used by the canonical-title-order tests.
-    fn title_with(playlist: &str, duration_secs: f64, size_bytes: u64, n_clips: usize) -> DiscTitle {
+    fn title_with(
+        playlist: &str,
+        duration_secs: f64,
+        size_bytes: u64,
+        n_clips: usize,
+    ) -> DiscTitle {
         let mut t = title_with_video(Codec::Hevc, Resolution::R2160p);
         t.playlist = playlist.into();
         t.duration_secs = duration_secs;
@@ -3317,13 +3344,24 @@ mod tests {
         const CAPACITY: u64 = 58_500_000_000; // 58.5 GB
         let mut titles = vec![
             // Title 1 in the raw MPLS order — virtual play-all
-            title_with("00020.mpls", 4.0 * 3600.0 + 13.0 * 60.0, 92_400_000_000, 253),
+            title_with(
+                "00020.mpls",
+                4.0 * 3600.0 + 13.0 * 60.0,
+                92_400_000_000,
+                253,
+            ),
             // Title 2 — actual movie
             title_with("00800.mpls", 2.0 * 3600.0 + 2.0 * 60.0, 57_200_000_000, 1),
         ];
         titles.sort_by(|a, b| Disc::canonical_title_order(a, b, CAPACITY));
-        assert_eq!(titles[0].playlist, "00800.mpls", "main feature should land at index 0");
-        assert_eq!(titles[1].playlist, "00020.mpls", "virtual play-all should be pushed back");
+        assert_eq!(
+            titles[0].playlist, "00800.mpls",
+            "main feature should land at index 0"
+        );
+        assert_eq!(
+            titles[1].playlist, "00020.mpls",
+            "virtual play-all should be pushed back"
+        );
     }
 
     /// Non-branching disc: longest 1-clip title is the movie. Sort
@@ -3333,12 +3371,15 @@ mod tests {
     fn canonical_order_preserves_natural_ranking_on_normal_disc() {
         const CAPACITY: u64 = 60_000_000_000;
         let mut titles = vec![
-            title_with("00100.mpls", 600.0, 5_000_000_000, 1),  // 10 min menu
+            title_with("00100.mpls", 600.0, 5_000_000_000, 1), // 10 min menu
             title_with("00800.mpls", 7320.0, 55_000_000_000, 1), // 2h02m main feature
             title_with("00200.mpls", 1800.0, 2_000_000_000, 1), // 30 min extra
         ];
         titles.sort_by(|a, b| Disc::canonical_title_order(a, b, CAPACITY));
-        assert_eq!(titles[0].playlist, "00800.mpls", "longest valid title still wins");
+        assert_eq!(
+            titles[0].playlist, "00800.mpls",
+            "longest valid title still wins"
+        );
         assert_eq!(titles[1].playlist, "00200.mpls");
         assert_eq!(titles[2].playlist, "00100.mpls");
     }

src/disc/read_error.rs

@@ -25,11 +25,25 @@ pub struct ReadCtx {
     /// Failed reads since the last success. Resets to 0 on success.
     /// Drives long-pause escalation on persistent failure.
     pub consecutive_failures: u64,
+    /// Failed OUTER batch reads since the last outer success — bisect
+    /// inner-sector failures are NOT counted here. Drives the
+    /// fast-entry damage-jump on Pass 1 (skip the disc-level grind
+    /// once we're clearly in a damaged region; Pass N will recover
+    /// the actual sectors). Reset on outer success.
+    pub consecutive_outer_failures: u64,
     /// Sliding window of recent read outcomes (true=ok, false=fail).
     /// Capped at `damage_window_max`. Drives damage-jump decisions.
     pub damage_window: Vec<bool>,
     pub damage_window_max: usize,
     pub damage_threshold_pct: usize,
+    /// Trigger a damage-jump after this many consecutive outer-batch
+    /// failures, even when the damage_window isn't full yet. Pass 1
+    /// uses a small value (4) so we don't spend ~40 minutes grinding
+    /// to fill a 16-block window before the first jump on a damage
+    /// zone we entered cleanly. Pass N uses a larger value (or
+    /// disables this — see `bisect_on_marginal`) because Pass N's
+    /// whole job IS to grind on the bad ranges.
+    pub fast_jump_threshold: u64,
     /// Multiplier applied to damage-jump distance. Doubles each jump,
     /// resets to 1 after `damage_window_max` consecutive good reads.
     pub jump_multiplier: u64,
@@ -43,38 +57,73 @@ pub struct ReadCtx {
     /// failed batch, so the handler doesn't recursively request another
     /// bisect on the inner-sector failures.
     pub bisecting: bool,
+    /// Whether to return `Bisect` on a marginal-media batch failure.
+    /// Pass 1 sweep sets this false: a failed batch becomes
+    /// SkipBlock (mark the whole 32-sector ECC block NonTrimmed,
+    /// advance, let Pass N recover the salvageable sectors with
+    /// proper recovery semantics). Pass N sets this true: bisection
+    /// is its core job, and it has the right tools (single-sector
+    /// reads, 60s recovery timeout, retry budget, escalating skip).
+    pub bisect_on_marginal: bool,
+    /// Count of consecutive firmware-wedge responses (HARDWARE_ERROR
+    /// or ILLEGAL_REQUEST sense keys) since the last successful read.
+    /// Pass 1 uses this to drive the wedge-skip path: each wedge
+    /// triggers a 1 GB jump + cooldown pause. Reaching
+    /// `WEDGE_ABORT_THRESHOLD` consecutive wedges with no good read
+    /// in between → real AbortPass.
+    pub wedge_count: u64,
 }
 
 impl ReadCtx {
-    /// Initial context for a Pass 1 sweep with the documented constants.
+    /// Initial context for a Pass 1 sweep. The job is "fast and
+    /// accurate, get the most data in the shortest time" — Pass N
+    /// is the one that grinds on the bad ranges. So bisect-on-
+    /// marginal is OFF (failed batches become SkipBlock; whole 32-
+    /// sector blocks marked NonTrimmed for Pass N to revisit), and
+    /// the damage-jump fast-path triggers after just 4 consecutive
+    /// outer-batch failures.
     pub fn for_sweep(batch: u16) -> Self {
         Self {
             batch,
             consecutive_good: 0,
             consecutive_failures: 0,
+            consecutive_outer_failures: 0,
             damage_window: Vec::with_capacity(16),
             damage_window_max: 16,
             damage_threshold_pct: 12,
+            fast_jump_threshold: 4,
             jump_multiplier: 1,
             not_ready_retries: 0,
             bridge_degradation_count: 0,
             bisecting: false,
+            bisect_on_marginal: false,
+            wedge_count: 0,
         }
     }
 
-    /// Initial context for a Pass 2-N patch.
+    /// Initial context for a Pass 2-N patch. Pass N's whole reason to
+    /// exist is to recover sectors Pass 1 skipped — bisection on
+    /// marginal media is part of the job, and the fast-jump
+    /// threshold is loose so we don't bail too early on a range that
+    /// has scattered good sectors mixed in.
     pub fn for_patch(batch: u16) -> Self {
         Self {
             batch,
             consecutive_good: 0,
             consecutive_failures: 0,
+            consecutive_outer_failures: 0,
             damage_window: Vec::with_capacity(16),
             damage_window_max: 16,
             damage_threshold_pct: 12,
+            // Pass N is allowed to grind: window-based jump only,
+            // matching the historical behaviour for patch passes.
+            fast_jump_threshold: u64::MAX,
             jump_multiplier: 1,
             not_ready_retries: 0,
             bridge_degradation_count: 0,
             bisecting: false,
+            bisect_on_marginal: true,
+            wedge_count: 0,
         }
     }
 
@@ -83,6 +132,16 @@ impl ReadCtx {
         self.consecutive_good += 1;
         self.consecutive_failures = 0;
         self.not_ready_retries = 0;
+        // Any successful read clears the wedge-skip counter — the
+        // drive recovered, so further wedges should reset the skip
+        // budget instead of accumulating toward a real abort.
+        self.wedge_count = 0;
+        // Outer-success only: a good single-sector read inside a
+        // bisect doesn't mean we've left the damaged batch. Only an
+        // outer-batch success resets the outer-failure counter.
+        if !self.bisecting {
+            self.consecutive_outer_failures = 0;
+        }
         self.damage_window.push(true);
         if self.damage_window.len() > self.damage_window_max {
             self.damage_window.remove(0);
@@ -128,6 +187,63 @@ const NOT_READY_MAX_RETRIES: u32 = 3;
 const BRIDGE_DEGRADATION_PAUSE_SECS: u64 = 15;
 const BRIDGE_DEGRADATION_MAX_RETRIES: u32 = 5;
 
+/// Pass 1 inter-error pause. ZERO on the clean path (sweep zooms
+/// past damage windows) but applied to every failed read so the
+/// drive's firmware gets cool-down time between damage-cluster
+/// exposures. Wedge-avoidance change 2026-05-10: pre-fix `pause_secs
+/// = 0` on Pass 1 errors caused back-to-back exposures that
+/// accumulated firmware wedge state.
+const PASS_1_FAIL_PAUSE_SECS: u64 = 5;
+
+/// Base of the damage-jump distance formula: `jump_sectors =
+/// JUMP_BASE_SECTORS × batch × jump_multiplier`. Bumped 2026-05-10
+/// from 256 → 1024 (4×) so the first damage-jump at batch=32 covers
+/// 64 MB instead of 16 MB. Empirically the BU40N's damage clusters
+/// are 100+ MB wide; 16 MB jumps landed inside the cluster and the
+/// re-read added to the firmware wedge counter. 64 MB → 128 MB
+/// (after one doubling) clears almost any single-cluster damage in
+/// 2 jumps.
+const JUMP_BASE_SECTORS: u64 = 1024;
+
+// Firmware-wedge skip policy for Pass 1 sweep
+// ===========================================
+//
+// When the BU40N (or similar drives) hits a physical-damage cluster,
+// its firmware can transition into a "wedge" state where it returns
+// HARDWARE_ERROR or ILLEGAL_REQUEST for every subsequent read —
+// often for many LBAs after the actual bad sector. Per project docs
+// "Bad-sector handling" rule #2: "Recovery requires eject+reload OR
+// significant cool-down."
+//
+// Pass 1's pre-fix behavior was to immediately AbortPass on the
+// first HARDWARE_ERROR / ILLEGAL_REQUEST, killing the rip at
+// whatever percentage it had reached. That's the wrong call when:
+//   - the damage zone may be small (jumping past it could resume
+//     normal reads), AND
+//   - even if the drive stays wedged, finishing the sweep gives us
+//     an honest mapfile for Pass N to attack later.
+//
+// New policy: treat wedge sense codes the same way the damage-window
+// treats persistent failure — JumpAhead by a large distance with a
+// cooldown pause. Allow up to WEDGE_ABORT_THRESHOLD consecutive
+// wedges (no successful read in between) before declaring the drive
+// truly stuck and surfacing AbortPass to autorip.
+
+/// One-gigabyte jump (1024 MiB) on each wedge. Big enough to clear
+/// almost any single-cluster damage zone we've seen.
+const WEDGE_JUMP_SECTORS: u64 = 524_288;
+/// Cooldown pause after each wedge. Per project docs the drive needs
+/// "significant cool-down"; 30 s strikes a balance between giving
+/// the drive a chance to recover and not stalling the rip if the
+/// drive is permanently stuck.
+const WEDGE_PAUSE_SECS: u64 = 30;
+/// Bail after this many consecutive wedges with no good read in
+/// between. At 1 GB jumps this lets us scan ~16 GB worth of fully
+/// wedged area before giving up — generous enough to clear most
+/// physical-damage clusters, bounded enough to not loop forever on
+/// a permanently bricked drive.
+const WEDGE_ABORT_THRESHOLD: u64 = 16;
+
 /// THE single error-handling entry point. Updates `ctx`, returns the
 /// action the caller must apply.
 ///
@@ -137,6 +253,13 @@ const BRIDGE_DEGRADATION_MAX_RETRIES: u32 = 5;
 pub fn handle_read_error(err: &Error, ctx: &mut ReadCtx) -> ReadAction {
     ctx.consecutive_failures += 1;
     ctx.consecutive_good = 0;
+    // Outer-failure counter — only OUTER batch failures count toward
+    // the fast-jump trigger. Bisect inner failures are part of
+    // recovering an already-failed batch and don't represent
+    // independent damage signal.
+    if !ctx.bisecting {
+        ctx.consecutive_outer_failures += 1;
+    }
 
     tracing::warn!(
         target: "freemkv::disc",
@@ -186,11 +309,54 @@ pub fn handle_read_error(err: &Error, ctx: &mut ReadCtx) -> ReadAction {
         ctx.not_ready_retries = 0;
     }
 
-    // 4. Hardware / illegal request: the drive said "no, won't do it".
-    //    Retrying won't change the answer. Surface to the outer layer
-    //    so it can eject + prompt user.
+    // 4. Hardware error / illegal request — the firmware-wedge family.
+    //    The drive transitioned into a fast-fail state where it
+    //    rejects reads near the LBA. Two policies:
+    //
+    //    Pass 1 sweep (bisect_on_marginal=false): the wedge is
+    //      *recoverable* by skipping. Jump a large distance ahead
+    //      (1 GB), pause for cooldown, mark the skipped region
+    //      NonTrimmed so Pass N revisits. Allow up to
+    //      WEDGE_ABORT_THRESHOLD consecutive wedges before truly
+    //      giving up. This replaces the pre-fix "abort the entire
+    //      rip on first wedge" behavior that caused 48%-and-die
+    //      failures on discs with one bad cluster.
+    //
+    //    Pass N patch (bisect_on_marginal=true): Pass N's job IS
+    //      single-sector recovery; a wedge means the drive won't
+    //      give us the specific sectors we asked for. Skipping
+    //      doesn't help here. Abort and let autorip decide whether
+    //      to retry, eject, or surface the failure to the user.
     if sense_key == scsi::SENSE_KEY_HARDWARE_ERROR || sense_key == scsi::SENSE_KEY_ILLEGAL_REQUEST {
-        return ReadAction::AbortPass;
+        if ctx.bisect_on_marginal {
+            return ReadAction::AbortPass;
+        }
+        // Pass 1 wedge-skip path.
+        if !ctx.bisecting {
+            ctx.wedge_count += 1;
+        }
+        if ctx.wedge_count >= WEDGE_ABORT_THRESHOLD {
+            tracing::warn!(
+                target: "freemkv::disc",
+                phase = "wedge_abort",
+                wedge_count = ctx.wedge_count,
+                threshold = WEDGE_ABORT_THRESHOLD,
+                "Pass 1 wedge-skip exhausted — drive appears permanently stuck"
+            );
+            return ReadAction::AbortPass;
+        }
+        tracing::warn!(
+            target: "freemkv::disc",
+            phase = "wedge_skip",
+            wedge_count = ctx.wedge_count,
+            jump_sectors = WEDGE_JUMP_SECTORS,
+            pause_secs = WEDGE_PAUSE_SECS,
+            "Pass 1 wedge detected — jumping ahead and pausing for drive cooldown"
+        );
+        return ReadAction::JumpAhead {
+            sectors: WEDGE_JUMP_SECTORS,
+            pause_secs: WEDGE_PAUSE_SECS,
+        };
     }
 
     // 5. Marginal media (MEDIUM_ERROR / ABORTED_COMMAND) on a multi-
@@ -198,11 +364,20 @@ pub fn handle_read_error(err: &Error, ctx: &mut ReadCtx) -> ReadAction {
     //    individually. Bisect into single-sector reads (gentler on the
     //    bridge too — shorter SCSI transactions). Avoid recursive
     //    bisect.
+    //
+    //    Pass 1 sweep sets `bisect_on_marginal=false` to skip this:
+    //    its job is "fast and accurate, get the most data in the
+    //    shortest time." Pass N is purpose-built to recover
+    //    individual sectors with proper recovery semantics, and Pass
+    //    1 grinding through 32-sector bisects costs ~2.5 min per bad
+    //    block AND fills the damage window slower than it should.
+    //    Whole-block NonTrimmed → SkipBlock → advance → Pass N
+    //    revisits.
     let is_marginal = matches!(
         sense_key,
         scsi::SENSE_KEY_MEDIUM_ERROR | scsi::SENSE_KEY_ABORTED_COMMAND
     );
-    if is_marginal && ctx.batch > 1 && !ctx.bisecting {
+    if is_marginal && ctx.batch > 1 && !ctx.bisecting && ctx.bisect_on_marginal {
         return ReadAction::Bisect;
     }
 
@@ -228,22 +403,77 @@ pub fn handle_read_error(err: &Error, ctx: &mut ReadCtx) -> ReadAction {
         bad_count * 100 / ctx.damage_window.len()
     };
 
-    let pause_secs = if ctx.consecutive_failures >= CONSECUTIVE_FAIL_LONG_PAUSE_THRESHOLD {
+    // Inter-error pause on Pass 1 — wedge avoidance.
+    //
+    // Pre-2026-05-10 Pass 1 ran `pause_secs = 0` on all errors so
+    // sweep would zoom past damage zones in seconds instead of
+    // minutes. Empirically on the BU40N this caused firmware-wedge
+    // events: each read failure leaves residual state in the drive's
+    // firmware, and back-to-back errors without cooldown accumulate
+    // toward a wedge threshold. We hit wedge on Dune Pt 2 after 5
+    // errors in 43 s spread over a 140 MB damage cluster — even
+    // though wall-clock pacing was slow, the drive had no breathing
+    // room between exposures.
+    //
+    // New policy: Pass 1 keeps `pause = 0` on the CLEAN path
+    // (successful reads zoom past damage windows fine), but every
+    // failed read takes a PASS_1_FAIL_PAUSE_SECS pause before the
+    // next read. Cost: ~5 s extra per scattered failure, ~30-60 s
+    // total in a damage cluster — trivial compared to the alternative
+    // of crashing the whole rip at 48%. Long failure streaks still
+    // escalate via CONSECUTIVE_FAIL_LONG_PAUSE.
+    let pause_secs = if !ctx.bisect_on_marginal {
+        if ctx.consecutive_failures >= CONSECUTIVE_FAIL_LONG_PAUSE_THRESHOLD {
+            CONSECUTIVE_FAIL_LONG_PAUSE_SECS
+        } else {
+            PASS_1_FAIL_PAUSE_SECS
+        }
+    } else if ctx.consecutive_failures >= CONSECUTIVE_FAIL_LONG_PAUSE_THRESHOLD {
         CONSECUTIVE_FAIL_LONG_PAUSE_SECS
     } else {
         POST_FAILURE_PAUSE_SECS
     };
 
