Decode the tie class header's +0x0c word: three LOD record counts

Range measured: every tie class on the disc, 1,864 classes over all 27 levels.

Method: read the word at class +0x0c as four bytes, and read the record counts
independently from the blocks themselves -- follow packets[lod] to the block,
walk to the 0xf6 terminator, and count the 16-byte records the block header's
data_offset names. Two paths to the same number, so the check is not circular.

Result: the byte at +0x0c is the LOD 0 record count, +0x0d is LOD 1's and +0x0e
is LOD 2's, on 1,864 of 1,864 classes. Classes with an absent LOD agree too:
a zero packets[lod] gives a zero count and the byte is zero.

This closes the file's oldest lie about the tie header. The comment in
TieClassHeader::parse read "0x0c: the 64 light-slot normals offset. Read past,
not decoded" -- inherited from R&C1's layout, where the word genuinely is an
offset. On R&C2 it is not an offset at all; it is three counts, and R&C1 carries
the same quadruple at +0x20. Skipping the field was the right action on the
wrong information: the field was unknown only because R&C1's map was applied to
R&C2's bytes.

TieClassHeader gains packet_counts, and
tie_class_header_counts_records_per_lod asserts the agreement.

The fourth byte, +0x0f, is left unnamed. It ranges 1..15, and the obvious
following of R&C1 -- the shader/ad-gif total -- disagrees on 1,827 of 1,864
classes, which is 2% agreement where a count would expect 100%. It is recorded
as unidentified rather than given a name that fits a fiftieth of the population.

