Commit Graph
9 Commits
Author SHA1 Message Date
Sam SungandClaude Opus 5.5 ce9d3f23cb The ground tilt bounds its duration with the reference's two compares, so a cameraTerrainTiltTimeMin above cameraTerrainTiltTimeMax, or a NaN, no longer panics the client (#459)
With Follow Terrain on, a `cameraTerrainTiltTimeMin` above `cameraTerrainTiltTimeMax`, or a NaN in either, no longer panics the client on the first slope. The ground channel bounded its tween duration with `f32::clamp(lo, hi)`, which panics when `lo > hi` or a bound is NaN.

- **The bound** is now the reference's two ordered compares (`0x50dd29`-`0x50dd83`, `lo = TimeMin × factor`, `hi = TimeMax × factor`): below `lo` takes `lo`, else at or above `hi`, or NaN, takes `hi`, else the duration stands. Nothing orders the pair, so an inverted one is an odd duration and never a fault. `lo` is compared unrounded, as it stays on the x87 stack (`0x50dd4c`), and `hi` as the f32 it is stored as (`0x50dd40`).
- **In flight too.** The armer `0x512490` returns 1 when the request matches the one in flight (`0x512555`), and the caller bounds the duration on any nonzero return (`0x50dd21`), so the bound runs again every frame a tilt is in flight. For an ordered pair that changes nothing; an inverted pair leans for `TimeMax × factor`, and a TimeMax changed mid-tilt bounds the tilt in flight.
- **The floor** that keeps a zero or NaN duration from stalling the tween moved from the stored duration to the divide, so the in-flight bound reads the raw value the reference reads.

Checked: nine tests, each derived from the reference's rule rather than the code, covering an inverted pair below, between and above, a NaN in each input, the unrounded `lo` compare, the in-flight bound, and the issue's case end to end (Min 20, Max 1 on a 20° slope). Each fails with its piece undone, the first ones with the old `clamp` panic (`min > max, or either was NaN`). The bound matches the reference's bytes over 20M random inputs with no bit difference. Gates green on macOS; no platform seam touched.

Fixes #304

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-27 13:39:41 +07:00
Sam SungandClaude Opus 5.5 9c9badf131 Comments in benilla-app, the player, entities, animation and targeting: one line of fact each, no history, no record numbers (#328)
Part of #41: the third of five `benilla-app` pieces. This one is the player, the entities, animation and targeting: the mover, the camera, movement on the wire, the body and creature animation drivers, spell visuals, equipment and corpses, and the target, cursor and by-name layers. Their comments are rewritten to the rule in `docs/METHOD.md`, which is what the code does and the reference fact behind it, in a line. 26,066 comment lines become 9,042.

Fixes #104. Fixes #108. Fixes #109.

- **The mechanical strip.** The citations that were only a private record number go first, and each edit is shown to be a pure removal.
- **Wrong facts corrected.** The rewrite corrects the comments that stated wrong facts. Among them:
  - The follow gate's `0x606ba0` is `CanCooperate` (a faction-group-mask test), not `CanAssist` (`0x6066f0`).
  - The facing report is sent from the move-command drain (`0x615fc6`); `0x617100`/`0x617170` are the transport-GUID heartbeat.
  - vmangos turns a corpse into bones under the corpse's own guid (`Map.cpp:3608-3649`), and the bones get new flags and zeroed item slots.
  - Two claims about the item models are refuted by the shipped data: not every helm stem ships all sixteen race and sex files, and some item emitters sit under a keyed billboard chain.
  - Docs sitting on the wrong item, stale function names, and disassembler labels, which become client addresses.
  - vmangos line citations are checked against the pinned vmangos revision.
- **Deviation labels.** "Deviation:" now marks only a choice with a reason. An unbuilt behaviour, or one whose reference side is untraced, says so plainly.
- **Verification.** Every file strips to identical code, and `cargo fmt --check` is clean. Six second readers compared every file old against new, restored the facts the rewrite had dropped, and checked each new citation against the code, the shipped data, vmangos and the reference.

Behaviour differences the rewrite surfaced are filed as their own issues (#293, #294, #301, #308, #316 and #317 among them). The comments here state what the code does. String literals are untouched; they are #92.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-25 19:06:42 +07:00
Sam SungandClaude Opus 5.5 7b66dac9db comments cite the reference by address, not by private note (#30)
Comments that pointed into the maintainer's private reverse-engineering notes (a note by name, a
path, a section or finding id, the analysis's own name for a client function, a process word) now
carry the fact itself, cited by the reference client's address, by vmangos file and line, or as
"the reference". Nobody outside could follow those pointers.

Comment text only, plus the test-assert and log strings that carried the same pointers. No claim
changes: a comment the notes have since corrected keeps its wording here, and correcting it is
separate work.

The two reference generators, `scripts/gen-reference-globals.py` and
`scripts/gen-reference-verb-events.py`, named the private notes' layout to find their inputs. They
now take the input files as arguments (`--capture`, `--framexml`, `--names`, `--extents`,
`--disasm`). Run on the same inputs, both write the committed tables row for row; only the two
header lines that named the old environment variable change.