-    // 7. Damage-jump: too many failures in window → skip ahead by an
-    //    escalating gap. Multiplier capped so we can't accidentally
-    //    skip the entire rest of the disc (observed 2026-05-07: a
-    //    saturated multiplier produced a 56 GB jump). Saturating
-    //    arithmetic on the sector calc as defence in depth.
-    const MAX_JUMP_MULTIPLIER: u64 = 64; // 64 × 256 × batch sectors
-    if ctx.damage_window.len() >= ctx.damage_window_max && bad_pct >= ctx.damage_threshold_pct {
+    // 7. Damage-jump: too many failures → skip ahead by an escalating
+    //    gap. Multiplier capped so we can't accidentally skip the
+    //    entire rest of the disc (observed 2026-05-07: a saturated
+    //    multiplier produced a 56 GB jump). Saturating arithmetic on
+    //    the sector calc as defence in depth.
+    //
+    //    Jump base bumped 2026-05-10 from 256 to 1024 sectors per
+    //    multiplier unit (= 64 MB first jump at batch=32, up from
+    //    16 MB). The smaller base routinely landed jumps back inside
+    //    damage clusters of 100+ MB, each landing adding to the
+    //    firmware's wedge counter. 64 MB initial + 128 MB second +
+    //    256 MB third clears almost any single-cluster damage
+    //    pattern we've seen in 2 jumps.
+    //
+    //    Two triggers, evaluated in order:
+    //
+    //    a. **Fast-entry** — `consecutive_outer_failures >= fast_jump_threshold`.
+    //       Fires on Pass 1 (threshold=4) so we don't spend ~40 min
+    //       grinding to fill a 16-block damage window before the
+    //       first jump on a damage zone we entered cleanly. Doesn't
+    //       fire on Pass N (threshold=u64::MAX).
+    //
+    //    b. **Window-based** — original behaviour: 12% bad in a
+    //       sliding window of 16 outer reads. Pass N's only path,
+    //       and Pass 1's fallback if the failures are scattered
+    //       enough that we don't hit the consecutive threshold.
+    const MAX_JUMP_MULTIPLIER: u64 = 64;
+    let fast_trigger = !ctx.bisecting && ctx.consecutive_outer_failures >= ctx.fast_jump_threshold;
+    let window_trigger =
+        ctx.damage_window.len() >= ctx.damage_window_max && bad_pct >= ctx.damage_threshold_pct;
+    if fast_trigger || window_trigger {
         let mult = ctx.jump_multiplier.min(MAX_JUMP_MULTIPLIER);
-        let sectors = 256u64.saturating_mul(ctx.batch as u64).saturating_mul(mult);
+        let sectors = JUMP_BASE_SECTORS
+            .saturating_mul(ctx.batch as u64)
+            .saturating_mul(mult);
         ctx.jump_multiplier = (ctx.jump_multiplier.saturating_mul(2)).min(MAX_JUMP_MULTIPLIER);
+        // Reset the outer-failure counter so a long damaged region
+        // doesn't keep firing fast-jump every read after the initial
+        // jump fired. The window-based trigger handles further jumps.
+        ctx.consecutive_outer_failures = 0;
         return ReadAction::JumpAhead {
             sectors,
             pause_secs: pause_secs + POST_JUMP_EXTRA_PAUSE_SECS,
@@ -285,16 +515,41 @@ mod tests {
         }
     }
 
+    fn illegal_request_err() -> Error {
+        Error::DiscRead {
+            sector: 100,
+            status: Some(2),
+            sense: Some(ScsiSense {
+                sense_key: scsi::SENSE_KEY_ILLEGAL_REQUEST,
+                asc: 0x24,
+                ascq: 0x00,
+            }),
+        }
+    }
+
     #[test]
-    fn medium_error_with_batch_gt_1_bisects() {
-        let mut ctx = ReadCtx::for_sweep(32);
+    fn pass_n_marginal_with_batch_gt_1_bisects() {
+        let mut ctx = ReadCtx::for_patch(32);
         let action = handle_read_error(&medium_err(), &mut ctx);
         assert_eq!(action, ReadAction::Bisect);
     }
 
+    #[test]
+    fn pass_1_marginal_skips_instead_of_bisecting() {
+        // Pass 1's job is "fast and accurate" — leave bisection to
+        // Pass N. A failed batch becomes SkipBlock (whole 32-sector
+        // block marked NonTrimmed for Pass N to revisit).
+        let mut ctx = ReadCtx::for_sweep(32);
+        let action = handle_read_error(&medium_err(), &mut ctx);
+        match action {
+            ReadAction::SkipBlock { .. } => {}
+            other => panic!("expected SkipBlock for Pass 1, got {other:?}"),
+        }
+    }
+
     #[test]
     fn medium_error_with_batch_1_skips() {
-        let mut ctx = ReadCtx::for_sweep(1);
+        let mut ctx = ReadCtx::for_patch(1);
         let action = handle_read_error(&medium_err(), &mut ctx);
         match action {
             ReadAction::SkipBlock { pause_secs } => assert!(pause_secs >= 1),
@@ -304,7 +559,7 @@ mod tests {
 
     #[test]
     fn medium_error_while_bisecting_does_not_recurse() {
-        let mut ctx = ReadCtx::for_sweep(32);
+        let mut ctx = ReadCtx::for_patch(32);
         ctx.bisecting = true;
         let action = handle_read_error(&medium_err(), &mut ctx);
         match action {
@@ -314,19 +569,165 @@ mod tests {
     }
 
     #[test]
-    fn hardware_error_aborts() {
+    fn pass_1_jumps_after_4_consecutive_outer_failures() {
+        // The fast-entry trigger: Pass 1 should JumpAhead after 4
+        // consecutive outer-batch failures, BEFORE the 16-block
+        // damage window has filled. Otherwise we spend ~40 minutes
+        // of bisecting/grinding to fill the window before the first
+        // jump on a damage zone we entered cleanly.
         let mut ctx = ReadCtx::for_sweep(32);
+        // First three should NOT jump (still under threshold of 4).
+        for _ in 0..3 {
+            let a = handle_read_error(&medium_err(), &mut ctx);
+            assert!(
+                !matches!(a, ReadAction::JumpAhead { .. }),
+                "should not jump until 4 consecutive outer failures"
+            );
+        }
+        // Fourth should jump.
+        let a = handle_read_error(&medium_err(), &mut ctx);
+        assert!(
+            matches!(a, ReadAction::JumpAhead { .. }),
+            "expected JumpAhead at 4th consecutive outer failure, got {a:?}"
+        );
+    }
+
+    #[test]
+    fn pass_n_does_not_fast_jump() {
+        // Pass N's whole reason to exist is to grind on bad ranges.
+        // It should NOT bail after 4 consecutive failures the way
+        // Pass 1 does — it bisects and skips with proper recovery.
+        let mut ctx = ReadCtx::for_patch(32);
+        for _ in 0..4 {
+            let a = handle_read_error(&medium_err(), &mut ctx);
+            assert!(
+                !matches!(a, ReadAction::JumpAhead { .. }),
+                "Pass N must not fast-jump; got {a:?}"
+            );
+        }
+    }
+
+    #[test]
+    fn outer_success_resets_consecutive_outer_failures() {
+        let mut ctx = ReadCtx::for_sweep(32);
+        for _ in 0..3 {
+            handle_read_error(&medium_err(), &mut ctx);
+        }
+        assert_eq!(ctx.consecutive_outer_failures, 3);
+        // An outer-success (bisecting=false) should reset the counter.
+        ctx.bisecting = false;
+        ctx.on_success();
+        assert_eq!(ctx.consecutive_outer_failures, 0);
+    }
+
+    #[test]
+    fn bisect_inner_success_does_not_reset_outer_counter() {
+        let mut ctx = ReadCtx::for_patch(32);
+        for _ in 0..3 {
+            handle_read_error(&medium_err(), &mut ctx);
+        }
+        assert_eq!(ctx.consecutive_outer_failures, 3);
+        // A successful inner-sector read during bisect is not the
+        // same as escaping the bad outer batch.
+        ctx.bisecting = true;
+        ctx.on_success();
+        assert_eq!(
+            ctx.consecutive_outer_failures, 3,
+            "bisect inner success must not reset outer-failure counter"
+        );
+    }
+
+    #[test]
+    fn pass_1_hardware_error_jumps_ahead_not_aborts() {
+        // New wedge-skip policy: Pass 1 (bisect_on_marginal=false)
+        // should JumpAhead with a 1 GB skip + cooldown pause instead
+        // of immediately aborting. Aborting on first wedge was the
+        // pre-fix behavior that killed rips at 48% on damaged discs.
+        let mut ctx = ReadCtx::for_sweep(32);
+        let action = handle_read_error(&hardware_err(), &mut ctx);
+        match action {
+            ReadAction::JumpAhead {
+                sectors,
+                pause_secs,
+            } => {
+                assert_eq!(sectors, WEDGE_JUMP_SECTORS);
+                assert_eq!(pause_secs, WEDGE_PAUSE_SECS);
+            }
+            other => panic!("expected JumpAhead, got {other:?}"),
+        }
+        assert_eq!(ctx.wedge_count, 1);
+    }
+
+    #[test]
+    fn pass_1_hardware_error_aborts_after_threshold() {
+        // After WEDGE_ABORT_THRESHOLD consecutive wedges with no good
+        // read in between, autorip should see a real AbortPass so it
+        // can surface "drive is stuck, power-cycle required" to the
+        // user — rather than looping forever on a permanently bricked
+        // drive.
+        let mut ctx = ReadCtx::for_sweep(32);
+        for i in 0..WEDGE_ABORT_THRESHOLD - 1 {
+            let action = handle_read_error(&hardware_err(), &mut ctx);
+            assert!(
+                matches!(action, ReadAction::JumpAhead { .. }),
+                "iter {i}: expected JumpAhead, got {action:?}"
+            );
+        }
+        // The Nth wedge crosses the threshold.
         let action = handle_read_error(&hardware_err(), &mut ctx);
         assert_eq!(action, ReadAction::AbortPass);
     }
 
     #[test]
-    fn long_failure_streak_extends_pause() {
-        let mut ctx = ReadCtx::for_sweep(1);
+    fn pass_1_good_read_resets_wedge_count() {
+        // A single successful read between wedges must clear the
+        // skip counter — otherwise a disc with a few scattered bad
+        // zones would eventually run out of skip budget even though
+        // the drive was recovering between zones.
+        let mut ctx = ReadCtx::for_sweep(32);
+        for _ in 0..(WEDGE_ABORT_THRESHOLD - 1) {
+            handle_read_error(&hardware_err(), &mut ctx);
+        }
+        assert_eq!(ctx.wedge_count, WEDGE_ABORT_THRESHOLD - 1);
+        ctx.on_success();
+        assert_eq!(ctx.wedge_count, 0);
+        // After the success, we should still get JumpAhead (not
+        // AbortPass) on the next wedge.
+        let action = handle_read_error(&hardware_err(), &mut ctx);
+        assert!(matches!(action, ReadAction::JumpAhead { .. }));
+    }
+
+    #[test]
+    fn pass_n_hardware_error_still_aborts() {
+        // Pass N (bisect_on_marginal=true) keeps the original
+        // AbortPass behavior — single-sector recovery can't make
+        // progress through a wedge, so the right answer is to bail
+        // and let the outer layer decide.
+        let mut ctx = ReadCtx::for_patch(1);
+        let action = handle_read_error(&hardware_err(), &mut ctx);
+        assert_eq!(action, ReadAction::AbortPass);
+    }
+
+    #[test]
+    fn pass_1_illegal_request_also_routes_to_wedge_skip() {
+        // ILLEGAL_REQUEST is the other half of the wedge family:
+        // drive saying "I won't parse your CDB" after entering the
+        // fast-fail state. Same treatment as HARDWARE_ERROR.
+        let mut ctx = ReadCtx::for_sweep(32);
+        let action = handle_read_error(&illegal_request_err(), &mut ctx);
+        assert!(matches!(action, ReadAction::JumpAhead { .. }));
+    }
+
+    #[test]
+    fn long_failure_streak_extends_pause_on_pass_n() {
+        // Pass N keeps the cooldown behaviour: after many consecutive
+        // failures, pauses extend to give the drive time to recover.
+        // Pass 1 explicitly does NOT pause — see
+        // `pass_1_does_not_pause_on_skip` below.
+        let mut ctx = ReadCtx::for_patch(1);
         for _ in 0..15 {
             handle_read_error(&medium_err(), &mut ctx);
         }
-        // After many consecutive failures we should be in the long-pause regime
         let final_action = handle_read_error(&medium_err(), &mut ctx);
         match final_action {
             ReadAction::SkipBlock { pause_secs } => {
@@ -339,6 +740,28 @@ mod tests {
         }
     }
 
+    #[test]
+    fn pass_1_pauses_briefly_on_skip_for_wedge_avoidance() {
+        // Pre-2026-05-10 Pass 1 ran pause_secs=0 on all errors (zoom
+        // past damage zones in seconds). That caused firmware wedges
+        // on the BU40N — back-to-back errors with no cooldown built
+        // up firmware state until the drive entered the wedge fast-
+        // fail mode. New policy: a brief inter-error pause (5 s) on
+        // Pass 1 to give the drive's firmware time to settle between
+        // damage-zone exposures. Successful reads remain zero-pause
+        // — only errors cost time, and only a few seconds per
+        // scattered failure. Trivial cost compared to crashing the
+        // whole rip at 48%.
+        let mut ctx = ReadCtx::for_sweep(32);
+        let action = handle_read_error(&medium_err(), &mut ctx);
+        match action {
+            ReadAction::SkipBlock { pause_secs } => {
+                assert_eq!(pause_secs, PASS_1_FAIL_PAUSE_SECS);
+            }
+            other => panic!("expected SkipBlock, got {other:?}"),
+        }
+    }
+
     #[test]
     fn damage_window_fills_then_jumps() {
         let mut ctx = ReadCtx::for_sweep(1);

src/labels/criterion.rs

@@ -3,7 +3,7 @@
 //! Clean structured XML with Content/Qualifier per stream and
 //! stream number mapping via playbackconfig.
 
-use super::{LabelPurpose, LabelQualifier, StreamLabel, StreamLabelType};
+use super::{LabelPurpose, LabelQualifier, ParseResult, StreamLabel, StreamLabelType, xml};
 use crate::sector::SectorReader;
 use crate::udf::UdfFs;
 use std::collections::HashMap;
@@ -12,7 +12,7 @@ pub fn detect(udf: &UdfFs) -> bool {
     super::jar_file_exists(udf, "streamproperties.xml")
 }
 
-pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<StreamLabel>> {
+pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<ParseResult> {
     let sp_data = super::read_jar_file(reader, udf, "streamproperties.xml")?;
     let sp_text = std::str::from_utf8(&sp_data).ok()?;
 
@@ -66,7 +66,8 @@ pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<StreamLab
     if labels.is_empty() {
         return None;
     }
-    Some(labels)
+    // High confidence: streamproperties.xml is fully structured.
+    Some(ParseResult::high(labels))
 }
 
 struct StreamInfo {
@@ -78,105 +79,68 @@ struct StreamInfo {
     qualifier: LabelQualifier,
 }
 
-fn parse_stream_infos(xml: &str) -> Vec<StreamInfo> {
+fn parse_stream_infos(text: &str) -> Vec<StreamInfo> {
     let mut infos = Vec::new();
-    let mut pos = 0;
 
-    while pos < xml.len() {
-        let (tag, stream_type) = if let Some(p) = xml[pos..].find("<AudioStreamInfos>") {
-            (p + pos, StreamLabelType::Audio)
-        } else if let Some(p) = xml[pos..].find("<SubtitleStreamInfos>") {
-            (p + pos, StreamLabelType::Subtitle)
-        } else {
-            break;
-        };
+    for (tag_name, stream_type) in [
+        ("AudioStreamInfos", StreamLabelType::Audio),
+        ("SubtitleStreamInfos", StreamLabelType::Subtitle),
+    ] {
+        let mut from = 0;
+        while let Some((start, end)) = xml::find_element(text, tag_name, from) {
+            let block = &text[start..end];
+            let id = xml::text(block, "ID").unwrap_or_default();
+            let lang_id = xml::text(block, "LangInfoID").unwrap_or_default();
+            let content = xml::text(block, "Content").unwrap_or_default();
+            let qualifier_str = xml::text(block, "Qualifier").unwrap_or_default();
 
-        let end_tag = match stream_type {
-            StreamLabelType::Audio => "</AudioStreamInfos>",
-            StreamLabelType::Subtitle => "</SubtitleStreamInfos>",
-        };
+            let (language, variant) = if lang_id.contains('_') {
+                let parts: Vec<&str> = lang_id.splitn(2, '_').collect();
+                (parts[0].to_lowercase(), parts[1].to_string())
+            } else {
+                (lang_id.to_lowercase(), String::new())
+            };
 
-        let block_end = match xml[tag..].find(end_tag) {
-            Some(p) => tag + p + end_tag.len(),
-            None => break,
-        };
+            let purpose = if content.eq_ignore_ascii_case("COMMENTARY") {
+                LabelPurpose::Commentary
+            } else {
+                LabelPurpose::Normal
+            };
 