crates/gc-extract/tests/disc.rs

@@ -2572,6 +2572,122 @@ fn tie_records_are_packet_headers() {
     );
 }
 
+/// The tie class header's `+0x0c` word is the three LOD record counts.
+///
+/// The word was read as R&C1's 64-light-slot *normals offset* and skipped for
+/// two commits, which was the right thing to do with the wrong information: the
+/// byte at `+0x0c` is not an offset at all. It is the **record count of the LOD
+/// 0 block**, `+0x0d` of LOD 1 and `+0x0e` of LOD 2 -- verified on **1,864 of
+/// 1,864** classes across all 27 levels. R&C1 carries the same quadruple at
+/// `+0x20`; R&C2 has it here.
+///
+/// The check is not circular: the counts are read from the header, and
+/// compared against the record tables the blocks themselves carry, which are
+/// reached through `packets[]` and delimited by the terminator. Two independent
+/// paths to the same number.
+///
+/// The fourth byte, `+0x0f`, is left unnamed. It is *not* the shader (ad-gif)
+/// total, which disagrees on 1,827 of the classes measured here.
+#[test]
+fn tie_class_header_counts_records_per_lod() {
+    require_disc!();
+    let data = std::env::var_os("GC_EXTRACTED")
+        .map(PathBuf::from)
+        .unwrap_or_else(|| {
+            let home = std::env::var("HOME").unwrap_or_else(|_| ".".into());
+            PathBuf::from(home).join("Documents/gc-extracted")
+        });
+    if !data.exists() {
+        eprintln!("skipped: unpack the disc first");
+        return;
+    }
+
+    use gc_formats::level_core::{ClassEntry, LevelCoreHeader};
+    use gc_formats::tie::{parse_tie_block, TieClassHeader, TieRecord};
+
+    let mut levels = 0usize;
+    let mut classes = 0usize;
+    let mut byte_matches = [0usize; 3];
+    let mut per_class_all_three = 0usize;
+    let mut fourth_bad = 0usize;
+    let mut fourth_range = (u8::MAX, 0u8);
+
+    for lv in 0..27usize {
+        let (Ok(idx), Ok(core)) = (
+            std::fs::read(data.join(format!("level{lv}/core_index.bin"))),
+            std::fs::read(data.join(format!("level{lv}/core_data.bin"))),
+        ) else {
+            continue;
+        };
+        let t = LevelCoreHeader::parse(&idx).expect("hdr").tie_classes;
+        levels += 1;
+
+        for c in 0..t.count as usize {
+            let o = t.offset as usize + c * 0x20;
+            let e = ClassEntry::parse(&idx[o..o + 0x20]).expect("entry");
+            let base = e.offset_in_asset_wad as usize;
+            if base == 0 || base + 0x30 > core.len() {
+                continue;
+            }
+            let Ok(hdr) = TieClassHeader::parse(&core[base..base + 0x48]) else {
+                continue;
+            };
+            // The record tables, from the blocks themselves.
+            let mut counts = [0usize; 3];
+            let mut shader_total = 0usize;
+            for (lod, count) in counts.iter_mut().enumerate() {
+                let off = hdr.packets[lod];
+                if off <= 0 {
+                    continue;
+                }
+                let s = base + off as usize;
+                if let Some(blk) = parse_tie_block(&core, s) {
+                    let recs: Vec<&[u8]> = blk.header.records(&core[s..]).collect();
+                    *count = recs.len();
+                    for rec in recs {
+                        if let Ok(r) = TieRecord::parse(rec) {
+                            shader_total += r.shader_count as usize;
+                        }
+                    }
+                }
+            }
+
+            classes += 1;
+            for (lod, slot) in byte_matches.iter_mut().enumerate() {
+                *slot += usize::from(hdr.packet_counts[lod] as usize == counts[lod]);
+            }
+            per_class_all_three +=
+                usize::from((0..3).all(|l| hdr.packet_counts[l] as usize == counts[l]));
+            if hdr.unknown_0f as usize != shader_total {
+                fourth_bad += 1;
+            }
+            fourth_range.0 = fourth_range.0.min(hdr.unknown_0f);
+            fourth_range.1 = fourth_range.1.max(hdr.unknown_0f);
+        }
+    }
+
+    assert!(levels >= 25, "only {levels} levels were read");
+    assert_eq!(classes, 1_864, "expected 1,864 tie classes, got {classes}");
+    for (lod, &matched) in byte_matches.iter().enumerate() {
+        assert_eq!(
+            matched,
+            classes,
+            "LOD{lod} record count byte disagrees on {} of {classes}",
+            classes - matched
+        );
+    }
+    assert_eq!(
+        per_class_all_three, classes,
+        "all three bytes match on {} of {classes}",
+        per_class_all_three
+    );
+    eprintln!(
+        "tie class header +0x0c: LOD0/LOD1/LOD2 record counts on {classes} classes; \
+         +0x0f range {}..{}, not the shader total on {}",
+        fourth_range.0, fourth_range.1, fourth_bad
+    );
+}
+
 /// Byte positions that hold the same value in every supplied record.
 fn constant_positions(records: &[&[u8]]) -> usize {
     let Some(first) = records.first() else {

crates/gc-formats/src/tie.rs

@@ -54,6 +54,21 @@ pub struct TieClassHeader {
     /// They are strictly increasing on only 461 of 1,864 classes, so neither one
     /// bounds the other -- the `0xf6` terminator does that instead.
     pub packets: [i32; 3],
+    /// `0x0c`: the record table's length of each LOD, as three `u8` counts
+    /// followed by one unidentified byte.
+    ///
+    /// The word was read as R&C1's 64-light-slot *normals offset* and skipped,
+    /// which was the right thing to do with the wrong information: the byte at
+    /// `+0x0c` is not an offset at all. It is the **record count of the LOD 0
+    /// block**, `+0x0d` of LOD 1, `+0x0e` of LOD 2, verified on **1,864 of
+    /// 1,864** classes on all 27 levels. R&C1 carries the same quadruple at
+    /// `+0x20`; R&C2 has it here.
+    ///
+    /// `+0x0f` is not the shader (ad-gif) total, which disagrees on 1,827 of
+    /// the classes measured; it is left unnamed rather than guessed at.
+    pub packet_counts: [u8; 3],
+    /// `0x0f`: unidentified. Not the shader total.
+    pub unknown_0f: u8,
     /// `0x10`, `0x14`, `0x18`: LOD switch distances in world units.
     pub near_dist: f32,
     pub mid_dist: f32,
@@ -73,8 +88,13 @@ impl TieClassHeader {
         for slot in packets.iter_mut() {
             *slot = b.i32("tie packet offset")?;
         }
-        // 0x0c: the 64 light-slot normals offset. Read past, not decoded.
+        // 0x0c: the three LOD record counts and one unidentified byte.
-        b.skip(4, "tie normals offset")?;
+        let packet_counts = [
+            b.u8("tie LOD0 record count")?,
+            b.u8("tie LOD1 record count")?,
+            b.u8("tie LOD2 record count")?,
+        ];
+        let unknown_0f = b.u8("tie class word 0x0f")?;
         let near_dist = b.f32("tie near distance")?;
         let mid_dist = b.f32("tie mid distance")?;
         let far_dist = b.f32("tie far distance")?;
@@ -85,6 +105,8 @@ impl TieClassHeader {
         let o_class = b.i32("tie class number")?;
         Ok(TieClassHeader {
             packets,
+            packet_counts,
+            unknown_0f,
             near_dist,
             mid_dist,
             far_dist,

