Player effects also leave out ambient false and show_particles true

MobEffectInstance.Details.MAP_CODEC uses optionalFieldOf("ambient", false)
and ("show_particles", true), so like amplifier 0 those defaults are not
written (the mob writer already does this). The harness gets a default
work dir, so tools/parity.py can run it with no arguments; the run-log
doc is dropped.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

crates/horizon-client/src/save.rs

@@ -4222,11 +4222,16 @@ impl Level {
                                     Tag::Byte((*amplifier).clamp(0, 255) as u8 as i8),
                                 ));
                             }
-                            effect.extend([
-                                ("ambient", Tag::Byte(i8::from(*ambient))),
-                                ("show_particles", Tag::Byte(i8::from(*particles))),
-                                ("show_icon", Tag::Byte(i8::from(*particles))),
-                            ]);
+                            // optionalFieldOf("ambient", false) and
+                            // ("show_particles", true): defaults are not written;
+                            // show_icon is always written.
+                            if *ambient {
+                                effect.push(("ambient", Tag::Byte(1)));
+                            }
+                            if !*particles {
+                                effect.push(("show_particles", Tag::Byte(0)));
+                            }
+                            effect.push(("show_icon", Tag::Byte(i8::from(*particles))));
                             Tag::compound(effect)
                         })
                         .collect(),
@@ -7200,11 +7205,12 @@ mod tests {
             for (entry, (id, duration, _, ambient, particles)) in entries.iter().zip(&effects) {
                 assert_eq!(entry.get("id"), Some(&string(id)));
                 assert_eq!(entry.get("duration"), Some(&Tag::Int(*duration)));
-                assert_eq!(entry.get("ambient"), Some(&Tag::Byte(i8::from(*ambient))));
-                assert_eq!(
-                    entry.get("show_particles"),
-                    Some(&Tag::Byte(i8::from(*particles)))
-                );
+                // MobEffectInstance.Details.MAP_CODEC leaves out ambient false
+                // and show_particles true.
+                let ambient_tag = ambient.then_some(Tag::Byte(1));
+                assert_eq!(entry.get("ambient"), ambient_tag.as_ref());
+                let particles_tag = (!*particles).then_some(Tag::Byte(0));
+                assert_eq!(entry.get("show_particles"), particles_tag.as_ref());
                 assert_eq!(
                     entry.get("show_icon"),
                     Some(&Tag::Byte(i8::from(*particles)))