-        let block = &xml[tag..block_end];
-        let id = extract_tag(block, "ID").unwrap_or_default();
-        let lang_id = extract_tag(block, "LangInfoID").unwrap_or_default();
-        let content = extract_tag(block, "Content").unwrap_or_default();
-        let qualifier_str = extract_tag(block, "Qualifier").unwrap_or_default();
+            let qualifier = match qualifier_str.to_ascii_uppercase().as_str() {
+                "SDH" => LabelQualifier::Sdh,
+                "DS" => LabelQualifier::DescriptiveService,
+                _ => LabelQualifier::None,
+            };
 
-        let (language, variant) = if lang_id.contains('_') {
-            let parts: Vec<&str> = lang_id.splitn(2, '_').collect();
-            (parts[0].to_lowercase(), parts[1].to_string())
-        } else {
-            (lang_id.to_lowercase(), String::new())
-        };
-
-        let purpose = match content.as_str() {
-            "COMMENTARY" => LabelPurpose::Commentary,
-            _ => LabelPurpose::Normal,
-        };
-
-        let qualifier = match qualifier_str.as_str() {
-            "SDH" => LabelQualifier::Sdh,
-            "DS" => LabelQualifier::DescriptiveService,
-            _ => LabelQualifier::None,
-        };
-
-        infos.push(StreamInfo {
-            id,
-            stream_type,
-            language,
-            variant,
-            purpose,
-            qualifier,
-        });
-        pos = block_end;
+            infos.push(StreamInfo {
+                id,
+                stream_type,
+                language,
+                variant,
+                purpose,
+                qualifier,
+            });
+            from = end;
+        }
     }
     infos
 }
 
-fn parse_playback_config(xml: &str, map: &mut HashMap<String, u16>) {
-    let mut pos = 0;
-    while pos < xml.len() {
-        let tag_start = if let Some(p) = xml[pos..].find("<AudioStreams>") {
-            Some(p + pos)
-        } else {
-            xml[pos..].find("<SubtitlesStreams>").map(|p| p + pos)
-        };
-
-        let tag_start = match tag_start {
-            Some(p) => p,
-            None => break,
-        };
-
-        let block_end = xml[tag_start..]
-            .find("</AudioStreams>")
-            .or_else(|| xml[tag_start..].find("</SubtitlesStreams>"))
-            .map(|p| tag_start + p + 20)
-            .unwrap_or(xml.len());
-
-        let block = &xml[tag_start..block_end];
-
-        if let (Some(stream_id_str), Some(info_id)) = (
-            extract_tag(block, "StreamID"),
-            extract_tag(block, "StreamInfo_ID"),
-        ) {
-            if let Ok(stream_num) = stream_id_str.parse::<u16>() {
-                map.insert(info_id, stream_num);
+fn parse_playback_config(text: &str, map: &mut HashMap<String, u16>) {
+    for tag_name in ["AudioStreams", "SubtitlesStreams"] {
+        let mut from = 0;
+        while let Some((start, end)) = xml::find_element(text, tag_name, from) {
+            let block = &text[start..end];
+            if let (Some(stream_id_str), Some(info_id)) = (
+                xml::text(block, "StreamID"),
+                xml::text(block, "StreamInfo_ID"),
+            ) {
+                if let Ok(stream_num) = stream_id_str.parse::<u16>() {
+                    map.insert(info_id, stream_num);
+                }
             }
+            from = end;
         }
-
-        pos = block_end;
     }
 }
-
-fn extract_tag(xml: &str, tag: &str) -> Option<String> {
-    let open = format!("<{tag}>");
-    let close = format!("</{tag}>");
-    let start = xml.find(&open)? + open.len();
-    let end = xml[start..].find(&close)? + start;
-    Some(xml[start..end].trim().to_string())
-}

src/labels/ctrm.rs

@@ -4,7 +4,7 @@
 //! When both exist, language_streams.txt provides structured types while
 //! menu_base.prop provides stream number → button name mapping.
 
-use super::{LabelPurpose, LabelQualifier, StreamLabel, StreamLabelType, vocab};
+use super::{LabelPurpose, LabelQualifier, ParseResult, StreamLabel, StreamLabelType, vocab};
 use crate::sector::SectorReader;
 use crate::udf::UdfFs;
 use std::collections::HashMap;
@@ -14,7 +14,7 @@ pub fn detect(udf: &UdfFs) -> bool {
         || super::jar_file_exists(udf, "language_streams.txt")
 }
 
-pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<StreamLabel>> {
+pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<ParseResult> {
     // Try language_streams.txt first (richer structured data)
     let ls_labels = parse_language_streams(reader, udf);
 
@@ -22,12 +22,18 @@ pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<StreamLab
     let mb_labels = parse_menu_base(reader, udf);
 
     // If we have both, merge: language_streams for structure, menu_base for names
-    match (ls_labels, mb_labels) {
-        (Some(ls), Some(mb)) => Some(merge(ls, mb)),
-        (Some(ls), None) => Some(ls),
-        (None, Some(mb)) => Some(mb),
-        (None, None) => None,
+    let labels = match (ls_labels, mb_labels) {
+        (Some(ls), Some(mb)) => merge(ls, mb),
+        (Some(ls), None) => ls,
+        (None, Some(mb)) => mb,
+        (None, None) => return None,
+    };
+    if labels.is_empty() {
+        return None;
     }
+    // High confidence: both language_streams.txt and menu_base.prop
+    // are structured key-value formats with documented types.
+    Some(ParseResult::high(labels))
 }
 
 fn merge(ls: Vec<StreamLabel>, mb: Vec<StreamLabel>) -> Vec<StreamLabel> {
@@ -172,6 +178,164 @@ fn parse_language_streams(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<
     Some(labels)
 }
 
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// Build a minimal menu_base.prop text and run `parse_menu_base`'s
+    /// inner logic via a temporary closure. This isolates the prop
+    /// parsing without needing a SectorReader.
+    fn parse_props(text: &str) -> Vec<StreamLabel> {
+        // Mirror the inner loop of parse_menu_base exactly. Kept
+        // separate so the test doesn't need disc fixtures.
+        use std::collections::HashMap;
+        let mut entries: HashMap<String, HashMap<String, String>> = HashMap::new();
+        for line in text.lines() {
+            let line = line.trim();
+            if line.is_empty() || line.starts_with('#') {
+                continue;
+            }
+            let Some(eq_pos) = line.find('=') else {
+                continue;
+            };
+            let full_key = &line[..eq_pos];
+            let value = &line[eq_pos + 1..];
+            if let Some(dot_pos) = full_key.rfind('.') {
+                entries
+                    .entry(full_key[..dot_pos].to_string())
+                    .or_default()
+                    .insert(full_key[dot_pos + 1..].to_string(), value.to_string());
+            }
+        }
+        let mut labels = Vec::new();
+        for (prefix, props) in &entries {
+            let is_audio = props
+                .get("class")
+                .is_some_and(|c| c.contains("AudioButton"))
+                || prefix.starts_with("audio_");
+            let is_subtitle = props
+                .get("class")
+                .is_some_and(|c| c.contains("SubtitleButton"))
+                || prefix.starts_with("subtitle_");
+            let stream_num_str = props
+                .get("streamNumber")
+                .or_else(|| props.get("audioStream"))
+                .or_else(|| props.get("subtitleStream"));
+            let stream_num: u16 = match stream_num_str.and_then(|s| s.parse().ok()) {
+                Some(n) if n > 0 => n,
+                _ => continue,
+            };
+            if !is_audio && !is_subtitle {
+                continue;
+            }
+            let name = props.get("name").cloned().unwrap_or_default();
+            let purpose = match vocab::purpose(&name) {
+                LabelPurpose::Normal if prefix.contains("comm") => LabelPurpose::Commentary,
+                p => p,
+            };
+            let qualifier = if is_subtitle {
+                vocab::qualifier(&name)
+            } else {
+                LabelQualifier::None
+            };
+            let stream_type = if is_audio {
+                StreamLabelType::Audio
+            } else {
+                StreamLabelType::Subtitle
+            };
+            let language = props
+                .get("audioLanguage")
+                .or_else(|| props.get("subtitleLanguage"))
+                .cloned()
+                .unwrap_or_default();
+            labels.push(StreamLabel {
+                stream_number: stream_num,
+                stream_type,
+                language,
+                name,
+                purpose,
+                qualifier,
+                codec_hint: String::new(),
+                variant: String::new(),
+            });
+        }
+        labels.sort_by_key(|l| (l.stream_type as u8, l.stream_number));
+        labels
+    }
+
+    #[test]
+    fn commentary_via_name() {
+        let labels = parse_props(
+            "audio_1.class=AudioButton\n\
+             audio_1.streamNumber=2\n\
+             audio_1.name=Director's Commentary\n\
+             audio_1.audioLanguage=eng\n",
+        );
+        assert_eq!(labels.len(), 1);
+        assert_eq!(labels[0].purpose, LabelPurpose::Commentary);
+        assert_eq!(labels[0].language, "eng");
+    }
+
+    #[test]
+    fn commentary_via_prefix_when_name_silent() {
+        let labels = parse_props(
+            "audio_commentary_1.class=AudioButton\n\
+             audio_commentary_1.streamNumber=2\n\
+             audio_commentary_1.name=Track 2\n\
+             audio_commentary_1.audioLanguage=eng\n",
+        );
+        assert_eq!(labels[0].purpose, LabelPurpose::Commentary);
+    }
+
+    #[test]
+    fn commenter_does_not_false_match_commentary() {
+        // Regression for the pre-refactor `name.contains("comment")`
+        // bug: this would wrongly classify a "Commenter Pro" track as
+        // Commentary. vocab::purpose enforces a word boundary.
+        let labels = parse_props(
+            "audio_1.class=AudioButton\n\
+             audio_1.streamNumber=2\n\
+             audio_1.name=Commenter Pro Track\n\
+             audio_1.audioLanguage=eng\n",
+        );
+        assert_eq!(labels[0].purpose, LabelPurpose::Normal);
+    }
+
+    #[test]
+    fn descriptive_via_name() {
+        let labels = parse_props(
+            "audio_1.class=AudioButton\n\
+             audio_1.streamNumber=3\n\
+             audio_1.name=English Descriptive Audio\n",
+        );
+        assert_eq!(labels[0].purpose, LabelPurpose::Descriptive);
+    }
+
+    #[test]
+    fn sdh_only_on_subtitles() {
+        // SDH applied to a subtitle stream.
+        let labels = parse_props(
+            "subtitle_1.class=SubtitleButton\n\
+             subtitle_1.streamNumber=4\n\
+             subtitle_1.name=English SDH\n",
+        );
+        assert_eq!(labels[0].qualifier, LabelQualifier::Sdh);
+    }
+
+    #[test]
+    fn sdh_not_applied_to_audio_stream_even_if_name_contains_sdh() {
+        // Audio streams should not pick up SDH (it's a subtitle
+        // concept). Edge case: badly-authored name happens to include
+        // "SDH" — we don't propagate it to audio metadata.
+        let labels = parse_props(
+            "audio_1.class=AudioButton\n\
+             audio_1.streamNumber=5\n\
+             audio_1.name=English SDH (track?)\n",
+        );
+        assert_eq!(labels[0].qualifier, LabelQualifier::None);
+    }
+}
+
 // ── menu_base.prop parser ──────────────────────────────────────────────────
 
 fn parse_menu_base(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<StreamLabel>> {
@@ -231,16 +395,20 @@ fn parse_menu_base(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<Str
         }
 
         let name = props.get("name").cloned().unwrap_or_default();
-        let name_lower = name.to_lowercase();
 
-        let purpose = if name_lower.contains("comment") || prefix.contains("comm") {
-            LabelPurpose::Commentary
-        } else {
-            LabelPurpose::Normal
+        // Purpose: ask vocab first (word-boundary matched — avoids the
+        // "Commenter" false positive the prior `name.contains("comment")`
+        // had). Then fall back to the structural prefix check
+        // (`audio_commentary.foo`-style keys group commentary streams
+        // regardless of display name).
+        let purpose = match vocab::purpose(&name) {
+            LabelPurpose::Normal if prefix.contains("comm") => LabelPurpose::Commentary,
+            p => p,
         };
 
-        let qualifier = if is_subtitle && name_lower.contains("sdh") {
-            LabelQualifier::Sdh
+        // Qualifier: only apply to subtitles (SDH is a subtitle concept).
+        let qualifier = if is_subtitle {
+            vocab::qualifier(&name)
         } else {
             LabelQualifier::None
         };

src/labels/dbp.rs

@@ -0,0 +1,261 @@
+//! "dbp" framework — Magnolia Pictures BD-J authoring shop (per
+//! `bd-live.magpictures.com` referenced in the disc's
+//! `com/dbp/bluray.MenuXlet.perm`). Detected on UHD discs whose
+//! `/BDMV/JAR/<x>.jar` (top-level, not in a subdir) contains
+//! `com/dbp/` package paths.
+//!
+//! Stream labels live as plain ASCII strings inside compiled `.class`
+//! files in the jar — a quirk of the menu-rendering layer encoding
+//! its TextField positions and content as constant strings the
+//! Java compiler retained in the class string pool. Format observed
+//! in the corpus (Civil War UHD, 2024):
+//!
+//! ```text
+//! LTextField,Audio1,English Dolby Atmos,Fontstrip_Composite,...
+//! RTextField,Audio2,English Descriptive Audio,Fontstrip_Composite,...
+//! HTextField,Subtitle1,English SDH,Fontstrip_Composite,...
+//! ATextField,Subtitle0,None,Fontstrip_Composite,...
+//! ```
+//!
+//! The single uppercase letter before `TextField` is string-pool
+//! prefix noise — the parser anchors on `TextField,` regardless of
+//! what precedes it. `Subtitle0` is the disable-subtitles menu
+//! button and is skipped (not a real subtitle stream).
+//!
+//! ## Implementation
+//!
+//! v2 (2026-05-10): rewritten on top of [`super::class_reader`] —
+//! iterates `CpInfo::Utf8` constant-pool entries instead of raw byte
+//! scanning each class file. Equivalent label coverage (the literal
+//! `TextField,...` strings live in the CP as Utf8 entries), but
+//! structurally cleaner: no false-positive risk from method bytecode
+//! or attribute names happening to contain `TextField,`. Language /
+//! purpose / qualifier classification moved to [`super::vocab`] so all
+//! Java-parser families share one source of truth.
+
+use super::class_reader::CpInfo;
+use super::{ParseResult, StreamLabel, StreamLabelType, jar, vocab};
+use crate::sector::SectorReader;
+use crate::udf::UdfFs;
+use std::collections::BTreeMap;
+
+/// dbp detect can't peek inside a jar without a SectorReader (the
+/// trait function only takes `&UdfFs`), so we trigger on the cheap
+/// signal "any top-level .jar in /BDMV/JAR/." That fires on every
+/// BD-J disc, but parse() does the real `com/dbp/` check and
+/// returns None on a mismatch — so this parser only ever consumes
+/// time on discs that fell through every earlier parser.
+pub fn detect(udf: &UdfFs) -> bool {
+    jar::has_any_top_level_jar(udf)
+}
+
+pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<ParseResult> {
+    jar::for_each_jar(reader, udf, |_entry_name, archive| {
+        if !jar::has_path_prefix(archive, "com/dbp/") {
+            return None;
+        }
+        let labels = scan_jar(archive);
+        if labels.is_empty() {
+            None
+        } else {
+            // High confidence: TextField,Audio1,... is a stable anchor
+            // pattern + vocab routes language/purpose/qualifier.
+            Some(ParseResult::high(labels))
+        }
+    })
+}
+
+fn scan_jar(archive: &mut jar::Jar) -> Vec<StreamLabel> {
+    // BTreeMap so we keep the highest-numbered (last-written) label
+    // for each stream slot deterministic across runs. The same
+    // TextField,Audio1,... string can appear in multiple classes
+    // (button-state variants, localization fallbacks). Last write
+    // wins — they should all agree, but the structure is defensive.
+    let mut audios: BTreeMap<u16, String> = BTreeMap::new();
+    let mut subs: BTreeMap<u16, String> = BTreeMap::new();
+
+    jar::for_each_class(archive, |_class_name, class| {
+        for (_idx, cp) in class.constant_pool.iter() {
+            if let CpInfo::Utf8(s) = cp {
+                collect_textfield(s, &mut audios, &mut subs);
+            }
+        }
+    });
+
+    let mut out = Vec::new();
+    for (num, label) in audios {
+        out.push(make_label(num, label, StreamLabelType::Audio));
+    }
+    for (num, label) in subs {
+        out.push(make_label(num, label, StreamLabelType::Subtitle));
+    }
+    out
+}
+
+fn collect_textfield(
+    s: &str,
+    audios: &mut BTreeMap<u16, String>,
+    subs: &mut BTreeMap<u16, String>,
+) {
+    // Anchor on "TextField," — the prefix character before it varies
+    // (string-pool ordering inside compiled Java) and is irrelevant.
+    let Some(idx) = s.find("TextField,") else {
+        return;
+    };
+    let after = &s[idx + "TextField,".len()..];
+    let mut parts = after.splitn(3, ',');
+    let kind_n = parts.next().unwrap_or("").trim();
+    let label = parts.next().unwrap_or("").trim();
+    if label.is_empty() {
+        return;
+    }
+    if let Some(rest) = kind_n.strip_prefix("Audio") {
+        if let Ok(n) = rest.parse::<u16>() {
+            audios.insert(n, label.to_string());
+        }
+    } else if let Some(rest) = kind_n.strip_prefix("Subtitle") {
+        if let Ok(n) = rest.parse::<u16>() {
+            // Subtitle0 is conventionally the "None / Off" disable
+            // button, not an actual subtitle stream.
+            if n > 0 {
+                subs.insert(n, label.to_string());
+            }
+        }
+    }
+}
+
+fn make_label(num: u16, label: String, stream_type: StreamLabelType) -> StreamLabel {
+    let lang_info = vocab::lang(&label);
+    let language = lang_info.map(|l| l.code).unwrap_or("").to_string();
+    let variant = lang_info.map(|l| l.variant).unwrap_or("").to_string();
+    let qualifier = vocab::qualifier(&label);
+    let purpose = vocab::purpose(&label);
+    StreamLabel {
+        stream_number: num,
+        stream_type,
+        language,
+        name: label,
+        purpose,
+        qualifier,
+        codec_hint: String::new(),
+        variant,
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::super::{LabelPurpose, LabelQualifier};
+    use super::*;
+
+    #[test]
+    fn collect_extracts_audio_and_subtitle_indices() {
+        let mut audios = BTreeMap::new();
+        let mut subs = BTreeMap::new();
+        let lines = [
+            "LTextField,Audio1,English Dolby Atmos,Fontstrip_Composite,296,763,275,25,left",
+            "RTextField,Audio2,English Descriptive Audio,Fontstrip_Composite,296,803,275,25,left",
+            "RTextField,Audio3,Spanish 5.1 Dolby Digital,Fontstrip_Composite,296,843,275,25,left",
+            "ATextField,Subtitle0,None,Fontstrip_Composite,1312,843,275,25,left",
+            "HTextField,Subtitle1,English SDH,Fontstrip_Composite,1312,763,275,25,left",
+            "DTextField,Subtitle2,Spanish,Fontstrip_Composite,1312,803,275,25,left",
+        ];
+        for s in &lines {
+            collect_textfield(s, &mut audios, &mut subs);
+        }
+        assert_eq!(audios.len(), 3);
+        assert_eq!(audios[&1], "English Dolby Atmos");
+        assert_eq!(audios[&2], "English Descriptive Audio");
+        assert_eq!(audios[&3], "Spanish 5.1 Dolby Digital");
+        // Subtitle0 ("None") is skipped — disable button, not a stream.
+        assert_eq!(subs.len(), 2);
+        assert_eq!(subs[&1], "English SDH");
+        assert_eq!(subs[&2], "Spanish");
+    }
+
+    #[test]
+    fn collect_ignores_non_textfield_strings() {
+        let mut audios = BTreeMap::new();
+        let mut subs = BTreeMap::new();
+        for s in [
+            "GraphicButton,SU_Audio",
+            "AudioMenu",
+            "CommentaryMenuAlternateScenes",
+            "PrimaryAudioControl",
+        ] {
+            collect_textfield(s, &mut audios, &mut subs);
+        }
+        assert!(audios.is_empty());
+        assert!(subs.is_empty());
+    }
+
+    #[test]
+    fn make_label_routes_via_vocab() {
+        let l = make_label(1, "English SDH".to_string(), StreamLabelType::Subtitle);
+        assert_eq!(l.language, "eng");
+        assert_eq!(l.qualifier, LabelQualifier::Sdh);
+        assert_eq!(l.purpose, LabelPurpose::Normal);
+    }
+
+    #[test]
+    fn make_label_descriptive_audio() {
+        let l = make_label(
+            2,
+            "English Descriptive Audio".to_string(),
+            StreamLabelType::Audio,
+        );
+        assert_eq!(l.language, "eng");
+        assert_eq!(l.purpose, LabelPurpose::Descriptive);
+    }
+
+    #[test]
+    fn make_label_commentary() {
+        let l = make_label(
+            3,
+            "English Director's Commentary".to_string(),
+            StreamLabelType::Audio,
+        );
+        assert_eq!(l.language, "eng");
+        assert_eq!(l.purpose, LabelPurpose::Commentary);
+    }
+
+    #[test]
+    fn make_label_compound_languages_populate_variant() {
+        let brazilian = make_label(1, "Brazilian Portuguese 5.1".into(), StreamLabelType::Audio);
+        assert_eq!(brazilian.language, "por");
+        assert_eq!(brazilian.variant, "Brazilian");
+
+        let castilian = make_label(1, "Castilian Spanish".into(), StreamLabelType::Audio);
+        assert_eq!(castilian.language, "spa");
+        assert_eq!(castilian.variant, "Castilian");
+
+        let canadian = make_label(
+            1,
+            "Canadian French Dolby Digital".into(),
+            StreamLabelType::Audio,
+        );
+        assert_eq!(canadian.language, "fra");
+        assert_eq!(canadian.variant, "Canadian");
+    }
+
+    #[test]
+    fn make_label_bare_language_has_empty_variant() {
+        let l = make_label(1, "English Dolby Atmos".into(), StreamLabelType::Audio);
+        assert_eq!(l.language, "eng");
+        assert_eq!(l.variant, "");
+    }
+
+    #[test]
+    fn make_label_unknown_language_is_empty() {
+        // vocab::lang returns None — make_label converts both fields to "".
+        let l = make_label(1, "Klingon Dolby Atmos".into(), StreamLabelType::Audio);
+        assert_eq!(l.language, "");
+        assert_eq!(l.variant, "");
+    }
+
+    #[test]
+    fn make_label_rnib_descriptive_service() {
+        let l = make_label(1, "English RNIB".into(), StreamLabelType::Subtitle);
+        assert_eq!(l.language, "eng");
+        assert_eq!(l.qualifier, LabelQualifier::DescriptiveService);
+    }
+}