docs/formats/core-data-regions.md

@@ -1722,6 +1722,39 @@ are verbatim, the grid is known, and three independent segmentation theories hav
 failed on the whole population. Further probing without the loader's own logic is
 not going to converge.
 
+## Tie class header: `+0x0c` is the three LOD record counts (2026-10-11)
+
+**Range measured:** every tie class on the disc -- 1,864 classes, all 27 levels.
+
+**Method:** read the word at class `+0x0c` as four bytes, and read the record
+counts independently from the blocks themselves: follow `packets[lod]` to the
+block, walk to the `0xf6` terminator, and count the 16-byte records the block
+header's `data_offset` names. Two paths to the same number, so the check is not
+circular.
+
+**Result: the byte at `+0x0c` is the LOD 0 record count, `+0x0d` is LOD 1's and
+`+0x0e` is LOD 2's, on 1,864 of 1,864 classes.** On every level the three
+counts agree, including the classes where a LOD is absent -- a zero
+`packets[lod]` gives a zero count, matching the byte.
+
+This closes the file's oldest lie about the tie header: the comment in
+`TieClassHeader::parse` said "0x0c: the 64 light-slot normals offset. Read past,
+not decoded", inherited from R&C1's layout where the word *is* an offset. Here
+it is not an offset at all -- it is three counts -- and R&C1 carries the same
+quadruple at `+0x20`. The earlier note was the right action on the wrong
+information: skipping an unknown field is correct, and the field was only
+unknown because R&C1's map was applied to R&C2's bytes.
+
+`TieClassHeader::packet_counts` carries the three, and
+`tie_class_header_counts_records_per_lod` asserts the agreement.
+
+**The fourth byte, `+0x0f`, is unidentified.** It ranges 1..15 across the disc.
+The obvious reading, following R&C1's `packet_count[3]` being `texture_count`,
+is a material count, and it is **not** the shader total: the sum of
+`shader_count` over a class's records disagrees on 1,827 of 1,864 classes.
+Left unnamed rather than guessed at -- a count that matches on 2% of the
+population is not a count.
+
 ## Tie: `i16[3]` selects the texcoord encoding, which explains the residual
 
 The texcoords are 12-bit fixed on 82.5% of vertices, not 100%. Grouped by the

docs/roadmap.md

@@ -275,6 +275,14 @@ Phase 1 hands over the extents; this phase works out what each one contains.
       V4_8 parent arrays, weight running zero to one across each band); verifying
       them needs the **camera distance at the savestate** so the weight can be
       computed and compared against the rendered result.
+- [x] The tie class header's **`+0x0c` word is the three LOD record
+      counts** — LOD 0 at `+0x0c`, LOD 1 at `+0x0d`, LOD 2 at `+0x0e` — verified
+      on **1,864 of 1,864** classes by reading the counts independently from
+      the blocks' own record tables. This was filed as "R&C1's normals offset,
+      read past, not decoded", which was the right action on the wrong map:
+      R&C1 carries the same quadruple at `+0x20` and it is not an offset on
+      R&C2. The fourth byte is unidentified and is *not* the shader total
+      (disagrees on 1,827 of 1,864).
 - [x] `tie` — **fully chained.** `resolve_tie_class` walks both indirections: the
       class table is in `core_index.bin`, and the entry's `offset_in_asset_wad`
       addresses `core_data.bin` (not `core_data.wad`). All **1,864** tie classes