Checked per file:
- the code is identical: each file with its comments stripped compares equal before and after; a
  changed string literal is compared under a mask that proves only literal text moved and every
  literal kept its line count;
- no client address the old text cited was dropped;
- a second reviewer checked every added address against its derivation;
- a detector for the notes' names, paths, ids, function names and process words finds none left;
- `scripts/gates.sh` is green.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-09-24 12:14:08 +07:00
SamandClaude Fable 5.1 aa207b4050 settings live in one registry, and the net drain starts to peel
- the CVars are one engine-side table that outlives every interface reload: a change fires its observer, the eight latched rows wait for RestartGx, and /console gains set, cvar_reset, cvar_default, cvarlist and help (2303)
- every push into the interface runs in one feed phase after the net drain and before the tick, held by a structural test on the built schedule (2304)
- subsystems own their packet handlers, dispatched in packet order; the auction house is the first family out of the 273-arm drain (2305)

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-09-17 16:47:09 +07:00
SamandClaude Fable 5.1 9e528e6f86 the video options window goes stock, and water collision returns whole
- the reference's own video options window loads hidden, so pfUI's video skin and any addon reaching for its sliders and check buttons by name finds real frames; our own options window is BenillaOptionsFrame now, and ESC still opens it (2177)
- GetScreenResolutions, GetCurrentResolution, SetScreenResolution, GetRefreshRates, GetVideoCaps, RestartGx and RestoreVideoDefaults answer off the monitor's own modes, so CT_Viewport loads clean (2177)
- three of those verbs were corrected against a second read of the binary: GetRefreshRates answers nothing on a device with no rates, GetVideoCaps always gives a number for triple buffering, and RestoreVideoDefaults writes the ten rows the verb writes (2177)
- the camera's water collision is back with both halves: the boom stops at a water surface, and a swimmer's camera keeps a constant 2/9 yd clearance over the waterline whatever the body underneath does (2179)

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-09-11 21:04:35 +07:00
SamandClaude Fable 5.1 a61b9159ac plate text settles, mobs stop sinking, and select goes with Lua 5.1
- a nameplate's name and level stop popping a pixel every other step as the plate slides, and the whole plate no longer jumps 19 px to the top of the screen on a rounding tie; the chat bubble takes the same fix (2172)
- creatures and other players stop sinking through the ground while the world is still streaming in, where a Stormwind patroller had been 31 yd under its street: a body our world cannot yet answer for holds the server's pose (2174)
- select and the string metatable are gone, as in 1.12's Lua 5.0: pfUI's heal prediction parses its messages with the right format, BuffCheck2's own select wins its fallback, and a method call on a string raises the reference's own error (2171)
- the camera's water collision comes back out for now: with the pivot a hair under the plane, a surface swimmer's camera snapped in and out against the waterline, so the boom ignores water again (2173)

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-09-11 21:03:05 +07:00
SamandClaude Fable 5.1 64296b306f dodge, parry and block answer, and the camera boom stops at the water
- GetDodgeChance, GetParryChance and GetBlockChance answer the numbers the character sheet already had, so the addons that ask for them get a figure instead of a nil (2166)
- with Follow Terrain on, pressing mouse-look no longer jolts the view: the terrain lean folds into the angle you drag and comes back out on release, and smart pivot holds its bias while a server-driven turn runs (2165)
- entering water, the camera's pivot eases down to the model's own swimming height, so the eye no longer sits at a different height for the same zoom (2165)
- the camera's water collision is back as what the option actually is: with it on, the camera boom stops at a bare water surface instead of passing through it; the pivot corridor that snapped last time is not rebuilt (2170)

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-09-11 21:02:31 +07:00
SamandClaude Fable 5.1 3b084ba052 pfUI's hdgraphic runs, the plates settle, and the water camera comes out
- pfUI's hdgraphic module runs: SetWorldDetail and GetWorldDetail are real verbs with the reference's three stops, and the near clip plane is a live setting at 0.100 yd rather than a frozen 0.111 (2163)
- nameplates stop drawing a black box on hover or click, stop jittering a frame behind a moving creature, and a unit that walks into range later gets its plate; a unit with no level wears no skull, and the wheel zooms over a plate (2168)
- the faint one-pixel box around every letter of a magnified overhead name is gone, each glyph now carrying its own transparent padding; interface text drawn at 1:1 is byte-identical (2169)
- the camera's water collision comes back out for now: its pivot corridor dropped the camera a yard in one frame at head depth, which is the snapping reported around the surface while swimming (2165)

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-09-11 21:02:02 +07:00
SamandClaude Fable 5.1 38e45badb9 nameplates are frames, guild names go overhead, and the camera pivots
- a nameplate is a real frame under WorldFrame, six regions and a health bar in the reference's order, so pfUI, ShaguTweaks and _Nameplates find and restyle plates instead of seeing none; the plate stays out of the mouse's way in this cut (2148)
- a player's guild shows under their overhead name in angle brackets, the line an NPC's subname already used, gated by the UnitNamePlayerGuild option at its stock on (2149)
- smart camera pivot is on, as 1.12 ships it: a mostly vertical drag with the camera pressed against something tilts the view instead of swinging the arm, easing home on release; in water the pivot rises for a swimmer (2149)
- Follow Terrain and head bob are built and ship off, as in 1.12: Follow Terrain leans the camera up or down the slope you are walking onto, and head bob sways the first-person eye in a figure of eight while you move (2149)

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-09-11 20:55:15 +07:00