src/labels/jar.rs

@@ -0,0 +1,151 @@
+//! BD-J jar utilities — common scaffolding for parsers that read
+//! `/BDMV/JAR/*.jar`.
+//!
+//! Composes with [`class_reader`](super::class_reader) for structured
+//! `.class` access. Used by `dbp` (string-pool scan via constant pool)
+//! and `deluxe` (bytecode pattern matching) — those parsers express
+//! "open every top-level jar, look at every .class inside" without
+//! repeating the zip-archive boilerplate.
+
+// `try_each_class` is staged for `labels::deluxe`, which needs the
+// early-return form to short-circuit class iteration on a match.
+// dead-code allow comes off when deluxe lands.
+#![allow(dead_code)]
+
+use super::class_reader::ClassFile;
+use crate::sector::SectorReader;
+use crate::udf::UdfFs;
+use std::io::Cursor;
+use zip::ZipArchive;
+
+/// In-memory zip archive: backed by a `Vec<u8>` read from UDF. Owns
+/// the buffer; callers pass it to [`has_path_prefix`], [`for_each_class`],
+/// etc.
+pub type Jar = ZipArchive<Cursor<Vec<u8>>>;
+
+/// True if `/BDMV/JAR/` contains at least one top-level `.jar` file
+/// (not under a subdir). Used by `detect()` in parsers whose real
+/// signal lives inside a jar — they can't open the jar without a
+/// `SectorReader`, so they use this cheap pre-check and do the real
+/// `com/<vendor>/` discriminator in `parse()`.
+pub fn has_any_top_level_jar(udf: &UdfFs) -> bool {
+    let Some(jar_dir) = udf.find_dir("/BDMV/JAR") else {
+        return false;
+    };
+    jar_dir
+        .entries
+        .iter()
+        .any(|e| !e.is_dir && e.name.to_lowercase().ends_with(".jar"))
+}
+
+/// Open every top-level `*.jar` entry in `/BDMV/JAR/` and yield each
+/// `(entry_name, Jar)` to `f`. Returns the first `Some(R)` the callback
+/// produces, or `None` if every jar was visited without a hit.
+///
+/// "Top-level" means entries directly under `/BDMV/JAR/`, not nested
+/// under a subdir. (Pixelogic, Criterion, Paramount, etc. put their
+/// data files inside `/BDMV/JAR/<x>/`; dbp and Deluxe put their jar
+/// directly at `/BDMV/JAR/<name>.jar`.)
+///
+/// Entries that fail to read from UDF or that aren't valid zips are
+/// silently skipped — same defensive shape as the existing dbp parser.
+pub fn for_each_jar<R, F>(reader: &mut dyn SectorReader, udf: &UdfFs, mut f: F) -> Option<R>
+where
+    F: FnMut(&str, &mut Jar) -> Option<R>,
+{
+    let jar_dir = udf.find_dir("/BDMV/JAR")?;
+    for entry in &jar_dir.entries {
+        if entry.is_dir {
+            continue;
+        }
+        if !entry.name.to_lowercase().ends_with(".jar") {
+            continue;
+        }
+        let path = format!("/BDMV/JAR/{}", entry.name);
+        let Ok(bytes) = udf.read_file(reader, &path) else {
+            continue;
+        };
+        let Ok(mut archive) = ZipArchive::new(Cursor::new(bytes)) else {
+            continue;
+        };
+        if let Some(r) = f(&entry.name, &mut archive) {
+            return Some(r);
+        }
+    }
+    None
+}
+
+/// True if any entry in this jar's central directory starts with
+/// `prefix`. Fast — only reads filenames, never extracts bytes.
+///
+/// Used by parsers as a cheap "is this MY framework's jar?" check
+/// (e.g. `has_path_prefix(archive, "com/dbp/")` for dbp,
+/// `has_path_prefix(archive, "com/bydeluxe/")` for Deluxe).
+pub fn has_path_prefix(archive: &mut Jar, prefix: &str) -> bool {
+    for i in 0..archive.len() {
+        if let Ok(f) = archive.by_index(i) {
+            if f.name().starts_with(prefix) {
+                return true;
+            }
+        }
+    }
+    false
+}
+
+/// Iterate every `.class` entry in the jar, parse it with
+/// [`class_reader`], and call `f` with `(entry_name, &ClassFile)`.
+///
+/// Entries that fail to read or parse are silently skipped — this is
+/// label-extraction code, robustness matters more than completeness.
+/// Callers that need to know which classes failed should use the
+/// lower-level [`class_reader`] API directly.
+pub fn for_each_class<F>(archive: &mut Jar, mut f: F)
+where
+    F: FnMut(&str, &ClassFile),
+{
+    for i in 0..archive.len() {
+        let Ok(mut entry) = archive.by_index(i) else {
+            continue;
+        };
+        if !entry.name().ends_with(".class") {
+            continue;
+        }
+        let name = entry.name().to_string();
+        let mut bytes = Vec::with_capacity(entry.size() as usize);
+        if std::io::Read::read_to_end(&mut entry, &mut bytes).is_err() {
+            continue;
+        }
+        let Ok(class) = ClassFile::parse(&bytes) else {
+            continue;
+        };
+        f(&name, &class);
+    }
+}
+
+/// Like [`for_each_class`] but allows the callback to short-circuit
+/// iteration. Returns the first `Some(R)` the callback produces.
+pub fn try_each_class<R, F>(archive: &mut Jar, mut f: F) -> Option<R>
+where
+    F: FnMut(&str, &ClassFile) -> Option<R>,
+{
+    for i in 0..archive.len() {
+        let Ok(mut entry) = archive.by_index(i) else {
+            continue;
+        };
+        if !entry.name().ends_with(".class") {
+            continue;
+        }
+        let name = entry.name().to_string();
+        let mut bytes = Vec::with_capacity(entry.size() as usize);
+        if std::io::Read::read_to_end(&mut entry, &mut bytes).is_err() {
+            continue;
+        }
+        let Ok(class) = ClassFile::parse(&bytes) else {
+            continue;
+        };
+        if let Some(r) = f(&name, &class) {
+            return Some(r);
+        }
+    }
+    None
+}

src/labels/mod.rs

@@ -7,11 +7,17 @@
 //!   3. Implement `pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<StreamLabel>>`
 //!   4. Add `mod myformat;` below and one line to `PARSERS` array
 
+pub(crate) mod class_reader;
 mod criterion;
 mod ctrm;
+mod dbp;
+mod deluxe;
+pub(crate) mod jar;
 mod paramount;
 mod pixelogic;
+pub(crate) mod text;
 pub mod vocab;
+pub(crate) mod xml;
 
 use crate::disc::{DiscTitle, Stream};
 use crate::sector::SectorReader;
@@ -68,18 +74,70 @@ pub enum LabelQualifier {
 
 // ── Parser registry ────────────────────────────────────────────────────────
 //
-// Each entry: (name, detect_fn, parse_fn)
-// Order = priority. First match wins. Highest quality output first.
+// Each entry: (name, detect_fn, parse_fn). Order = tiebreaker only —
+// the registry picks the highest-confidence parse result, falling back
+// to array order on confidence ties.
 
 type DetectFn = fn(&UdfFs) -> bool;
-type ParseFn = fn(&mut dyn SectorReader, &UdfFs) -> Option<Vec<StreamLabel>>;
+type ParseFn = fn(&mut dyn SectorReader, &UdfFs) -> Option<ParseResult>;
+
+/// Per-parser claim of how reliable its output is. Used by the
+/// registry to pick between parsers when more than one matches (e.g.
+/// a disc that has both `bluray_project.bin` and `playlists.xml`).
+///
+/// A parser SHOULD return `High` only when its full schema was
+/// extracted with no fallback or guessing. `Medium` is for matched-
+/// but-degraded outputs (some streams missing fields, fingerprint
+/// matched but a sub-table couldn't be decoded, etc.). The registry
+/// prefers `High` over `Medium`; ties fall to array order.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)]
+pub enum Confidence {
+    Medium,
+    High,
+}
+
+/// Successful parser result. `None` from `parse()` still means "this
+/// isn't my disc" (no labels at all); `Some(ParseResult { labels, .. })`
+/// with `labels.is_empty()` is also a "no labels" case but reachable
+/// via the analyzer (used by deluxe today to signal "I recognized the
+/// framework but Phase D not yet implemented").
+#[derive(Debug, Clone)]
+pub struct ParseResult {
+    pub labels: Vec<StreamLabel>,
+    pub confidence: Confidence,
+}
+
+impl ParseResult {
+    /// Convenience for the common "I parsed N labels with full schema
+    /// coverage" case.
+    pub fn high(labels: Vec<StreamLabel>) -> Self {
+        ParseResult {
+            labels,
+            confidence: Confidence::High,
+        }
+    }
+
+    /// Convenience for "I matched but had to fall back on some fields".
+    pub fn medium(labels: Vec<StreamLabel>) -> Self {
+        ParseResult {
+            labels,
+            confidence: Confidence::Medium,
+        }
+    }
+}
 
 const PARSERS: &[(&str, DetectFn, ParseFn)] = &[
     ("paramount", paramount::detect, paramount::parse),
     ("criterion", criterion::detect, criterion::parse),
     ("pixelogic", pixelogic::detect, pixelogic::parse),
     ("ctrm", ctrm::detect, ctrm::parse),
-    // ("deluxe",  deluxe::detect,     deluxe::parse),  // TODO: bytecode parser
+    // dbp and deluxe both detect on "any top-level .jar in /BDMV/JAR/"
+    // (every BD-J disc trips that) and do the real vendor-prefix check
+    // in parse(). Order between them is the tiebreaker on equal
+    // confidence; dbp goes first because its parse path is cheaper
+    // (constant-pool iteration vs. deluxe's bytecode walking).
+    ("dbp", dbp::detect, dbp::parse),
+    ("deluxe", deluxe::detect, deluxe::parse),
 ];
 
 /// Search disc for config files, extract labels, apply to streams.
@@ -90,7 +148,18 @@ pub fn apply(reader: &mut dyn SectorReader, udf: &UdfFs, titles: &mut [DiscTitle
     if labels.is_empty() {
         return;
     }
+    apply_labels(&labels, titles);
+}
 
+/// Apply a pre-extracted set of labels to titles' streams. Match
+/// labels to streams by (stream_type, 1-based stream_number per type).
+/// Audio streams update `purpose` + `label` (codec/variant info; never
+/// English purpose text). Subtitle streams update `qualifier` and the
+/// `forced` flag.
+///
+/// Extracted from `apply()` so the matching logic is unit-testable
+/// without needing a SectorReader / UdfFs.
+pub(crate) fn apply_labels(labels: &[StreamLabel], titles: &mut [DiscTitle]) {
     for title in titles.iter_mut() {
         let mut audio_idx: u16 = 0;
         let mut sub_idx: u16 = 0;
@@ -268,14 +337,151 @@ fn generate_audio_label(
 }
 
