ZInflate (z_inflate_new) holds the bit reader, the RFC tables and every table and scratch list a block
builds, rebuilt in place: an inflate allocates nothing, and a worker thread inflates in a context of its
own (made on the program's thread). z_inflate_in / z_uncompress_in / z_gunzip_in take it; z_inflate /
z_uncompress / z_gunzip and every caller (image, atlas, tiled, render3d's png_decode) are unchanged and
work in RtInflateState's one context. The old per-call lit/dist tables were also leaked on an error
return; there are none now. Compiles: Maroon Lake headless, examples/library/tiled_p2 and tiled_p6.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The typed buffers are slices: words/floats/fixeds/doubles/pointers(n) make
zeroed, bounds-checked []int/[]float/... and the type names mean them. buffer(n)
is a []byte, with text_of, Fs.read_bytes/write_bytes and view(xs, start, n).
bytes(), indexing a raw pointer or bytes, free, resize, Memory.*, raw file calls,
data_of and C externs are refused outside unsafe { } / unsafe function, and a
project's own files may write unsafe only with --unsafe; the runtime and packages
are the platform. A slice passed to an extern goes as its data.
What the change found: Sync's atomics on a slice header, words(n) uninitialised,
input's fixed axes in ints, truetype's fixed outlines as ints, skin matrices
typed int, gl_shader's source table made from raw bytes. render3d gets safe
entry points (safe_api.ludic). Rendering is byte-identical; a frame costs the same.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`Gl.*` binds the whole OpenGL 4.1 core API — every entry point of the
platform gl3.h with every GL_* constant, generated by `ludic-dev glgen`
with per-call ABI thunks. Windowed builds get an NSOpenGLContext on the
existing window at Retina resolution; headless builds render into an
offscreen CGL context, so a program that uses Gl.* renders and
screenshots identically under the test harness. It links gl.ll, the
thunks and OpenGL.framework only when used; every other build stays
byte-identical.
packages/ludic.render3d is a physically based renderer written on that
surface: HDRI image-based lighting, GPU-generated terrain with scanned
PBR materials, CDLOD, cascaded shadows, glTF with skinning, instanced
vegetation with impostors, procedural grass, water, SSAO, and an HDR
pipeline with bloom, auto-exposure and ACES.
It also carries this session's work on it: the terrain at half its cost
(10.3 -> 5.4 ms of frame), the streaming hitch that got worse the longer
you played, a resize that emptied the world, and the packaging that lets
a game use the renderer from its own repository — `ludic assets`, the
material manifest shipping with the package, and shader lookup falling
back to the install root. See changes/ for each, with its numbers.
The camping game that drove all of it has moved out to its own
repository, Maroon Lake; examples/rendering/smooth.ludic stays as the
renderer's example here.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
- runtime: HEADLESS_FRAME_PATH, STICK_DEADZONE / STICK_LEFT_X/Y, key and
byte codes as char literals throughout (`k == 'w'`, `fill(rt_map, ' ', …)`)
- ludic.gameplay/stats: drop the duplicate `stat_field` (it answered "atk"
for every build stat); Stats.base uses stats_field_name
- ludic.shooter: compare aim modes and fire patterns with AimMode.* and
WeaponPattern.* instead of raw ints; STICK_RIGHT_X/Y
- ludic.npcai: DecisionMade / brain_set_state use AiState.*
- examples/games/menu.ludic uses Font.load / Ui.* with FONT_PATH and
BACKDROP named; strings.ludic header says what it prints
- whole tree: `x = x + 1` → `x += 1` (single-term right-hand sides only),
`0 - x` → `-x`, ASCII codes → char literals; every .ludic and every
```ludic fence reformatted with the fixed formatter (whitespace only)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The three parsing primitives the TMX path needs that were not already in
the tree (zlib inflate + the JSON reader already shipped):
- Xml.* — a minimal, deterministic pure-Ludic XML reader for the
element/attribute/CDATA subset TMX/TSX/TX use, spliced on demand. Handles
nested elements, single/double-quoted attributes, text + <![CDATA[…]]>,
comments, the <?xml?> prolog and <!DOCTYPE>, the five predefined entities
and numeric character references.
- Base64.* — standard base64 (RFC 4648) decode + a matching pure-Ludic
encoder; the decoder ignores the whitespace Tiled wraps into <data>.
Splicing Base64.* also pulls in inflate.ludic so a plain tool can run the
full base64 -> zlib/gzip decode chain.
- z_gunzip — gzip framing (RFC 1952) over the existing DEFLATE inflater:
skip the 10-byte header + optional fields, inflate the body, ignore the
CRC32/ISIZE trailer.
Golden corpus vendored under assets/tiled-fixtures/ (mapeditor/tiled
examples, attributed, + hand-authored multi-encoding fixtures). New
example library/tiled_p0.ludic proves all three (21 assertions); docs
pages + inventory added so check-impl stays green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Expand the abbreviated pointer types to full words on the language surface:
ptr -> pointer (a raw address / FFI handle)
ptrs -> pointers (a buffer of pointers)
The Ludic type name is distinct from LLVM's own `ptr` spelling: llty() maps
`pointer`/`pointers` to LLVM `ptr`, and the emitted IR keeps `ptr`, so only
the Ludic-level surface changes. Rewrites type annotations across all
sources, the 8 hardcoded pointer type-tags, the `pointers`-buffer indexing
in emit_addr, the grammars/LSP/JetBrains tokens, and the docs
(type-ptr -> type-pointer, type-ptrs -> type-pointers). int/bool keep their
conventional short spelling (like Math).
