feat(accuracysnes): emulation mode uses its own vector table (A6.02)

CHANGELOG.md

@@ -11,6 +11,21 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Added
 
+- **Emulation mode uses its own vector table (`A6.02`).** `COP` goes through `$FFF4`, sixteen bytes
+  from the native `$FFE4`, and a core that keeps one set of vectors — or picks the table from
+  something other than the E flag — lands in the wrong handler. Nothing about that is visible in
+  ordinary code, since a game's `COP` handler is usually the same routine either way.
+
+  **The cart could not see it either.** Both vectors pointed at the same trampoline, so a core taking
+  the native table in emulation mode ran the same handler and passed. The runtime now gives the
+  emulation `COP` vector its own trampoline and its own RAM pointer, which is what makes the two
+  distinguishable at all. Both handlers are installed and live, so taking the wrong table is a wrong
+  answer rather than a hang, and the failure code says which vector was used.
+
+  `BRK` deliberately does not get the same treatment: in emulation, `$FFFE` is shared between `IRQ`
+  and `BRK`, so a pointer behind it could not mean "the BRK handler" unambiguously, and splitting it
+  would invent a distinction the machine does not have.
+
 - **`(dp,X)`'s pointer fetch wraps inside bank `$00`, and `N`/`Z` are valid in decimal (`A2.09`,
   `A7.04`).** The pointer fetch is a separate piece of code in most cores and is the one place the
   `d,X` no-bank-crossing rule is easy to forget, because it happens before the mode looks like an

docs/accuracysnes-coverage.md

@@ -13,7 +13,7 @@ Every sub-group of Part V is enumerated, so this is a **complete** statement of
 | `A3` | 10 | 8 | 0 | A3.06, A3.08 |
 | `A4` | 10 | 5 | 0 | A4.03, A4.04, A4.05, A4.09, A4.10 |
 | `A5` | 15 | 9 | 0 | A5.09, A5.10, A5.16, A5.18, A5.19, A5.20 |
-| `A6` | 15 | 8 | 0 | A6.02, A6.09, A6.10, A6.11, A6.12, A6.13, A6.15 |
+| `A6` | 15 | 9 | 0 | A6.09, A6.10, A6.11, A6.12, A6.13, A6.15 |
 | `A7` | 5 | 5 | 0 | — |
 | `A8` | 6 | 2 | 0 | A8.01, A8.04, A8.05, A8.06 |
 | `A9` | 3 | 3 | 0 | — |
@@ -52,7 +52,7 @@ Every sub-group of Part V is enumerated, so this is a **complete** statement of
 | `F1` | 22 | 1 | 0 | F1.01, F1.03, F1.04, F1.05, F1.06, F1.07, F1.08, F1.09, F1.10, F1.11, F1.12, F1.13, F1.14, F1.15, F1.16, F1.17, F1.18, F1.19, F1.20, F1.21, F1.22 |
 | `G1` | 18 | 8 | 0 | G1.01, G1.03, G1.05, G1.06, G1.07, G1.13, G1.15, G1.16, G1.17, G1.18 |
 
-**189 of 443** enumerated assertion rows covered by an on-cart test, plus **50** covered only by a rendered scene (`docs/adr/0013`) — **239 of 443** in total.
+**190 of 443** enumerated assertion rows covered by an on-cart test, plus **50** covered only by a rendered scene (`docs/adr/0013`) — **240 of 443** in total.
 
 The two columns are kept apart on purpose. An on-cart result means the same thing on any emulator and on real hardware; a rendered scene needs a host holding the golden. Adding them into one figure would quietly change what the number claims.
 

docs/accuracysnes-plan.md

@@ -13,7 +13,7 @@ AccuracySNES closed ticket **T-04**. The follow-on tickets minted here are **T-0
 
 | | |
 |---|---|
-| Tests | **233** (221 scoring + 11 golden vectors + 1 region SKIP per image) |
+| Tests | **234** (222 scoring + 11 golden vectors + 1 region SKIP per image) — *tests, not assertions; see the note below the table* |
 | Rendered scenes | **50**, all cross-validated (`docs/adr/0013`) |
 | Pass rate | **100.00%**, floor enforced at 1.00 by `tests/accuracysnes.rs` |
 | Cross-validated | RustySNES and Mesen2 agree on every test; snes9x agrees on every test but five, all recorded reference bugs with citations in `scripts/accuracysnes/crossval.sh`. Both images. |
@@ -23,6 +23,12 @@ AccuracySNES closed ticket **T-04**. The follow-on tickets minted here are **T-0
 These counts are maintained by hand and will drift. **`docs/accuracysnes-coverage.md` is the
 authority**: it is regenerated with the ROM, so it cannot.
 
+**The test count and the assertion count are different numbers and neither bounds the other.** One
+test routinely carries several assertions with distinct failure codes, and several tests routinely
+share one enumerated assertion (`E6.02` is four tests for one row). Reading the coverage figure off
+the test count, or the reverse, has now been done once by a human and once by a review bot in the
+same fortnight, so it is written here as well as in §2.
+
 Phase A shipped Group A. Phase B has so far shipped the register-observable half of Group C — the
 OAM/VRAM/CGRAM port mechanics, the H/V counters, the two open-bus latches, the version nibbles, the
 Mode 7 multiply, the sprite over-flags, the VRAM access window, and the overscan vblank boundary —

tests/roms/AccuracySNES/ERROR_CODES.md

@@ -469,6 +469,14 @@ Provenance: **Documented** (WDC 65C816 datasheet; 6502.org 65c816opcodes decimal
 | 3 | `$06` | SED: $79 + $79 did not give $58, so the BCD addition itself is wrong |
 | 4 | `$08` | N was set after a decimal $79 + $79 = $58, so the flags describe the binary sum $F2 |
 
+### A6.10 — Emulation COP vector
+
+Provenance: **Documented** (WDC 65C816 datasheet, vector table; SNESdev Wiki vectors). Kind: scored.
+
+| Code | Byte | Meaning |
+|---|---|---|
+| 1 | `$02` | COP in emulation mode did not vector through $FFF4 — the $B0 marker means it took the native table's $FFE4 instead, and $00 means it reached neither handler |
+
 ## Group C
 
 ### C1.01 — OAM word write/read