 fn extract(reader: &mut dyn SectorReader, udf: &UdfFs) -> Vec<StreamLabel> {
-    for (_name, detect, parse) in PARSERS {
-        if detect(udf) {
-            if let Some(labels) = parse(reader, udf) {
-                return labels;
+    let mut best: Option<(&'static str, ParseResult)> = None;
+    for (name, detect, parse) in PARSERS {
+        if !detect(udf) {
+            continue;
+        }
+        tracing::info!(parser = name, "label parser detected");
+        let Some(result) = parse(reader, udf) else {
+            continue;
+        };
+        if result.labels.is_empty() {
+            continue;
+        }
+        // Pick highest confidence. Equal confidence → first wins
+        // (array order tiebreaker).
+        match &best {
+            None => best = Some((name, result)),
+            Some((_, b)) if result.confidence > b.confidence => best = Some((name, result)),
+            _ => {}
+        }
+    }
+    match best {
+        Some((name, r)) => {
+            tracing::info!(
+                parser = name,
+                confidence = ?r.confidence,
+                label_count = r.labels.len(),
+                "label parser selected",
+            );
+            r.labels
+        }
+        None => {
+            tracing::info!("no label parser matched");
+            Vec::new()
+        }
+    }
+}
+
+/// Diagnostic introspection — returns the parser that matched, the
+/// labels it emitted, and the inventory of files under `/BDMV/JAR/*/`
+/// that the discriminators looked at. Intended for `freemkv-tools
+/// labels-analyze` and corpus regression tooling, not production code
+/// paths. The matching/parsing logic is identical to [`extract`]; only
+/// the return shape is richer (includes confidence, all detected
+/// parsers, and any parsers that produced empty results).
+#[doc(hidden)]
+pub fn analyze(reader: &mut dyn SectorReader, udf: &UdfFs) -> LabelAnalysis {
+    let inventory = jar_inventory(udf);
+    let mut parsers_detected: Vec<&'static str> = Vec::new();
+    let mut all_results: Vec<(&'static str, ParseResult)> = Vec::new();
+
+    for (name, detect, parse) in PARSERS {
+        if !detect(udf) {
+            continue;
+        }
+        tracing::info!(parser = name, "label parser detected");
+        parsers_detected.push(name);
+        if let Some(r) = parse(reader, udf) {
+            all_results.push((name, r));
+        }
+    }
+
+    // Selection logic mirrors `extract`: highest confidence + non-empty,
+    // array order tiebreaker.
+    let chosen = all_results
+        .iter()
+        .filter(|(_, r)| !r.labels.is_empty())
+        .max_by(|(_, a), (_, b)| {
+            // Cmp first by confidence (higher first), then position
+            // (earlier first). max_by yields the maximum, so we
+            // invert the index comparison.
+            a.confidence
+                .cmp(&b.confidence)
+                .then(std::cmp::Ordering::Equal)
+        });
+
+    let (parser, confidence, labels) = match chosen {
+        Some((name, r)) => (Some(*name), Some(r.confidence), r.labels.clone()),
+        None => (None, None, Vec::new()),
+    };
+
+    if parsers_detected.is_empty() {
+        tracing::info!("no label parser matched");
+    } else if parser.is_none() {
+        tracing::info!(
+            detected = ?parsers_detected,
+            "label parsers detected but produced no labels"
+        );
+    }
+
+    LabelAnalysis {
+        parser,
+        parsers_detected,
+        confidence,
+        jar_inventory: inventory,
+        labels,
+    }
+}
+
+/// Result of [`analyze`].
+#[doc(hidden)]
+#[derive(Debug, Clone)]
+pub struct LabelAnalysis {
+    /// Which parser was SELECTED — the one whose `ParseResult` had
+    /// the highest confidence among non-empty results (array order
+    /// tiebreaker). `None` means either no parser recognized the
+    /// disc, OR every parser that recognized it returned no labels.
+    /// Use `parsers_detected` to disambiguate.
+    pub parser: Option<&'static str>,
+    /// Confidence of the selected parser, `None` if no parser was
+    /// selected.
+    pub confidence: Option<Confidence>,
+    /// Every parser whose discriminator matched, in registry order.
+    /// Distinguishes "we recognized this disc but couldn't extract
+    /// labels" from "we don't recognize this disc at all" — the
+    /// former points at a parser bug or a truncated capture, the
+    /// latter points at a missing parser.
+    pub parsers_detected: Vec<&'static str>,
+    /// Filenames found under any `/BDMV/JAR/*/` subdirectory, deduped
+    /// and sorted. Helps spot unknown authoring formats when no
+    /// parser detected.
+    pub jar_inventory: Vec<String>,
+    /// Raw labels emitted by the selected parser (empty if `parser`
+    /// is `None`).
+    pub labels: Vec<StreamLabel>,
+}
+
+/// List filenames found under any `/BDMV/JAR/<x>/` subdirectory of
+/// the disc. Deduped, sorted. Returns an empty vec if no JAR dir is
+/// present.
+fn jar_inventory(udf: &UdfFs) -> Vec<String> {
+    let Some(jar_dir) = udf.find_dir("/BDMV/JAR") else {
+        return Vec::new();
+    };
+    let mut out: Vec<String> = Vec::new();
+    for entry in &jar_dir.entries {
+        if entry.is_dir {
+            for child in &entry.entries {
+                if !child.is_dir && !out.contains(&child.name) {
+                    out.push(child.name.clone());
+                }
             }
         }
     }
-    Vec::new()
+    out.sort();
+    out
 }
 
 // ── Shared helpers ─────────────────────────────────────────────────────────
@@ -311,3 +517,414 @@ pub(crate) fn read_jar_file(
     let path = find_jar_file(udf, filename)?;
     udf.read_file(reader, &path).ok().filter(|d| !d.is_empty())
 }
+
+// ── Registry-level tests ────────────────────────────────────────────────────
+
+#[cfg(test)]
+mod registry_tests {
+    use super::*;
+
+    /// Lock the parser roster + order. If someone reorders the array
+    /// or adds/removes a parser, this test forces them to update the
+    /// expectation explicitly. The order is load-bearing: first
+    /// matching `parse()` wins, so reordering changes which parser
+    /// claims a disc on overlapping detect signals.
+    ///
+    /// dbp + deluxe MUST stay at the end (their detect triggers on
+    /// "any BD-J disc"; placing them earlier would short-circuit the
+    /// stricter parsers above them).
+    #[test]
+    fn parsers_registry_order_locked() {
+        let names: Vec<&str> = PARSERS.iter().map(|(n, _, _)| *n).collect();
+        assert_eq!(
+            names,
+            vec![
+                "paramount",
+                "criterion",
+                "pixelogic",
+                "ctrm",
+                "dbp",
+                "deluxe"
+            ],
+            "PARSERS array order changed — confirm dbp + deluxe stay last \
+             (loose detect, real check in parse), and stricter parsers \
+             (paramount/criterion/pixelogic/ctrm — all file-presence \
+             gated detect) stay first."
+        );
+    }
+
+    /// Per-parser sanity: every parser has both detect and parse
+    /// hooked up. Catches accidental nullification (e.g. someone
+    /// stubbing `parse` to always-None during a refactor).
+    #[test]
+    fn parsers_registry_all_entries_populated() {
+        for (name, detect, parse) in PARSERS {
+            // Function pointers can't be Null in safe Rust, so the
+            // assertion is just that the array entry was constructed
+            // — which the iter above already implies. The test
+            // exists to fail compile if someone changes the tuple
+            // shape (e.g. adds a 4th field) without updating callers,
+            // and as a marker for "these parsers exist."
+            let _ = (name, detect, parse);
+        }
+        assert!(!PARSERS.is_empty(), "PARSERS array must not be empty");
+    }
+}
+
+// ── apply() integration tests ──────────────────────────────────────────────
+//
+// End-to-end coverage for the apply_labels + fill_defaults pipeline
+// without needing a SectorReader / UdfFs. Synthetic DiscTitle +
+// StreamLabel inputs, assert on the resulting Stream field values.
+
+#[cfg(test)]
+mod apply_tests {
+    use super::*;
+    use crate::disc::{
+        AudioChannels, AudioStream, Codec, ColorSpace, FrameRate, HdrFormat, Resolution,
+        SampleRate, SubtitleStream, VideoStream,
+    };
+
+    fn audio(pid: u16, codec: Codec, channels: AudioChannels, language: &str) -> Stream {
+        Stream::Audio(AudioStream {
+            pid,
+            codec,
+            channels,
+            language: language.into(),
+            sample_rate: SampleRate::S48,
+            secondary: false,
+            purpose: LabelPurpose::Normal,
+            label: String::new(),
+        })
+    }
+
+    fn subtitle(pid: u16, language: &str) -> Stream {
+        Stream::Subtitle(SubtitleStream {
+            pid,
+            codec: Codec::Pgs,
+            language: language.into(),
+            forced: false,
+            qualifier: LabelQualifier::None,
+            codec_data: None,
+        })
+    }
+
+    fn video() -> Stream {
+        Stream::Video(VideoStream {
+            pid: 0x1011,
+            codec: Codec::Hevc,
+            resolution: Resolution::R2160p,
+            frame_rate: FrameRate::F23_976,
+            hdr: HdrFormat::Hdr10,
+            color_space: ColorSpace::Bt2020,
+            secondary: false,
+            label: String::new(),
+        })
+    }
+
+    fn title_with(streams: Vec<Stream>) -> DiscTitle {
+        DiscTitle {
+            playlist: "00800.mpls".into(),
+            playlist_id: 800,
+            duration_secs: 7200.0,
+            size_bytes: 0,
+            clips: Vec::new(),
+            streams,
+            chapters: Vec::new(),
+            extents: Vec::new(),
+            content_format: crate::disc::ContentFormat::BdTs,
+            codec_privates: Vec::new(),
+        }
+    }
+
+    fn audio_label(num: u16, lang: &str, codec_hint: &str, variant: &str) -> StreamLabel {
+        StreamLabel {
+            stream_number: num,
+            stream_type: StreamLabelType::Audio,
+            language: lang.into(),
+            name: String::new(),
+            purpose: LabelPurpose::Normal,
+            qualifier: LabelQualifier::None,
+            codec_hint: codec_hint.into(),
+            variant: variant.into(),
+        }
+    }
+
+    #[test]
+    fn apply_attaches_codec_hint_and_variant_to_audio() {
+        let mut titles = vec![title_with(vec![
+            video(),
+            audio(0x1100, Codec::TrueHd, AudioChannels::Surround51, "eng"),
+        ])];
+        let labels = vec![audio_label(1, "eng", "Dolby Atmos", "")];
+        apply_labels(&labels, &mut titles);
+
+        if let Stream::Audio(a) = &titles[0].streams[1] {
+            assert_eq!(a.label, "Dolby Atmos");
+        } else {
+            panic!("expected audio stream");
+        }
+    }
+
+    #[test]
+    fn apply_combines_variant_and_codec_hint() {
+        let mut titles = vec![title_with(vec![audio(
+            0x1100,
+            Codec::TrueHd,
+            AudioChannels::Surround51,
+            "por",
+        )])];
+        let labels = vec![audio_label(1, "por", "Dolby Atmos", "Brazilian")];
+        apply_labels(&labels, &mut titles);
+
+        if let Stream::Audio(a) = &titles[0].streams[0] {
+            assert_eq!(a.label, "(Brazilian) Dolby Atmos");
+        } else {
+            panic!("expected audio stream");
+        }
+    }
+
+    #[test]
+    fn apply_sets_purpose_on_audio_commentary() {
+        let mut titles = vec![title_with(vec![audio(
+            0x1100,
+            Codec::Ac3,
+            AudioChannels::Stereo,
+            "eng",
+        )])];
+        let labels = vec![StreamLabel {
+            stream_number: 1,
+            stream_type: StreamLabelType::Audio,
+            language: "eng".into(),
+            name: String::new(),
+            purpose: LabelPurpose::Commentary,
+            qualifier: LabelQualifier::None,
+            codec_hint: String::new(),
+            variant: String::new(),
+        }];
+        apply_labels(&labels, &mut titles);
+
+        if let Stream::Audio(a) = &titles[0].streams[0] {
+            assert_eq!(a.purpose, LabelPurpose::Commentary);
+            // Label stays empty: no codec/variant; purpose is conveyed
+            // structurally, NOT as English text.
+            assert_eq!(a.label, "");
+        } else {
+            panic!("expected audio stream");
+        }
+    }
+
+    #[test]
+    fn apply_uses_name_fallback_only_for_normal_purpose() {
+        // Name fallback fires when purpose=Normal and codec/variant are empty.
+        let mut titles = vec![title_with(vec![audio(
+            0x1100,
+            Codec::TrueHd,
+            AudioChannels::Surround71,
+            "eng",
+        )])];
+        let labels = vec![StreamLabel {
+            stream_number: 1,
+            stream_type: StreamLabelType::Audio,
+            language: "eng".into(),
+            name: "Director's Cut Edition".into(),
+            purpose: LabelPurpose::Normal,
+            qualifier: LabelQualifier::None,
+            codec_hint: String::new(),
+            variant: String::new(),
+        }];
+        apply_labels(&labels, &mut titles);
+
+        if let Stream::Audio(a) = &titles[0].streams[0] {
+            assert_eq!(a.label, "Director's Cut Edition");
+        } else {
+            panic!("expected audio stream");
+        }
+    }
+
+    #[test]
+    fn apply_name_fallback_suppressed_for_non_normal_purpose() {
+        // Name fallback must NOT fire when purpose != Normal — the
+        // CLI is responsible for rendering purpose text.
+        let mut titles = vec![title_with(vec![audio(
+            0x1100,
+            Codec::Ac3,
+            AudioChannels::Stereo,
+            "eng",
+        )])];
+        let labels = vec![StreamLabel {
+            stream_number: 1,
+            stream_type: StreamLabelType::Audio,
+            language: "eng".into(),
+            name: "Commentary by Director".into(),
+            purpose: LabelPurpose::Commentary,
+            qualifier: LabelQualifier::None,
+            codec_hint: String::new(),
+            variant: String::new(),
+        }];
+        apply_labels(&labels, &mut titles);
+        if let Stream::Audio(a) = &titles[0].streams[0] {
+            assert_eq!(a.label, "", "label must not contain English purpose text");
+            assert_eq!(a.purpose, LabelPurpose::Commentary);
+        }
+    }
+
+    #[test]
+    fn apply_sets_qualifier_on_subtitle_sdh() {
+        let mut titles = vec![title_with(vec![subtitle(0x1200, "eng")])];
+        let labels = vec![StreamLabel {
+            stream_number: 1,
+            stream_type: StreamLabelType::Subtitle,
+            language: "eng".into(),
+            name: String::new(),
+            purpose: LabelPurpose::Normal,
+            qualifier: LabelQualifier::Sdh,
+            codec_hint: String::new(),
+            variant: String::new(),
+        }];
+        apply_labels(&labels, &mut titles);
+
+        if let Stream::Subtitle(s) = &titles[0].streams[0] {
+            assert_eq!(s.qualifier, LabelQualifier::Sdh);
+            // SDH doesn't flip the `forced` flag.
+            assert!(!s.forced);
+        } else {
+            panic!("expected subtitle");
+        }
+    }
+
+    #[test]
+    fn apply_flips_forced_flag_on_subtitle_forced_qualifier() {
+        let mut titles = vec![title_with(vec![subtitle(0x1200, "eng")])];
+        let labels = vec![StreamLabel {
+            stream_number: 1,
+            stream_type: StreamLabelType::Subtitle,
+            language: "eng".into(),
+            name: String::new(),
+            purpose: LabelPurpose::Normal,
+            qualifier: LabelQualifier::Forced,
+            codec_hint: String::new(),
+            variant: String::new(),
+        }];
+        apply_labels(&labels, &mut titles);
+        if let Stream::Subtitle(s) = &titles[0].streams[0] {
+            assert_eq!(s.qualifier, LabelQualifier::Forced);
+            assert!(s.forced);
+        }
+    }
+
+    #[test]
+    fn apply_indexes_streams_by_type_separately() {
+        // Audio and subtitle each have their own 1-based index; an
+        // Audio #2 label maps to the 2nd audio stream, not the 2nd
+        // stream overall (which could be a subtitle).
+        let mut titles = vec![title_with(vec![
+            video(),
+            audio(0x1100, Codec::TrueHd, AudioChannels::Surround51, "eng"),
+            subtitle(0x1200, "eng"),
+            audio(0x1101, Codec::Ac3, AudioChannels::Stereo, "fra"),
+        ])];
+        let labels = vec![
+            audio_label(1, "eng", "Dolby Atmos", ""),
+            audio_label(2, "fra", "Dolby Digital", ""),
+            StreamLabel {
+                stream_number: 1,
+                stream_type: StreamLabelType::Subtitle,
+                language: "eng".into(),
+                name: String::new(),
+                purpose: LabelPurpose::Normal,
+                qualifier: LabelQualifier::Sdh,
+                codec_hint: String::new(),
+                variant: String::new(),
+            },
+        ];
+        apply_labels(&labels, &mut titles);
+
+        // Audio #1
+        if let Stream::Audio(a) = &titles[0].streams[1] {
+            assert_eq!(a.label, "Dolby Atmos");
+        }
+        // Audio #2 (4th stream overall)
+        if let Stream::Audio(a) = &titles[0].streams[3] {
+            assert_eq!(a.label, "Dolby Digital");
+        }
+        // Subtitle #1
+        if let Stream::Subtitle(s) = &titles[0].streams[2] {
+            assert_eq!(s.qualifier, LabelQualifier::Sdh);
+        }
+    }
+
+    #[test]
+    fn apply_ignores_labels_for_nonexistent_streams() {
+        // A label for stream #99 with no matching stream is a no-op.
+        let mut titles = vec![title_with(vec![audio(
+            0x1100,
+            Codec::TrueHd,
+            AudioChannels::Surround51,
+            "eng",
+        )])];
+        let labels = vec![audio_label(99, "fra", "Dolby Digital", "")];
+        apply_labels(&labels, &mut titles);
+        if let Stream::Audio(a) = &titles[0].streams[0] {
+            assert_eq!(a.label, "", "label must be untouched");
+        }
+    }
+
+    #[test]
+    fn apply_empty_labels_does_not_touch_streams() {
+        let mut titles = vec![title_with(vec![audio(
+            0x1100,
+            Codec::TrueHd,
+            AudioChannels::Surround51,
+            "eng",
+        )])];
+        apply_labels(&[], &mut titles);
+        if let Stream::Audio(a) = &titles[0].streams[0] {
+            assert_eq!(a.label, "");
+        }
+    }
+
+    // ── fill_defaults() tests ───────────────────────────────────────────────
+
+    #[test]
+    fn fill_defaults_generates_audio_label_when_empty() {
+        let mut titles = vec![title_with(vec![audio(
+            0x1100,
+            Codec::TrueHd,
+            AudioChannels::Surround71,
+            "eng",
+        )])];
+        fill_defaults(&mut titles);
+        if let Stream::Audio(a) = &titles[0].streams[0] {
+            assert_eq!(a.label, "Dolby TrueHD 7.1");
+        }
+    }
+
+    #[test]
+    fn fill_defaults_preserves_existing_audio_label() {
+        let mut titles = vec![title_with(vec![Stream::Audio(AudioStream {
+            pid: 0x1100,
+            codec: Codec::TrueHd,
+            channels: AudioChannels::Surround71,
+            language: "eng".into(),
+            sample_rate: SampleRate::S48,
+            secondary: false,
+            purpose: LabelPurpose::Normal,
+            label: "Pre-set Atmos".into(),
+        })])];
+        fill_defaults(&mut titles);
+        if let Stream::Audio(a) = &titles[0].streams[0] {
+            assert_eq!(a.label, "Pre-set Atmos");
+        }
+    }
+
+    #[test]
+    fn fill_defaults_generates_video_label_with_hdr() {
+        let mut titles = vec![title_with(vec![video()])];
+        fill_defaults(&mut titles);
+        if let Stream::Video(v) = &titles[0].streams[0] {
+            assert!(v.label.contains("4K"), "expected 4K, got {}", v.label);
+            assert!(v.label.contains("HDR10"), "expected HDR10, got {}", v.label);
+        }
+    }
+}

src/labels/paramount.rs

@@ -12,7 +12,7 @@
 //!   sub_com1_idx="23,24,25" />
 //! ```
 
-use super::{LabelPurpose, LabelQualifier, StreamLabel, StreamLabelType};
+use super::{LabelPurpose, LabelQualifier, ParseResult, StreamLabel, StreamLabelType, xml};
 use crate::sector::SectorReader;
 use crate::udf::UdfFs;
 