Reseeded; C-free fixpoint holds; all suites green (45/24/29); site + check.py OK.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Expand the function-declaration keyword to the full word across the whole
language and toolchain:
fn name(...) -> T { ... } -> function name(...) -> T { ... }
Done as a self-hosting migration: teach the parser both spellings, reseed,
rewrite every .ludic definition to `function`, then drop `fn`. The compiler
now rejects `fn`. Touches the parser, all selfhost/tools/runtime/example/test
sources, the grammars (TextMate shared+vscode, ludic_syntax.h, JetBrains
LudicTokens.kt), the LSP and formatter, the Python doc/vocab tools
(check-impl, check-docs, validate, palette, test-lsp), and the docs
(fences, prose, kw-fn -> kw-function).
Reseeded; C-free bootstrap fixpoint holds. All suites green (45 regression,
24 self-host, 29 tool); the docs site generates and check.py passes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Groups B and C of the leftover-primitive cleanup — renames, not new machinery,
and deliberately NO unsafe_ prefix (a __-prefix is itself a C convention, and an
`unsafe` marker carries no signal in a fully-manual-memory language with no safe
subset to contrast against).
memory: mem_free -> free mem_realloc -> resize mem_set -> fill
ptr_add -> offset
process: os_argc -> arg_count os_arg -> arg os_exit -> exit
os_system -> run os_getenv -> getenv read_byte -> read_char
dead: mem_copy, os_time, write_byte (0 uses) — deleted
Two reseeds: accept both old and new names in the intrinsic dispatch, then
migrate every call site and drop the old names. file_open/read/write/seek/tell/
close are left as-is — they're the domain-prefixed syscall layer wrapped by
read_file, not the argc/argv-style C-ness the audit targeted; a `File` type is a
separate, larger design if wanted.
test.sh's CLI smoke updated (os_exit -> exit); check-vocabulary's grammar marker
moved off the deleted names. Reseeded (22243 lines); C-free fixpoint holds;
goldens identical; 18/18; vocab + doc-fences clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Allocation reads as intent, not malloc:
mem_alloc(64) -> bytes(64) (64 bytes -> a byte buffer)
mem_alloc(w * h * 4) -> words(w * h) (w*h 32-bit words)
bytes(n) mallocs n bytes and returns a plain pointer (byte-indexed); words(n)
mallocs n*4 bytes and returns a `words` pointer (int-indexed). Since words(X) and
mem_alloc(X*4) allocate the identical number of bytes, the migration cannot change
any allocation size — the `* 4` factor just moves from the argument into the
allocator name, pairing naturally with the Phase-7j `words` retyping
(`var fb: words = words(w * h)`).
Migrated 102 sites (mem_alloc(E*4) -> words(E), else bytes(E)); deleted the
mem_alloc intrinsic. mem_realloc/free/copy/set stay as the low-level
reallocation/free family. Reseeded (22565 lines); C-free fixpoint holds; goldens
byte-identical; 18/18; vocab + doc-fences clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
32-bit word access is indexing now, not peek32/poke32:
peek32(ui_rx, i) -> ui_rx[i] (reads an int)
poke32(rt_fb, i, c) -> rt_fb[i] = c (writes an int)
A buffer typed `words` (a pointer whose elements are i32) indexes with `w[i]`
as a full int; a plain `ptr`/`str` keeps byte indexing. emit_index_addr picks the
element type from the base's type — byte-identical IR to the old intrinsics, so
the migration reproduces the compiler and every golden render exactly.
The ~60 word buffers (rt_fb, tt_*/gc_* font tables, png_px/spr_px pixels, ui_*
layout arrays) were retyped from `ptr` to `words` scope-aware (per-function, so
the s/out/p byte-vs-word name collisions across functions stay correct), then the
254 peek32/poke32 sites migrated to indexing. A scope-analysis miss left 12
buffers (gc_*, sc_d, ui_rx/ui_ry) un-retyped — caught as a menu golden diff and
fixed. No true mixed byte+word access exists on any one variable, so a per-buffer
element type is sound.
Reseeded (22527 lines); C-free fixpoint holds; goldens byte-identical; 18/18;
vocab (byte/words types in, peek32/poke32 out) + doc-fences clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Raw byte access is now indexing, not C-style peek/poke:
peek8(src, i) -> src[i] (reads a byte, widened to int)
poke8(out, j, r) -> out[j] = r (narrows the int to a byte)
E_INDEX on a non-slice pointer/string lowers to a `getelementptr i8` + load/zext
(read) or trunc/store (write) — byte-identical to the old peek8/poke8, so the
migration reproduces the compiler exactly. A "byte" element type (llty i8) drives
the widen/narrow. The compiler's own byte work now reads naturally, e.g.