docs/parity/player-effect-amplifier-default.md

@@ -1,88 +0,0 @@
-# Player saved effect amplifier default
-
-H100 changes only `save.rs::Level::player_tag`: omit an amplifier whose modeled
-value is zero. The existing clamp to 0..255 and unsigned Byte conversion stay
-unchanged for other values. The loader, live effects, other effect fields and
-separate Mob writer are untouched.
-
-Pinned Java 26.3 `MobEffectInstance.Details` builds its map codec with
-`ExtraCodecs.UNSIGNED_BYTE.optionalFieldOf("amplifier", 0)`. PR6369 changed only
-the Mob writer. PR9061 supplied the Player unsigned-byte reload/clamp behavior,
-which this fix preserves.
-
-## Actual Game proof
-
-Frozen base: `0d21d5520ec26c49b774bf1533e53ea8af81278d`.
-Script: `tools/test_player_effect_amplifier_parity.py`, SHA-256
-`2b2023bb710b24dac97953ee87765a9c1585c3c90ce4842cb3f0eed2e023605c`.
-Fresh Java reference SHA-256:
-`b4a8db3e861ac399f61093270580a5e2917882a40fa2ce9c1ca615763011f805`.
-
-Real Player commands apply/clear an infinite Luck effect at amplifiers 0/1/255.
-Saved reads use the existing `/data get entity @s` route: Game synchronizes the
-live player and calls `Level::player_tag`, the same Player saved encoder used by
-world/player file writes. No observer or packet implementation was added. This
-proves that saved-NBT projection, without a save/reopen or disk-I/O claim.
-
-The predeclared 71 command records compare as complete raw JSON, including all
-messages/styles and callbacks, without masks/tolerances. UUID/type/Health20,
-position/mode, load/floor/air, exact effect id/duration/type, one-effect count,
-particle/icon flags, effective nonambient state, real effect removal and zero
-relative world-clock controls stay strict. Each amplifier read has fresh
-result/success sentinels. Whole effect/body compounds and absolute startup
-clocks are separate complete raw traces, not whole-record equality claims.
-
-Untouched Game matches 68/71; fixed Game matches 71/71 with identical script
-and fresh Java reference bytes. Exactly three old records differ for amplifier 0:
-
-| Record | Fresh Java | Untouched Game |
-| --- | --- | --- |
-| Saved amplifier read | Missing field; callback`[false,0]` | `0b`; callback`[true,0]` |
-| Success store | `0` | `1` |
-| Field-absence predicate | Test passed;`[true,1]` | Test failed, count1;`[false,0]` |
-
-The positive 1 and 255 controls pass; 255 saves as typed Byte -1. Live effect
-presence is independently witnessed by successful real effect removal, rather
-than inferred from the saved amplifier's absence.
-
-The resource-free save regression starts both with no raw input and with stale
-raw effects. It checks 0 omission and Byte 1/-128/-1, unchanged id/duration/flags,
-unrelated raw data and unchanged live model. On the old writer it fails at
-`Some(Byte(0))` versus`None`; the existing 255-reload and ended-effect controls pass.
-
-Source checkpoint `932d351e115e9a3e31f29eaaba0614004bbb3621` adds unrelated
-region compression selection/tests in `save.rs`; the Player writer and command
-read path are unchanged. The paired runtime proof remains qualified to 0d21.
-
-## Retained whole-effect diagnostic
-
-The original 68-record program/reference and complete old output remain local
-and executable. It matches 62/68: the three H100 records above plus
-`amp0-count`, `amp1-count`, `amp255-count`, which read whole effect lists.
-Those lists also contain native `ambient:0b`, whereas Java omits the default.
-The 71-record program replaces those three compound comparisons with exact
-first-entry/absent-second-entry controls. It preserves their full raw reads as
-traces. No ambient/default-flag repair belongs to H100; no whole-diagnostic equality claim. The fixed diagnostic matches 65/68; every
-remaining exact saved-list value difference is below (all three callbacks
-remain `[true,1]`).
-
-
-| Label | Java saved list | Fixed Game saved list |
-| --- | --- | --- |
-| `amp0-count` | `[{duration: -1, show_icon: 0b, id: "minecraft:luck", show_particles: 0b}]` | `[{duration: -1, show_icon: 0b, ambient: 0b, id: "minecraft:luck", show_particles: 0b}]` |
-| `amp1-count` | `[{duration: -1, show_icon: 0b, amplifier: 1b, id: "minecraft:luck", show_particles: 0b}]` | `[{duration: -1, show_icon: 0b, ambient: 0b, amplifier: 1b, id: "minecraft:luck", show_particles: 0b}]` |
-| `amp255-count` | `[{duration: -1, show_icon: 0b, amplifier: -1b, id: "minecraft:luck", show_particles: 0b}]` | `[{duration: -1, show_icon: 0b, ambient: 0b, amplifier: -1b, id: "minecraft:luck", show_particles: 0b}]` |
-
-Run in a fresh destination using the existing command harness:
-
-```sh
-python3 tools/test_player_effect_amplifier_parity.py --reference-only \
-  --work-dir .local/player-effect-reference
-python3 tools/test_player_effect_amplifier_parity.py --binary /path/to/horizon-client \
-  --work-dir .local/player-effect-native --reference .local/player-effect-reference/vanilla.jsonl
-```
-
-The native invocation independently reruns Java and requires the exact same
-reference bytes. All commands, original diagnostic scripts, raw records and
-traces are retained locally, including the earlier 68-record compound admission
-failure. The focused 71-record proof independently admits every actual subject/state.

tools/test_player_effect_amplifier_parity.py

@@ -85,7 +85,11 @@ def main():
         i = sys.argv.index('--reference'); reference = Path(sys.argv[i+1]).resolve()
         del sys.argv[i:i+2]
     if '--work-dir' not in sys.argv:
-        raise SystemExit('a fresh --work-dir is required')
+        # tools/parity.py runs every harness with no arguments: a fresh folder under .local.
+        import shutil
+        default = Path(__file__).resolve().parents[1] / '.local' / 'player-effect-amplifier-parity'
+        shutil.rmtree(default, ignore_errors=True)
+        sys.argv += ['--work-dir', str(default)]
     build = Path(sys.argv[sys.argv.index('--work-dir')+1]).resolve()
     assert not build.exists(), 'preserve earlier attempts; use a fresh directory'
     build.mkdir(parents=True)