@@ -20,7 +20,7 @@ pub fn detect(udf: &UdfFs) -> bool {
     super::jar_file_exists(udf, "playlists.xml")
 }
 
-pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<StreamLabel>> {
+pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<ParseResult> {
     let data = super::read_jar_file(reader, udf, "playlists.xml")?;
     let text = std::str::from_utf8(&data).ok()?;
 
@@ -30,8 +30,8 @@ pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<StreamLab
     let mut labels = Vec::new();
 
     // Parse audio streams
-    if let Some(aud) = extract_attr(&feature, "aud") {
-        let com_idx = extract_attr(&feature, "aud_com1_idx").and_then(|s| s.parse::<usize>().ok());
+    if let Some(aud) = xml::attr(&feature, "aud") {
+        let com_idx = xml::attr(&feature, "aud_com1_idx").and_then(|s| s.parse::<usize>().ok());
 
         for (i, lang) in aud.split(',').enumerate() {
             let lang = lang.trim();
@@ -57,12 +57,12 @@ pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<StreamLab
     }
 
     // Parse subtitle streams
-    if let Some(sub) = extract_attr(&feature, "sub") {
-        let forced: Vec<bool> = extract_attr(&feature, "forced_sub")
+    if let Some(sub) = xml::attr(&feature, "sub") {
+        let forced: Vec<bool> = xml::attr(&feature, "forced_sub")
             .map(|s| s.split(',').map(|f| f.trim() == "1").collect())
             .unwrap_or_default();
 
-        let com_indices: Vec<usize> = extract_attr(&feature, "sub_com1_idx")
+        let com_indices: Vec<usize> = xml::attr(&feature, "sub_com1_idx")
             .map(|s| s.split(',').filter_map(|i| i.trim().parse().ok()).collect())
             .unwrap_or_default();
 
@@ -100,32 +100,29 @@ pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<StreamLab
     if labels.is_empty() {
         return None;
     }
-    Some(labels)
+    // High confidence: paramount's playlists.xml is fully structured
+    // and we extract every documented field.
+    Some(ParseResult::high(labels))
 }
 
 /// Find the feature playlist element (the one with the most audio tracks).
-fn find_feature_playlist(xml: &str) -> Option<String> {
+fn find_feature_playlist(text: &str) -> Option<String> {
     let mut best: Option<String> = None;
     let mut best_aud_count = 0;
+    let mut from = 0;
 
-    let mut pos = 0;
-    while let Some(start) = xml[pos..].find("<playlist ") {
-        let abs_start = pos + start;
-        let end = match xml[abs_start..].find("/>") {
-            Some(p) => abs_start + p + 2,
-            None => break,
-        };
-        let element = &xml[abs_start..end];
+    while let Some((start, end)) = xml::find_element(text, "playlist", from) {
+        let element = &text[start..end];
 
-        // Prefer name="Feature" explicitly
-        if let Some(name) = extract_attr(element, "name") {
+        // Prefer name="Feature" explicitly.
+        if let Some(name) = xml::attr(element, "name") {
             if name.eq_ignore_ascii_case("Feature") {
                 return Some(element.to_string());
             }
         }
 
-        // Otherwise pick the one with the most audio streams
-        if let Some(aud) = extract_attr(element, "aud") {
+        // Otherwise pick the one with the most audio streams.
+        if let Some(aud) = xml::attr(element, "aud") {
             let count = aud.split(',').count();
             if count > best_aud_count {
                 best_aud_count = count;
@@ -133,15 +130,7 @@ fn find_feature_playlist(xml: &str) -> Option<String> {
             }
         }
 
-        pos = end;
+        from = end;
     }
     best
 }
-
-/// Extract an XML attribute value from an element string.
-fn extract_attr(element: &str, name: &str) -> Option<String> {
-    let needle = format!("{name}=\"");
-    let start = element.find(&needle)? + needle.len();
-    let end = element[start..].find('"')? + start;
-    Some(element[start..end].to_string())
-}

src/labels/pixelogic.rs

@@ -5,9 +5,13 @@
 //!
 //! Token format: `{lang}_{codec?}_{purpose?}_{region?}_`
 
-use super::{LabelPurpose, LabelQualifier, StreamLabel, StreamLabelType, vocab};
+use super::{
+    Confidence, LabelPurpose, LabelQualifier, ParseResult, StreamLabel, StreamLabelType, text,
+    vocab,
+};
 use crate::sector::SectorReader;
 use crate::udf::UdfFs;
+use std::sync::atomic::{AtomicBool, Ordering};
 
 /// Known audio codec tokens
 const AUDIO_CODECS: &[&str] = &["MLP", "AC3", "DTS", "DDL", "WAV", "AC"];
@@ -20,9 +24,18 @@ pub fn detect(udf: &UdfFs) -> bool {
     super::jar_file_exists(udf, "bluray_project.bin")
 }
 
-pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<StreamLabel>> {
+pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<ParseResult> {
     let data = super::read_jar_file(reader, udf, "bluray_project.bin")?;
-    let strings = extract_strings(&data);
+    // min_len=4 matches the prior local extract_strings impl. The token
+    // grammar is `{lang3}_{codec?}_{purpose?}_{region?}_` so the
+    // shortest meaningful run is 4 chars (lang + underscore).
+    let strings = text::extract_ascii_strings(&data, 4);
+
+    // Tracked across all parse_token calls in this run: did any stream
+    // hit an unrecognized token component (skip-unknown path)? If yes
+    // we downgrade confidence to Medium — the labels are still valid
+    // but the corpus surfaced something we don't catalogue.
+    let saw_unknown = AtomicBool::new(false);
 
     let mut labels = Vec::new();
     let mut in_feature = false;
@@ -50,7 +63,7 @@ pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<StreamLab
             continue;
         }
 
-        if let Some(label) = parse_token(s) {
+        if let Some(label) = parse_token_inner(s, Some(&saw_unknown)) {
             match label.stream_type {
                 StreamLabelType::Audio => {
                     audio_num += 1;
@@ -73,10 +86,15 @@ pub fn parse(reader: &mut dyn SectorReader, udf: &UdfFs) -> Option<Vec<StreamLab
     if labels.is_empty() {
         return None;
     }
-    Some(labels)
+    let confidence = if saw_unknown.load(Ordering::Relaxed) {
+        Confidence::Medium
+    } else {
+        Confidence::High
+    };
+    Some(ParseResult { labels, confidence })
 }
 
-fn parse_token(s: &str) -> Option<StreamLabel> {
+fn parse_token_inner(s: &str, saw_unknown: Option<&AtomicBool>) -> Option<StreamLabel> {
     let clean = s.trim().trim_start_matches('\t').trim_end_matches('_');
     let parts: Vec<&str> = clean.split('_').collect();
     if parts.len() < 2 {
@@ -127,7 +145,18 @@ fn parse_token(s: &str) -> Option<StreamLabel> {
         } else if part.starts_with("PGStream") {
             is_subtitle = true;
         } else {
-            return None;
+            // Unknown token component — skip this single part rather
+            // than discarding the entire stream record. Pre-refactor
+            // behavior was `return None` here, which silently dropped
+            // any stream containing a single uncatalogued token (e.g.
+            // a new codec ID or framework variant). Better to surface
+            // what we know than discard a whole stream over one part,
+            // but flag the parse as Medium-confidence so callers know
+            // some data was elided.
+            tracing::debug!(part = %part, "pixelogic: unrecognized token component, skipping");
+            if let Some(flag) = saw_unknown {
+                flag.store(true, Ordering::Relaxed);
+            }
         }
     }
 
@@ -153,22 +182,72 @@ fn parse_token(s: &str) -> Option<StreamLabel> {
     })
 }
 
-fn extract_strings(data: &[u8]) -> Vec<String> {
-    let mut strings = Vec::new();
-    let mut current = String::new();
+// extract_strings removed — replaced by super::text::extract_ascii_strings(data, 4).
 
-    for &b in data {
-        if (0x20..0x7f).contains(&b) {
-            current.push(b as char);
-        } else {
-            if current.len() > 3 {
-                strings.push(current.clone());
-            }
-            current.clear();
-        }
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn parse_token_basic_audio() {
+        let l = parse_token_inner("eng_MLP_", None).unwrap();
+        assert_eq!(l.stream_type, StreamLabelType::Audio);
+        assert_eq!(l.language, "eng");
+        assert_eq!(l.codec_hint, "TrueHD");
+        assert_eq!(l.purpose, LabelPurpose::Normal);
     }
-    if current.len() > 3 {
-        strings.push(current);
+
+    #[test]
+    fn parse_token_basic_subtitle_sdh() {
+        let l = parse_token_inner("eng_SDH_", None).unwrap();
+        assert_eq!(l.stream_type, StreamLabelType::Subtitle);
+        assert_eq!(l.language, "eng");
+        assert_eq!(l.qualifier, LabelQualifier::Sdh);
+    }
+
+    #[test]
+    fn parse_token_commentary() {
+        let l = parse_token_inner("eng_MLP_ACOM_", None).unwrap();
+        assert_eq!(l.stream_type, StreamLabelType::Audio);
+        assert_eq!(l.purpose, LabelPurpose::Commentary);
+    }
+
+    #[test]
+    fn parse_token_descriptive() {
+        let l = parse_token_inner("eng_AC3_ADES_", None).unwrap();
+        assert_eq!(l.purpose, LabelPurpose::Descriptive);
+    }
+
+    #[test]
+    fn parse_token_with_region() {
+        let l = parse_token_inner("eng_MLP_US_", None).unwrap();
+        assert_eq!(l.language, "eng");
+        assert_eq!(l.variant, "US");
+    }
+
+    #[test]
+    fn parse_token_unknown_component_does_not_kill_stream() {
+        // Regression: pre-refactor, an unrecognized token part returned
+        // None for the whole stream, silently dropping it. New
+        // behavior: skip the unknown part, surface what we know.
+        let l = parse_token_inner("eng_MLP_FUTUREFLAG_FOR_", None).unwrap();
+        assert_eq!(l.stream_type, StreamLabelType::Audio);
+        assert_eq!(l.language, "eng");
+        assert_eq!(l.codec_hint, "TrueHD");
+        assert_eq!(l.qualifier, LabelQualifier::Forced);
+    }
+
+    #[test]
+    fn parse_token_no_audio_or_subtitle_signal_returns_none() {
+        // A token that has only a language and an unknown part with
+        // no audio/subtitle classifier should still return None —
+        // there's no way to file it as a stream.
+        assert!(parse_token_inner("eng_UNKNOWN_", None).is_none());
+    }
+
+    #[test]
+    fn parse_token_rejects_non_lang_prefix() {
+        assert!(parse_token_inner("XX_MLP_", None).is_none());
+        assert!(parse_token_inner("ENG_MLP_", None).is_none()); // uppercase not accepted as ISO 639-2
     }
-    strings
 }

src/labels/text.rs

@@ -0,0 +1,89 @@
+//! Text-extraction helpers used by parsers that scan binary blobs for
+//! embedded label strings.
+//!
+//! Promoted from two near-duplicate implementations:
+//! - `pixelogic::extract_strings` (`bluray_project.bin`, min_len=4)
+//! - `dbp::extract_printable` (`.class` files in jars, min_len=5)
+//!
+//! Single implementation, threshold passed in. Callers that have a
+//! more structured parse path (e.g. `class_reader` for .class) should
+//! prefer that — this helper is for genuinely unstructured input.
+
+/// Walk `data`, emit every maximal run of printable-ASCII bytes
+/// (`0x20..=0x7E`) whose length is at least `min_len`.
+///
+/// Non-printable bytes (including `\t`, `\n`, NUL) terminate the
+/// current run. Output strings are guaranteed valid UTF-8 (they're
+/// pure 7-bit ASCII). Strings shorter than `min_len` are dropped.
+pub fn extract_ascii_strings(data: &[u8], min_len: usize) -> Vec<String> {
+    let mut out = Vec::new();
+    let mut current = String::new();
+    for &b in data {
+        if (0x20..=0x7E).contains(&b) {
+            current.push(b as char);
+        } else if current.len() >= min_len {
+            out.push(std::mem::take(&mut current));
+        } else {
+            current.clear();
+        }
+    }
+    if current.len() >= min_len {
+        out.push(current);
+    }
+    out
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn extracts_simple_runs() {
+        let got = extract_ascii_strings(b"hello\0world\0", 3);
+        assert_eq!(got, vec!["hello", "world"]);
+    }
+
+    #[test]
+    fn applies_minimum_length() {
+        let got = extract_ascii_strings(b"hi\0ok\0longer\0", 5);
+        assert_eq!(got, vec!["longer"]);
+    }
+
+    #[test]
+    fn treats_tab_and_newline_as_separators() {
+        // \t (0x09) and \n (0x0A) are below 0x20, so they break runs.
+        let got = extract_ascii_strings(b"alpha\tbeta\ngamma", 3);
+        assert_eq!(got, vec!["alpha", "beta", "gamma"]);
+    }
+
+    #[test]
+    fn handles_trailing_run_without_terminator() {
+        // Run at the very end of the buffer should still be emitted.
+        let got = extract_ascii_strings(b"prefix\0tail", 3);
+        assert_eq!(got, vec!["prefix", "tail"]);
+    }
+
+    #[test]
+    fn rejects_high_bit_bytes() {
+        // 0x80+ is non-printable per this helper's definition.
+        let mut buf = b"good".to_vec();
+        buf.push(0xC3);
+        buf.push(0xA9);
+        buf.extend_from_slice(b"more");
+        let got = extract_ascii_strings(&buf, 3);
+        assert_eq!(got, vec!["good", "more"]);
+    }
+
+    #[test]
+    fn empty_input_returns_empty() {
+        let got = extract_ascii_strings(&[], 1);
+        assert!(got.is_empty());
+    }
+
+    #[test]
+    fn min_len_zero_emits_singletons() {
+        // Pathological but well-defined.
+        let got = extract_ascii_strings(b"a\0b", 0);
+        assert_eq!(got, vec!["a", "b"]);
+    }
+}

src/labels/vocab.rs

@@ -1,19 +1,40 @@
-//! Shared label vocabulary — values we are 100% confident about.
+//! Shared label vocabulary — canonical mappings used by ≥1 label parser.
 //!
 //! Labels come from BD-J authoring tool files (bluray_project.bin,
-//! playlists.xml, menu_base.prop, etc.) — NOT from BD spec fields.
-//! This is NOT for MPLS/CLPI/STN data. Those follow the BD spec directly.
+//! playlists.xml, menu_base.prop, .class string pools, etc.) — NOT
+//! from BD spec fields. This module is the central, regression-tested
+//! source of truth for:
 //!
-//! Rules:
-//! - Only map values we are 100% certain about (published codec names).
-//! - Unknown codes (csp, eda, cf, etc.) pass through raw from disc.
-//! - The app/CLI handles display text, not the lib.
+//! - Codec brand name aliases (`MLP` → `TrueHD`).
+//! - English / multi-word language name → ISO 639-2 code.
+//! - English text → [`LabelPurpose`] (Commentary / Descriptive / etc.).
+//! - English text → [`LabelQualifier`] (SDH / Forced / Descriptive Service).
+//!
+//! Rules of engagement (carried over from
+//! `(internal)/memory/feedback_label_data_rules.md`):
+//!
+//! 1. Only map values we are 100% certain about — published codec
+//!    names, well-known ISO 639-2 mappings, vendor-documented purpose
+//!    keywords.
+//! 2. Unknown codes / unrecognized phrases pass through raw or return
+//!    `None`. We never guess.
+//! 3. Matching is case-insensitive and word-boundary-aware where
+//!    relevant (so "Commenter" doesn't match "commentary"). Anchoring
+//!    on whole tokens is the responsibility of this module — callers
+//!    pass raw text, we handle it.
+//!
+//! This module is NOT for BD spec STN codec IDs; those decode in
+//! `mpls.rs` separately.
+
+use super::{LabelPurpose, LabelQualifier};
+
+// ── Codec aliases ────────────────────────────────────────────────────────────
 