`is_slice_ty` is `t[0] == 91 and t[1] == 93`.
Subtlety fixed on the way: g_addr_ty (the out-param carrying the indexed element
type) must be set AFTER evaluating the index expression, since a member/index in
the index would otherwise clobber it — doing it early made a byte read load a
full pointer from a byte address and crash the self-compile.
Two reseeds: add byte-index support keeping peek8/poke8, then migrate 148 call
sites and delete the intrinsics (+ the now-dead emit_gep_i8). Vocabulary drops
peek8/poke8. Reseeded (22604 lines); C-free fixpoint holds; goldens identical;
18/18; vocab clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
`null` is now a real pointer literal and null-tests are comparisons, instead of
`ptr_null()` and `ptr_is_null(x)`:
ptr_null() -> null
ptr_is_null(x) -> (x == null)
not ptr_is_null(x) -> (x != null)
Mechanics: a new E_NULL primary (`null`, like true/false) lowers to the `null`
pointer; emit_bin's comparison path now picks `ptr` vs `i32` from operand type
(via llty), so `==`/`!=` work on any pointer/record/slice. The two intrinsics are
deleted.
Two reseeds: (A) add the literal + ptr comparison keeping the intrinsics; (B)
migrate all 182 call sites (compiler + runtime, via a balanced-paren script that
skips string-literal args and rewrites `not ptr_is_null` to `!= null`) and delete
the intrinsics. Node/Val/Buf/Tok field defaults now read `ptr = null`.
Vocabulary drops the two from LUDIC_INTRINSICS; `null` joins true/false as a
language constant (grammar + ludic_syntax.h). LANGUAGE.md notes the literal.
Reseeded (21664 lines); C-free fixpoint holds; goldens identical; 17/17; vocab +
doc-fences clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
First step of the "stop feeling like C" pass. Bitwise ops were functions
(`band(x, MASK)`, `shl(a, 3)`); they are now real operators:
band -> & bor -> | bxor -> ^ shl -> << shr -> >> bnot -> ~
Precedence is Go-style so the C footgun is gone: `<<`/`>>`/`&` bind like `*`,
`|`/`^` like `+`, both tighter than comparison — `flags & MASK == 0` parses as
`(flags & MASK) == 0`. `>>` is logical (lshr), matching the old `shr`.
Mechanics: lexer tokenizes `<< >> & | ^ ~`; p_mul takes `<< >> &`, p_add takes
`| ^`, p_unary takes `~`; emit_bin routes them through the existing int arith
path (arith_code gains and/or/xor/shl/lshr) and E_UN handles `~`. The six
intrinsics are deleted.
Delivered as two reseeds: (A) add the operators keeping the intrinsics, (B)
migrate every call site to operator form (85 lines across compiler + runtime,
via a balanced-paren call->operator script; two multi-line big-endian reads in
image/truetype done by hand) and delete the intrinsics. Vocabulary drops the six
from LUDIC_INTRINSICS (ludic_syntax.h, grammar, LudicTokens.kt) and the grammar
gains a bitwise-operator rule. LANGUAGE.md precedence table rewritten.
Reseeded (21724 lines); C-free fixpoint holds; goldens byte-identical (the PNG
and TrueType decoders lean on these ops); 17/17; vocab + doc-fences clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bindings now signal mutability the way Rust/Swift do, instead of `let` meaning
"local" and `var` meaning "module-level":
- `let x = e` -> immutable binding; a later `x = …` is a compile error
(`cannot assign to immutable 'x' … use var`).
- `var x = e` -> mutable binding, at local OR module scope (position decides
scope; the keyword decides mutability).
- `const` -> unchanged (compile-time).
Immutability is of the *binding*, not the object: `let n = new Node; n.kind = 1`
is fine (mutation through the reference); only rebinding `n` is rejected. The
check lives in emit_assign — a direct `name =` whose target is a `let` local
(loc_mut == 0) errors; field/index targets and `var`/param/loop bindings are
unaffected.
Delivered as three reseeds so the self-hosting compiler never had to compile
source its own rules would reject:
A) add `var` as a local statement + per-local mutability tracking (loc_mut),
no enforcement;
B) migrate every reassigned `let` -> `var` across the compiler, runtime and
examples (337 declarations), driven by a per-function, string/comment-aware
scan (binding targets only, never `x.f =` / `x[i] =`);
C) turn on the check. bootstrap-cfree (compiler vs its own source) and every
golden build (which splices the runtime) then proved zero reassigned `let`
was missed anywhere.
Also folded in: removed leftover debug instrumentation in block() (a `cur=` /
print_int(777…) trace on the separator-error path) and fixed parse.ludic's stale
header comment (no more `struct`). Reseeded (21711 lines); C-free fixpoint holds;
goldens identical; 17/17; vocab + doc-fences clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Self-hosted compiler (selfhost/*.ludic), runtime, examples, editor tooling,
and docs. Phase 1 of the syntax-redesign cohesion pass has landed:
edge-system fix, signature-query, when-alias, and the documentation truth-pass.
Suite green (14/14), C-free bootstrap fixpoint holds.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