 /// Map a codec identifier found in label data to its display name.
 ///
 /// These are well-known codec identifiers used across multiple BD-J
-/// authoring tools. NOT BD spec STN codec IDs — those are decoded
-/// separately in mpls.rs.
+/// authoring tools. Unknown codes pass through unchanged so callers
+/// can still surface vendor-specific tokens we haven't catalogued.
 pub fn codec(code: &str) -> &str {
     match code {
         "MLP" => "TrueHD",
@@ -25,3 +46,405 @@ pub fn codec(code: &str) -> &str {
         _ => code,
     }
 }
+
+// ── Language: English / multi-word names → ISO 639-2 ─────────────────────────
+
+/// Result of [`lang`] — ISO code + human-readable regional variant.
+///
+/// `code` is ISO 639-2 (always 3 lowercase letters).
+/// `variant` is the regional dialect as a human-readable English word
+/// (`"Brazilian"`, `"Castilian"`, `"Canadian"`, `"Simplified"`, ...)
+/// or `""` when the input names just a bare language without
+/// dialect ("Spanish" → variant=""). The variant matches the
+/// convention pixelogic / ctrm / criterion already use for their
+/// `StreamLabel::variant` field: a short display token the UI can
+/// surface verbatim.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub struct LangInfo {
+    pub code: &'static str,
+    pub variant: &'static str,
+}
+
+/// Map a free-form language label fragment to an ISO 639-2 code AND
+/// (where applicable) its regional variant.
+///
+/// Handles both bare English names ("English", "Spanish") and the
+/// multi-word vendor variants we've seen in the corpus ("Brazilian
+/// Portuguese", "Castilian Spanish", "Canadian French"). Match is
+/// case-insensitive; longer compound phrases win over their bare
+/// counterparts (so "Brazilian Portuguese" returns
+/// `LangInfo { code: "por", variant: "Brazilian" }`, not consumed by
+/// the bare "Portuguese" entry).
+///
+/// Bare-name matches return `variant: ""`.
+///
+/// Returns `None` for unrecognized input — callers decide whether to
+/// fall back to MPLS spec codes, pass through raw, or drop the stream.
+/// Never guesses.
+///
+/// Why the variant: the prior `lang() -> Option<&str>` shape silently
+/// dropped regional dialect info. "Brazilian Portuguese 5.1" became
+/// `language="por", variant=""` — UI displayed plain "Portuguese"
+/// even though the disc had explicitly labeled this stream Brazilian.
+/// Capturing the variant here parallels how pixelogic and ctrm
+/// populate `StreamLabel::variant` from their own region tables.
+pub fn lang(text: &str) -> Option<LangInfo> {
+    let lower = text.to_lowercase();
+    // Multi-word compounds first — longest-match wins.
+    for (needle, code, variant) in COMPOUND_LANGS {
+        if lower.contains(needle) {
+            return Some(LangInfo { code, variant });
+        }
+    }
+    // Bare names: word-boundary match (avoid "english" inside "englishman"
+    // or any other accidental substring).
+    for (needle, code) in BARE_LANGS {
+        if has_word(&lower, needle) {
+            return Some(LangInfo { code, variant: "" });
+        }
+    }
+    None
+}
+
+const COMPOUND_LANGS: &[(&str, &str, &str)] = &[
+    ("brazilian portuguese", "por", "Brazilian"),
+    ("euro portuguese", "por", "European"),
+    ("european portuguese", "por", "European"),
+    ("castilian spanish", "spa", "Castilian"),
+    ("latin american spanish", "spa", "Latin American"),
+    ("latin spanish", "spa", "Latin American"),
+    ("canadian french", "fra", "Canadian"),
+    ("parisian french", "fra", "Parisian"),
+    ("australian english", "eng", "Australian"),
+    ("austrailian english", "eng", "Australian"), // disc-corpus typo, keep matching
+    ("british english", "eng", "British"),
+    ("simplified chinese", "zho", "Simplified"),
+    ("traditional chinese", "zho", "Traditional"),
+    ("mandarin chinese", "zho", "Mandarin"),
+    ("cantonese chinese", "zho", "Cantonese"),
+];
+
+const BARE_LANGS: &[(&str, &str)] = &[
+    ("english", "eng"),
+    ("spanish", "spa"),
+    ("french", "fra"),
+    ("german", "deu"),
+    ("italian", "ita"),
+    ("japanese", "jpn"),
+    ("chinese", "zho"),
+    ("mandarin", "zho"),
+    ("cantonese", "zho"),
+    ("portuguese", "por"),
+    ("polish", "pol"),
+    ("czech", "ces"),
+    ("hungarian", "hun"),
+    ("dutch", "nld"),
+    ("korean", "kor"),
+    ("arabic", "ara"),
+    ("hindi", "hin"),
+    ("turkish", "tur"),
+    ("thai", "tha"),
+    ("swedish", "swe"),
+    ("norwegian", "nor"),
+    ("danish", "dan"),
+    ("finnish", "fin"),
+    ("hebrew", "heb"),
+    ("russian", "rus"),
+    ("greek", "ell"),
+    ("vietnamese", "vie"),
+    ("indonesian", "ind"),
+    ("malay", "msa"),
+    ("ukrainian", "ukr"),
+    ("romanian", "ron"),
+    ("bulgarian", "bul"),
+    ("croatian", "hrv"),
+    ("serbian", "srp"),
+    ("slovak", "slk"),
+    ("slovenian", "slv"),
+    ("estonian", "est"),
+    ("latvian", "lav"),
+    ("lithuanian", "lit"),
+    ("icelandic", "isl"),
+    ("basque", "eus"),
+    ("catalan", "cat"),
+    ("galician", "glg"),
+];
+
+// ── Purpose ──────────────────────────────────────────────────────────────────
+
+/// Classify a free-form English label string into a [`LabelPurpose`].
+///
+/// Recognized keywords (case-insensitive, word-boundary matched):
+/// - "commentary", "director's commentary" → `Commentary`
+/// - "descriptive", "description", "audio description", "described" → `Descriptive`
+/// - "score", "music only" → `Score`
+/// - "ime" (alternate music for closing themes etc.) → `Ime`
+/// - anything else → `Normal`
+///
+/// Word-boundary matching means "Commentary track" matches but
+/// "Commenter Pro audio" does not.
+pub fn purpose(text: &str) -> LabelPurpose {
+    let lower = text.to_lowercase();
+    // Multi-word compounds first — they're more specific.
+    if lower.contains("audio description") || lower.contains("descriptive service") {
+        return LabelPurpose::Descriptive;
+    }
+    if lower.contains("music only") {
+        return LabelPurpose::Score;
+    }
+    if has_word(&lower, "commentary") {
+        return LabelPurpose::Commentary;
+    }
+    if has_word(&lower, "descriptive")
+        || has_word(&lower, "description")
+        || has_word(&lower, "described")
+    {
+        return LabelPurpose::Descriptive;
+    }
+    if has_word(&lower, "score") {
+        return LabelPurpose::Score;
+    }
+    if has_word(&lower, "ime") {
+        return LabelPurpose::Ime;
+    }
+    LabelPurpose::Normal
+}
+
+// ── Qualifier ────────────────────────────────────────────────────────────────
+
+/// Classify a free-form English label string into a [`LabelQualifier`].
+///
+/// Recognized keywords (case-insensitive, word-boundary matched):
+/// - "sdh", "captions" → `Sdh`
+/// - "forced", "forced narrative" → `Forced`
+/// - "rnib", "descriptive service" → `DescriptiveService`
+/// - anything else → `None`
+///
+/// SDH (Subtitles for the Deaf and Hard of hearing) wins over Forced
+/// when both keywords are present, because an SDH track is its own
+/// stream regardless of whether the player flags it as "forced".
+pub fn qualifier(text: &str) -> LabelQualifier {
+    let lower = text.to_lowercase();
+    if has_word(&lower, "sdh") || has_word(&lower, "captions") {
+        return LabelQualifier::Sdh;
+    }
+    if lower.contains("descriptive service") || has_word(&lower, "rnib") {
+        return LabelQualifier::DescriptiveService;
+    }
+    if has_word(&lower, "forced") {
+        return LabelQualifier::Forced;
+    }
+    LabelQualifier::None
+}
+
+// ── Internal: word-boundary matching ────────────────────────────────────────
+
+/// True if `needle` appears in `haystack` surrounded by non-alphanumeric
+/// boundaries (or string ends). `haystack` is assumed lowercase already.
+///
+/// This is the load-bearing primitive for `lang` / `purpose` /
+/// `qualifier`: bare-token matchers MUST use it, otherwise we match
+/// "english" inside "englishman" and "sdh" inside "lambdash". The
+/// existing parsers used `.contains()` and got lucky on the corpus;
+/// vocab guarantees the boundary.
+fn has_word(haystack: &str, needle: &str) -> bool {
+    if needle.is_empty() {
+        return false;
+    }
+    let bytes = haystack.as_bytes();
+    let nb = needle.as_bytes();
+    let mut i = 0;
+    while i + nb.len() <= bytes.len() {
+        if &bytes[i..i + nb.len()] == nb {
+            let before = if i == 0 { None } else { Some(bytes[i - 1]) };
+            let after = bytes.get(i + nb.len()).copied();
+            let bound = |c: Option<u8>| match c {
+                None => true,
+                Some(b) => !b.is_ascii_alphanumeric(),
+            };
+            if bound(before) && bound(after) {
+                return true;
+            }
+        }
+        i += 1;
+    }
+    false
+}
+
+// ── Tests ────────────────────────────────────────────────────────────────────
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn codec_known_aliases() {
+        assert_eq!(codec("MLP"), "TrueHD");
+        assert_eq!(codec("AC3"), "Dolby Digital");
+        assert_eq!(codec("AC"), "Dolby Digital");
+        assert_eq!(codec("DDL"), "Dolby Digital Plus");
+        assert_eq!(codec("atmos"), "Dolby Atmos");
+    }
+
+    #[test]
+    fn codec_unknown_passes_through() {
+        assert_eq!(codec("FX9"), "FX9");
+        assert_eq!(codec(""), "");
+    }
+
+    fn li(code: &'static str, variant: &'static str) -> LangInfo {
+        LangInfo { code, variant }
+    }
+
+    #[test]
+    fn lang_bare_names_have_empty_variant() {
+        assert_eq!(lang("English"), Some(li("eng", "")));
+        assert_eq!(lang("english"), Some(li("eng", "")));
+        assert_eq!(lang("Spanish 5.1 Dolby Digital"), Some(li("spa", "")));
+        assert_eq!(lang("japanese"), Some(li("jpn", "")));
+        assert_eq!(lang("Italian"), Some(li("ita", "")));
+    }
+
+    #[test]
+    fn lang_compounds_carry_variant() {
+        assert_eq!(
+            lang("Brazilian Portuguese 5.1"),
+            Some(li("por", "Brazilian"))
+        );
+        assert_eq!(lang("Castilian Spanish"), Some(li("spa", "Castilian")));
+        assert_eq!(
+            lang("Canadian French Dolby Digital"),
+            Some(li("fra", "Canadian"))
+        );
+        assert_eq!(
+            lang("Latin American Spanish"),
+            Some(li("spa", "Latin American"))
+        );
+        assert_eq!(lang("Simplified Chinese"), Some(li("zho", "Simplified")));
+        assert_eq!(lang("British English"), Some(li("eng", "British")));
+    }
+
+    #[test]
+    fn lang_compounds_win_over_bare() {
+        // Brazilian Portuguese must map to (por, Brazilian) via the
+        // compound rule, not be intercepted by bare "portuguese"
+        // (which would yield (por, "") and lose the variant).
+        assert_eq!(lang("Brazilian Portuguese").unwrap().variant, "Brazilian");
+        assert_eq!(lang("Canadian French").unwrap().variant, "Canadian");
+    }
+
+    #[test]
+    fn lang_unknown_returns_none() {
+        assert_eq!(lang("Klingon Dolby Atmos"), None);
+        assert_eq!(lang(""), None);
+        assert_eq!(lang("eng"), None); // 3-letter codes are not English names
+    }
+
+    #[test]
+    fn lang_word_boundary_avoids_substring_false_positive() {
+        // No false positive — "engineering" must NOT match "english".
+        assert_eq!(lang("Audio Engineering Demo"), None);
+    }
+
+    #[test]
+    fn purpose_recognizes_commentary() {
+        assert_eq!(purpose("English Commentary"), LabelPurpose::Commentary);
+        assert_eq!(purpose("Director's Commentary"), LabelPurpose::Commentary);
+    }
+
+    #[test]
+    fn purpose_recognizes_descriptive() {
+        assert_eq!(
+            purpose("English Descriptive Audio"),
+            LabelPurpose::Descriptive
+        );
+        assert_eq!(purpose("Audio Description"), LabelPurpose::Descriptive);
+        assert_eq!(purpose("Described Video"), LabelPurpose::Descriptive);
+    }
+
+    #[test]
+    fn purpose_descriptive_service_routes_to_descriptive() {
+        // "Descriptive Service" is qualifier territory but the purpose
+        // implication is Descriptive — vocab::purpose treats it as such.
+        assert_eq!(
+            purpose("English Descriptive Service"),
+            LabelPurpose::Descriptive
+        );
+    }
+
+    #[test]
+    fn purpose_word_boundary_avoids_commenter_false_positive() {
+        // "Commenter Pro audio" does NOT match "commentary" — the
+        // existing dbp/ctrm hand-rolls would have. Vocab is stricter.
+        assert_eq!(purpose("Commenter Pro audio track"), LabelPurpose::Normal);
+    }
+
+    #[test]
+    fn purpose_recognizes_score() {
+        assert_eq!(purpose("Music Only"), LabelPurpose::Score);
+        assert_eq!(purpose("Isolated Score"), LabelPurpose::Score);
+    }
+
+    #[test]
+    fn purpose_unknown_is_normal() {
+        assert_eq!(purpose("English Dolby Atmos"), LabelPurpose::Normal);
+        assert_eq!(purpose(""), LabelPurpose::Normal);
+    }
+
+    #[test]
+    fn qualifier_recognizes_sdh() {
+        assert_eq!(qualifier("English SDH"), LabelQualifier::Sdh);
+        assert_eq!(qualifier("English Captions"), LabelQualifier::Sdh);
+    }
+
+    #[test]
+    fn qualifier_recognizes_forced() {
+        assert_eq!(qualifier("English Forced"), LabelQualifier::Forced);
+        assert_eq!(qualifier("Forced Narrative"), LabelQualifier::Forced);
+    }
+
+    #[test]
+    fn qualifier_recognizes_descriptive_service() {
+        assert_eq!(
+            qualifier("English RNIB"),
+            LabelQualifier::DescriptiveService
+        );
+        assert_eq!(
+            qualifier("English Descriptive Service"),
+            LabelQualifier::DescriptiveService
+        );
+    }
+
+    #[test]
+    fn qualifier_sdh_wins_over_forced_when_both_present() {
+        // SDH track is its own stream regardless of forced flag.
+        assert_eq!(qualifier("English Forced SDH"), LabelQualifier::Sdh);
+    }
+
+    #[test]
+    fn qualifier_unknown_is_none() {
+        assert_eq!(qualifier("English"), LabelQualifier::None);
+        assert_eq!(qualifier(""), LabelQualifier::None);
+    }
+
+    #[test]
+    fn has_word_basic() {
+        assert!(has_word("english forced", "english"));
+        assert!(has_word("english forced", "forced"));
+        assert!(has_word("english", "english"));
+    }
+
+    #[test]
+    fn has_word_rejects_substring() {
+        assert!(!has_word("engineering", "english"));
+        assert!(!has_word("englishman", "english"));
+        assert!(!has_word("aenglish", "english"));
+    }
+
+    #[test]
+    fn has_word_punctuation_boundary() {
+        // "(SDH)" is a valid boundary — parentheses count as non-alphanum.
+        assert!(has_word("english (sdh)", "sdh"));
+        assert!(has_word("commentary,extra,info", "commentary"));
+    }
+}

src/labels/xml.rs

@@ -0,0 +1,452 @@
+//! Tolerant XML scraping helpers — promoted from two near-duplicate
+//! hand-rolls in `paramount.rs` (attribute extraction) and `criterion.rs`
+//! (tag-text extraction).
+//!
+//! These are NOT a full XML parser. They handle the subset of XML the
+//! BD-J authoring tools we've seen actually emit: ASCII tag/attr
+//! names, no entity references inside label strings, optional XML
+//! namespaces. Hardening goals over the prior `find("<tag>")` /
+//! `find(r#"name=""#)` matchers:
+//!
+//! 1. **Case-insensitive** tag and attribute names — vendors casing
+//!    is inconsistent across authoring-tool revisions.
+//! 2. **Namespace-aware** — strip an optional `ns:` prefix so
+//!    `<ns:playlist>` and `<playlist>` both match.
+//! 3. **Whitespace-tolerant** — multiple/tab/newline characters
+//!    around `=` between attribute name and value; whitespace inside
+//!    the opening tag.
+//! 4. **Quote-style tolerant** — both `"value"` and `'value'`.
+//! 5. **Self-closing tag handling** — `<tag />` and `<tag/>` both
+//!    work; [`text`] returns `Some("")` for empty content.
+//!
+//! Out of scope (intentionally simple): XML entity decoding
+//! (`&amp;`, `&lt;`, etc.), CDATA sections, comments, processing
+//! instructions, DTD declarations. None of the BD-J authored disc
+//! data we've observed exercises any of those — labels are plain
+//! ASCII/Latin-1 in attribute values.
+
+/// Extract the value of attribute `name` from one XML element
+/// fragment (e.g. `<playlist name="Feature" id="00222" />`).
+///
+/// Returns the raw attribute text (no entity decoding) or `None` if
+/// the attribute isn't present. Empty string for `name=""` is
+/// represented as `Some("")`.
+pub fn attr(element: &str, name: &str) -> Option<String> {
+    let bytes = element.as_bytes();
+    let name_lower = name.to_ascii_lowercase();
+    let name_bytes = name_lower.as_bytes();
+    let mut i = 0;
+    while i + name_bytes.len() < bytes.len() {
+        // Find the next position where `name=` could start. We need
+        // a word boundary before the name (whitespace or `<` or `:`).
+        if i > 0 && is_name_char(bytes[i - 1]) {
+            i += 1;
+            continue;
+        }
+        if !slice_eq_ignore_case(&bytes[i..i + name_bytes.len()], name_bytes) {
+            i += 1;
+            continue;
+        }
+        let after_name = i + name_bytes.len();
+        // The character immediately after the name must not be a
+        // name-continuation (otherwise we matched a prefix like
+        // `lang_id` when looking for `lang`).
+        if after_name < bytes.len() && is_name_char(bytes[after_name]) {
+            i = after_name;
+            continue;
+        }
+        // Walk past whitespace, then `=`, then more whitespace, then
+        // the opening quote.
+        let mut j = after_name;
+        while j < bytes.len() && is_ws(bytes[j]) {
+            j += 1;
+        }
+        if j >= bytes.len() || bytes[j] != b'=' {
+            i = j.max(i + 1);
+            continue;
+        }
+        j += 1; // past '='
+        while j < bytes.len() && is_ws(bytes[j]) {
+            j += 1;
+        }
+        if j >= bytes.len() {
+            return None;
+        }
+        let quote = bytes[j];
+        if quote != b'"' && quote != b'\'' {
+            // Unquoted attribute values aren't part of well-formed
+            // XML (HTML5 allows them, XML doesn't). Skip.
+            i = j;
+            continue;
+        }
+        let value_start = j + 1;
+        let close = bytes[value_start..].iter().position(|&b| b == quote)?;
+        let value = &element[value_start..value_start + close];
+        return Some(value.to_string());
+    }
+    None
+}
+
+/// Extract the trimmed text content of the first occurrence of
+/// `<tag>...</tag>` in `xml`. Returns `None` if the tag isn't found
+/// or its opening tag is malformed.
+///
+/// Whitespace around the inner text is stripped. Self-closing
+/// `<tag />` yields `Some("")`. Nested same-name tags are NOT
+/// handled — the first close encountered wins (this matches the
+/// prior behavior in criterion.rs).
+pub fn text(xml: &str, tag: &str) -> Option<String> {
+    let (open_end, body_start) = find_open_tag(xml, tag, 0)?;
+    // Self-closing — already consumed in find_open_tag if `/>`.
+    if open_end == body_start {
+        // Means find_open_tag returned the same offset twice for
+        // self-closing form. (Not currently the case in our impl,
+        // but defensive.)
+        return Some(String::new());
+    }
+    // For self-closing tags, body_start is past `/>` and we have no
+    // content. Detect by checking the char at body_start - 1 was `/`.
+    if body_start >= 2 && &xml[body_start - 2..body_start] == "/>" {
+        return Some(String::new());
+    }
+    // Find the matching close tag. Case-insensitive + namespace-aware.
+    let close_start = find_close_tag(xml, tag, body_start)?;
+    Some(xml[body_start..close_start].trim().to_string())
+}
+
+/// Locate the next `<tag>` opening AND its closing `</tag>` in
+/// `xml`, starting at byte offset `from`. Returns `(element_start,
+/// element_end)` — `element_start` is the `<` of the opening tag,
+/// `element_end` is one past the `>` of the closing tag. Useful for
+/// iterating over repeated elements like `<playlist>` blocks in
+/// `paramount`.
+///
+/// Self-closing elements return the same offset for body_end as the
+/// element_end (i.e. `element_end - element_start` includes only the
+/// `<tag .../>` text).
+pub fn find_element(xml: &str, tag: &str, from: usize) -> Option<(usize, usize)> {
+    let bytes = xml.as_bytes();
+    let mut i = from;
+    while i < bytes.len() {
+        if bytes[i] != b'<' {
+            i += 1;
+            continue;
+        }
+        // Try matching tag name at i+1 (after `<`).
+        let after_lt = i + 1;
+        if !matches_tag_name_at(bytes, after_lt, tag) {
+            i += 1;
+            continue;
+        }
+        // Found an open tag at offset i. Walk to find the closing `>`
+        // of the open tag itself.
+        let mut j = after_lt;
+        // Skip past the tag name (and optional namespace prefix).
+        while j < bytes.len() && (is_name_char(bytes[j]) || bytes[j] == b':') {
+            j += 1;
+        }
+        // Walk attributes — track quoting state.
+        let mut self_closing = false;
+        while j < bytes.len() {
+            match bytes[j] {
+                b'>' => {
+                    j += 1;
+                    break;
+                }
+                b'/' if j + 1 < bytes.len() && bytes[j + 1] == b'>' => {
+                    self_closing = true;
+                    j += 2;
+                    break;
+                }
+                b'"' | b'\'' => {
+                    let q = bytes[j];
+                    j += 1;
+                    while j < bytes.len() && bytes[j] != q {
+                        j += 1;
+                    }
+                    if j < bytes.len() {
+                        j += 1;
+                    }
+                }
+                _ => j += 1,
+            }
+        }
+        if self_closing {
+            return Some((i, j));
+        }
+        // Find matching close. Doesn't handle nested same-name; OK
+        // for our authoring-tool subset.
+        let close_start = find_close_tag(xml, tag, j)?;
+        let close_end = find_byte(bytes, b'>', close_start)? + 1;
+        return Some((i, close_end));
+    }
+    None
+}
+
+// ── Internal helpers ───────────────────────────────────────────────────────
+
+/// True if `bytes[start..]` opens a tag named `tag`, allowing an
+/// optional `ns:` namespace prefix. Comparison is case-insensitive.
+/// The character after the tag name must not be a name-continuation
+/// (so `<player>` doesn't match `<play>`).
+fn matches_tag_name_at(bytes: &[u8], start: usize, tag: &str) -> bool {
+    let tag_lower = tag.to_ascii_lowercase();
+    let tag_bytes = tag_lower.as_bytes();
+    // Skip optional `prefix:` (one or more name chars + `:`).
+    let mut name_start = start;
+    let mut scan = start;
+    while scan < bytes.len() && is_name_char(bytes[scan]) {
+        scan += 1;
+    }
+    if scan < bytes.len() && bytes[scan] == b':' {
+        name_start = scan + 1;
+    }
+    if name_start + tag_bytes.len() > bytes.len() {
+        return false;
+    }
+    if !slice_eq_ignore_case(&bytes[name_start..name_start + tag_bytes.len()], tag_bytes) {
+        return false;
+    }
+    // Boundary: char after the tag name must be `>`, `/`, whitespace.
+    let after = name_start + tag_bytes.len();
+    if after >= bytes.len() {
+        return false;
+    }
+    matches!(bytes[after], b'>' | b'/' | b' ' | b'\t' | b'\n' | b'\r')
+}
+
+/// Find the offset of the next `</tag>` (or `</ns:tag>`) in `xml`
+/// starting at `from`. Case-insensitive; returns the offset of the
+/// `<`. None if not found.
+fn find_close_tag(xml: &str, tag: &str, from: usize) -> Option<usize> {
+    let bytes = xml.as_bytes();
+    let mut i = from;
+    while i + 2 < bytes.len() {
+        if bytes[i] == b'<' && bytes[i + 1] == b'/' {
+            // Check tag name (with optional namespace).
+            if matches_tag_name_at(bytes, i + 2, tag) {
+                return Some(i);
+            }
+        }
+        i += 1;
+    }
+    None
+}
+
+/// Find the open tag of `<tag>` in `xml` starting at `from`. Returns
+/// `(after_open_lt, after_open_gt)` — the offsets are: just past
+/// the `<` of the open tag, and just past the `>` of the open tag.
+fn find_open_tag(xml: &str, tag: &str, from: usize) -> Option<(usize, usize)> {
+    let bytes = xml.as_bytes();
+    let (elem_start, _) = find_element(xml, tag, from)?;
+    let after_lt = elem_start + 1;
+    // Find the `>` that ends the open tag (handling quoted attrs).
+    let mut j = after_lt;
+    while j < bytes.len() {
+        match bytes[j] {
+            b'>' => return Some((after_lt, j + 1)),
+            b'/' if j + 1 < bytes.len() && bytes[j + 1] == b'>' => {
+                return Some((after_lt, j + 2));
+            }
+            b'"' | b'\'' => {
+                let q = bytes[j];
+                j += 1;
+                while j < bytes.len() && bytes[j] != q {
+                    j += 1;
+                }
+                if j < bytes.len() {
+                    j += 1;
+                }
+            }
+            _ => j += 1,
+        }
+    }
+    None
+}
+
+fn find_byte(bytes: &[u8], target: u8, from: usize) -> Option<usize> {
+    bytes[from..]
+        .iter()
+        .position(|&b| b == target)
+        .map(|p| p + from)
+}
+
+/// True if `c` can be part of an XML name token (rough). We accept
+/// alphanumerics, `_`, `-`, `.`.
+fn is_name_char(c: u8) -> bool {
+    c.is_ascii_alphanumeric() || c == b'_' || c == b'-' || c == b'.'
+}
+
+fn is_ws(c: u8) -> bool {
+    matches!(c, b' ' | b'\t' | b'\n' | b'\r')
+}
+
+fn slice_eq_ignore_case(a: &[u8], b_lower: &[u8]) -> bool {
+    if a.len() != b_lower.len() {
+        return false;
+    }
+    a.iter()
+        .zip(b_lower.iter())
+        .all(|(&x, &y)| x.to_ascii_lowercase() == y)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn attr_basic() {
+        assert_eq!(
+            attr(r#"<playlist name="Feature" id="00222" />"#, "name"),
+            Some("Feature".into())
+        );
+        assert_eq!(
+            attr(r#"<playlist name="Feature" id="00222" />"#, "id"),
+            Some("00222".into())
+        );
+    }
+
+    #[test]
+    fn attr_case_insensitive_name() {
+        assert_eq!(
+            attr(r#"<playlist Name="Feature" />"#, "name"),
+            Some("Feature".into())
+        );
+        assert_eq!(
+            attr(r#"<playlist NAME="Feature" />"#, "Name"),
+            Some("Feature".into())
+        );
+    }
+
+    #[test]
+    fn attr_accepts_single_quotes() {
+        assert_eq!(
+            attr(r#"<playlist name='Feature' />"#, "name"),
+            Some("Feature".into())
+        );
+    }
+
+    #[test]
+    fn attr_whitespace_around_equals() {
+        assert_eq!(
+            attr(r#"<playlist name = "Feature" />"#, "name"),
+            Some("Feature".into())
+        );
+        assert_eq!(
+            attr("<playlist name\n=\n\"Feature\" />", "name"),
+            Some("Feature".into())
+        );
+    }
+
+    #[test]
+    fn attr_missing_returns_none() {
+        assert_eq!(attr(r#"<playlist name="X" />"#, "id"), None);
+        assert_eq!(attr("", "name"), None);
+    }
+
+    #[test]
+    fn attr_no_substring_false_positive() {
+        // Looking for "lang" should NOT match "lang_id" or
+        // "language" because of the name-char boundary check.
+        assert_eq!(attr(r#"<x lang_id="fra" language="eng" />"#, "lang"), None);
+    }
+
+    #[test]
+    fn attr_empty_value() {
+        assert_eq!(attr(r#"<x name="" id="1" />"#, "name"), Some("".into()));
+    }
+
+    #[test]
+    fn text_basic() {
+        assert_eq!(text("<x>hello</x>", "x"), Some("hello".into()));
+        assert_eq!(
+            text("<x>  hello world  </x>", "x"),
+            Some("hello world".into())
+        );
+    }
+
+    #[test]
+    fn text_case_insensitive_tag() {
+        assert_eq!(text("<X>foo</X>", "x"), Some("foo".into()));
+        assert_eq!(text("<Foo>bar</foo>", "foo"), Some("bar".into()));
+    }
+
+    #[test]
+    fn text_namespace_prefix() {
+        assert_eq!(text("<ns:tag>value</ns:tag>", "tag"), Some("value".into()));
+        assert_eq!(text("<foo:Bar>v</foo:Bar>", "bar"), Some("v".into()));
+    }
+
+    #[test]
+    fn text_self_closing() {
+        assert_eq!(text("<x/>", "x"), Some("".into()));
+        assert_eq!(text("<x />", "x"), Some("".into()));
+        assert_eq!(text("<x  attr=\"y\" />", "x"), Some("".into()));
+    }
+
+    #[test]
+    fn text_with_attrs() {
+        assert_eq!(
+            text(r#"<x id="1" name="y">hello</x>"#, "x"),
+            Some("hello".into())
+        );
+    }
+
+    #[test]
+    fn text_missing_close_returns_none() {
+        assert_eq!(text("<x>hello", "x"), None);
+    }
+
+    #[test]
+    fn text_skips_inner_tags_naively() {
+        // Limitation noted: nested same-name tags aren't handled.
+        // Different-name nesting works (we just return everything
+        // between the open and close).
+        assert_eq!(
+            text("<x><y>nested</y></x>", "x"),
+            Some("<y>nested</y>".into())
+        );
+    }
+
+    #[test]
+    fn find_element_basic() {
+        let xml = r#"<x />  <y attr="1">body</y>"#;
+        let (s, e) = find_element(xml, "y", 0).unwrap();
+        assert_eq!(&xml[s..e], r#"<y attr="1">body</y>"#);
+    }
+
+    #[test]
+    fn find_element_self_closing() {
+        let xml = r#"<x />"#;
+        let (s, e) = find_element(xml, "x", 0).unwrap();
+        assert_eq!(&xml[s..e], "<x />");
+    }
+
+    #[test]
+    fn find_element_handles_quoted_gt_in_attr() {
+        // A `>` inside a quoted attribute value should not terminate
+        // the open tag prematurely.
+        let xml = r#"<x attr="foo>bar">body</x>"#;
+        let (s, e) = find_element(xml, "x", 0).unwrap();
+        assert_eq!(&xml[s..e], r#"<x attr="foo>bar">body</x>"#);
+    }
+
+    #[test]
+    fn find_element_iteration() {
+        let xml = "<p>a</p><p>b</p><p>c</p>";
+        let mut positions = Vec::new();
+        let mut from = 0;
+        while let Some((s, e)) = find_element(xml, "p", from) {
+            positions.push(&xml[s..e]);
+            from = e;
+        }
+        assert_eq!(positions, vec!["<p>a</p>", "<p>b</p>", "<p>c</p>"]);
+    }
+
+    #[test]
+    fn find_element_with_namespace() {
+        let xml = r#"<root><ns:item id="1" /></root>"#;
+        let (s, e) = find_element(xml, "item", 0).unwrap();
+        assert_eq!(&xml[s..e], r#"<ns:item id="1" />"#);
+    }
+}

src/lib.rs

@@ -84,7 +84,7 @@ pub(crate) mod identity;
 pub(crate) mod ifo;
 pub(crate) mod io;
 pub mod keydb;
-pub(crate) mod labels;
+pub mod labels;
 pub(crate) mod mpls;
 pub mod mux;
 pub mod pes;

src/mux/disc.rs

@@ -657,11 +657,8 @@ mod tests {
         // produces no frames, but the call still routes through the blanket
         // dispatch into Stream::read.
         let mut frames = 0usize;
-        loop {
-            match src.read().expect("read") {
-                Some(_) => frames += 1,
-                None => break,
-            }
+        while src.read().expect("read").is_some() {
+            frames += 1;
             if frames > 1024 {
                 panic!("unexpected unbounded frame stream from empty title");
             }

tests/integration_progress_and_halt.rs

@@ -459,11 +459,17 @@ fn test_disc_copy_completes_full_disc_with_failing_reader() {
     let _ = std::fs::remove_file(&iso_path);
     let _ = std::fs::remove_file(libfreemkv::disc::mapfile_path_for(&iso_path));
 
-    // Hard bound — even at 0 ms per read, 1024 sectors with skip-forward
-    // should complete in well under a second on any host.
+    // Hard bound — Pass 1 must NOT infinite-loop on a fully-failing
+    // reader. The threshold accommodates the 2026-05-10 wedge-
+    // avoidance pause (PASS_1_FAIL_PAUSE_SECS = 5 s on each failed
+    // batch). With batch=32 and 1024 sectors that's up to ~5 batch
+    // failures + a few damage-jump pauses before fast-trigger jumps
+    // us past end-of-disc — well-bounded total, ~20-30 s typical.
+    // The point of this test is "finishes cleanly, not infinitely",
+    // not "completes in milliseconds."
     assert!(
-        elapsed < Duration::from_secs(5),
-        "Pass 1 took {elapsed:?} on a 2 MB synthetic disc — expected < 5 s"
+        elapsed < Duration::from_secs(60),
+        "Pass 1 took {elapsed:?} on a 2 MB synthetic disc — expected < 60 s (not infinite)"
     );
 
     // Per RIP_DESIGN.md §2.1: Pass 1 must reach end of disc regardless of
@@ -624,19 +630,27 @@ fn test_disc_copy_marks_failed_ecc_blocks_as_nontrimmed() {
     let _ = std::fs::remove_file(&iso_path);
     let _ = std::fs::remove_file(libfreemkv::disc::mapfile_path_for(&iso_path));
 
-    // Pass 1 reads every batch at bpt=32 (no batch reduction, no skip-ahead).
-    // BlockSizeFailingReader fails on multi-sector reads but succeeds on single-sector.
-    // Bridge degradation handling retries failed batches as individual sectors,
-    // so all data is recovered as Finished. bytes_good == total_bytes is correct.
+    // Pass 1's job is "fast and accurate, get the most data in the
+    // shortest time." It no longer bisects on marginal media — that's
+    // Pass N's purpose-built role. So a BlockSizeFailingReader that
+    // fails on multi-sector reads and succeeds on single-sector
+    // results in: every batch fails → SkipBlock → whole 32-sector
+    // ECC block marked NonTrimmed → Pass N (Disc::patch) revisits and
+    // recovers via single-sector reads with proper recovery semantics.
+    //
+    // Pass 1 alone:
     assert_eq!(
-        result.bytes_good, total_bytes,
-        "Pass 1 at bpt=32 recovers all sectors via single-sector retry on MEDIUM_ERROR"
+        result.bytes_good, 0,
+        "Pass 1 doesn't bisect on marginal media — failed batches become NonTrimmed for Pass N to revisit"
     );
     assert_eq!(
-        result.bytes_pending, 0,
-        "no pending sectors after full recovery"
+        result.bytes_pending, total_bytes,
+        "every sector is NonTrimmed (pending) after Pass 1, awaiting Pass N"
+    );
+    assert!(
+        !result.complete,
+        "complete=false because NonTrimmed regions remain (Pass N's work)"
     );
-    assert!(result.complete, "complete=true when all sectors recovered");
 }
 
 // ── 9. PassProgress carries separate unreadable vs pending byte counts ─────

tests/pass_n_size_aware_skip.rs

@@ -30,8 +30,10 @@ struct PatternedSectorReader {
     trace: Arc<Mutex<Vec<(u32, u16)>>>,
 }
 
+type ReadTrace = Arc<Mutex<Vec<(u32, u16)>>>;
+
 impl PatternedSectorReader {
-    fn new(capacity: u32, bad_lbas: HashSet<u32>) -> (Self, Arc<Mutex<Vec<(u32, u16)>>>) {
+    fn new(capacity: u32, bad_lbas: HashSet<u32>) -> (Self, ReadTrace) {
         let trace = Arc::new(Mutex::new(Vec::new()));
         (
             Self {